Physical AI Brief
Daily cross-source signals for the Physical AI supply chain — silicon photonics, CPO, VLA models, humanoid hardware, embodied AI. Three streams, one page, zero filler.
328 items today · 263 arxiv · 3 SEC 8-K · 62 humanoid · 0 CN photonics
01 ARXIV · PHYSICAL AI PAPERS
263 items- arxiv:2611.09894 · cs.CVVideo-Coder: Reinforcing Agentic Coding for Video GenerationChaoyang Wang, Sicheng Gao, Junan Hu, Kunyang Li +6
Recent years have witnessed remarkable progress in video generation in terms of visual quality and temporal consistency. However, generating videos with precise spatial structure, coherent motion dynamics, and controllable scene composition remains challenging. In this paper, we introduce Video-Coder, the first attempt to train a code-driven video generation agent, which autonomously writes and executes code to construct a dynamic 3D scene (Draft Video) that provides structure and motion signals to downstream video generators. To support this paradigm, we design a dedicated data pipeline and curate two high-quality datasets, Video-Coder-SFT-6k and Video-Coder-RL-2k, which contain diverse prompts and ground-truth reference videos. We further develop the Video- Coder Harness, a unified interactive environment integrating code execution, 3D scene rendering, and execution feedback for agent training and inference. We then train Video-Coder with SFT fol- lowed by harness-native agentic reinforcement learning with dual reward feedback, which combines execution-based and video-based rewards. Additionally, we design an annealing training strategy that gradually reduces the execution reward over time, allowing the model to rely more on the video reward in the later training stages. Experiments show that Video-Coder achieves significant gains, improving Minimax-H3 by approximately 7 points on PhyGenBench and 24 points on VBench-2.0. We hope this work can serve as an open foundation for coding agents in video generation, and we fully open-source our data, models, and code
agentagentic - arxiv:2611.09892 · cs.ROOn Agentic Manipulation PoliciesChen Chen, Omid Bagherieh, Naveen Nikamanth Appiah Balaji, Jesse T. Gonzalez +9
Recent work on robot manipulation has focused on specialist policies trained on demonstrations. We show that generalist AI models can control manipulators more reliably than specialist policies such as vision-language-action models (VLAs) and world-action models (WAMs). We systematically analyze agentic manipulation policies (AMPs), in which a generalist AI model commands the manipulator directly, adapting its actions as the task unfolds. We present AMP1, a reference harness that runs unchanged across single-arm, bimanual, and mobile manipulation. AMP1 achieves state-of-the-art results on the RoboLab-120 and CaP-Gym benchmarks, with no demonstrations or task-specific training. On RoboLab-120, AMP1 reaches an 85.6\% success rate zero-shot, compared to 49.2\% and 42.9\% for leading VLA and WAM policies. The agent succeeds zero-shot in 104 of 108 real-world trials spanning twelve tasks and three robot instances, reducing failure rates by 20x and 15x relative to strong VLA and WAM baselines. The basic harness supplies no grasping, pushing, or placing skills; the agent develops its own strategies, such as searching for hidden objects, using objects as tools, and calibrating its cameras. With the addition of a persistent memory store, the agent can also learn from experience: an augmented harness with a memory representation that is carried across episodes lowers the failure rate on RoboDojo bimanual manipulation tasks by 4.6x relative to zero-shot deployment. Our analysis suggests that AMPs are substantially more general and robust than specialist manipulation policies, providing a promising foundation for robot manipulation.
vision-language-actionvlamanipulationmanipulatorgraspmemory - arxiv:2611.09891 · cs.AICPU-ARC: Agentic Reasoning Caching for on-CPU Reasoning ReusePeter Belcak, Greg Heinrich, Saurav Muralidharan, Shizhe Diao +1
Large language models reinvent the wheel all the time. They conflate task-solving with the execution of the solutions, which forces them to solve every problem from scratch. We show that once a task has been reasoned through, the application of its solution to a new problem instance can be done very efficiently and with little involvement from language models. We introduce agentic reasoning caching, a method for constructing robust CPU-executable agentic reasoning caches (ARCs) from LLM reasoning traces. ARCs show on-par performance with frontier LLMs across a range of difficult mathematical and scientific tasks, are fully explainable, continually trainable, robust to input perturbations, and 16-291x cheaper to invoke than the source model.
agentic - arxiv:2611.09888 · cs.ROScenario Orchestration for Closed-loop Policy Evaluation of Autonomous DrivingFrieda Rong, Zhaoyue Wang, Veer Kapadia, Vishaka Rajani +5
Scenario-based simulation is central to evaluating autonomous-driving policies, but a scenario scripted before execution (i.e., with predetermined, fixed actor routes and speeds) may not preserve its intended interaction once the tested ego policy responds in closed loop. Scripted scenarios cannot adapt to unanticipated ego behavior, while learned traffic actors may remain reactive without reliably realizing a particular scenario intent. We introduce a scenario orchestrator that continuously monitors the joint traffic state and adapts non-ego actors through re-timing, re-routing, and role reassignment for creating a specific scenario intent. At each step, the orchestrator solves a constrained optimization problem that preserves the intended interaction while limiting interventions and maintaining traffic feasibility. We further distinguish scenario realization from policy failure, allowing a test to first be judged based on whether the intended interaction occurred and then by how the ego policy responded. Experiments in the CARLA simulator across multiple scenario families and driving policies show that orchestration realizes targeted interactions more reliably than scripted and learned actor-level baselines, particularly under variations in initial conditions and ego behavior. These results support closed-loop scenario orchestration as a tool for controlled and reproducible policy evaluation.
policy evaluation - arxiv:2611.09880 · cs.CVZeoBench: A Benchmark for Self-Supervised Learning on 3D Zeolite RepresentationsAaron Sun, Yachan Liu, Ping Yang, Peng Bai +1
Zeolites are an important class of porous crystalline materials with broad applications in gas storage, separations, and catalysis. Although hundreds of thousands of experimentally realized and hypothetical zeolite structures are known, identifying materials optimized for a target application remains challenging because property labels are expensive to obtain and available only for limited structure-molecule pairs. While prior work has demonstrated that 3D neural representations can predict zeolite properties effectively, it remains unclear which representation-learning strategies are most effective in the low-data regime relevant to materials discovery. We present ZeoBench, a benchmark for self-supervised learning on dense 3D representations of all-silica zeolites, and evaluate the resulting features on six existing downstream adsorption-property prediction tasks, across a range of training set sizes. Our study compares self-supervised 3D pretraining on volumetric zeolite data, adaptations of image-pretrained 2D ConvNets and vision transformers to 3D via multiview and channel-inflation strategies, and 3D ConvNets pretrained on out-of-domain volumetric datasets, alongside hand-crafted descriptors and supervised baselines. Using our pretraining and evaluation framework, we find self-supervised 3D representations outperform existing methods and achieve state-of-the-art performance across both in-domain label regimes and for the unseen domain of metal organic frameworks. Learned representations consistently outperform hand-crafted descriptors, and convolutional architectures are generally more effective than vision transformers in this setting. These results establish ZeoBench as a practical benchmark for 3D representation learning in zeolites and provide guidance for data-limited adsorption property prediction.
benchmarkevaluation framework - arxiv:2611.09877 · cs.ROLocating Contact on Tendon-Driven Continuum Robots From Tendon Measurements Alone via Constrained Moving Horizon EstimationHaitham El-Hussieny
Tendon-driven continuum robots can reach through constrained spaces, but contact along their bodies is invisible to the sensors at their motors. We show that this is a structural limitation rather than a tuning problem: the map from contact forces to tendon measurements has a null space of dimension at least seven, so no estimator that leaves the contact support free can locate contact. When the obstacle geometry is known, however, non-penetration and complementarity make the contact force locally identifiable, and we prove that the resulting system with unknown contact inputs is locally exponentially strongly detectable. We present a constrained moving horizon estimator (MHE) that encodes these conditions and runs in real time as one quadratic program per sample, and we give a local, mode-wise robust stability result for it. In a finite- element benchmark of a two-section robot pressing and sliding on a wall, the estimator located contact to 0.9 mm and estimated contact force to 0.22 N, compared with 65 mm and 0.72 N for a tuned extended Kalman filter (EKF) with state projection. The estimator tolerated millimeter- level errors in the obstacle position and unmodeled wall friction, under which the projected EKF failed. Under randomized model mismatch over 40 trials, including unmodeled tendon friction, it remained better in force and location in every trial. Without contact, the MHE offered no advantage over the EKF, which confirms that its benefit comes from the contact constraints. This report extends a letter submitted to IEEE Control Systems Letters with complete proofs and additional experiments.
benchmark - arxiv:2611.09872 · cs.AIOn the Reliability of Majority-Vote Pseudo-Labels in Test-Time Reinforcement LearningHaoyu Wang, Ru Wang, Yuanyuan Wang, Jiaxian Guo +1
Test-time reinforcement learning (TTRL) adapts pretrained language models to unlabeled target-domain data using self-generated pseudo-labels, whose reliability remains poorly understood. We analyze majority-vote supervision for TTRL with online Direct Preference Optimization (DPO) and decompose its error into population-majority bias and prompt-dependent finite-rollout variation. Because population-majority bias is generally unidentifiable without external supervision, we focus on mitigating the downstream effects of finite-rollout variation. We propose \emph{Prompt-Conditional Distributionally Robust DPO (PC-Dr.DPO)}, which conditions within-prompt robust reweighting on observed vote reliability to make updates from sampling-sensitive prompts more conservative. We further introduce a cache-based mechanism to reduce self-supervised forgetting during online adaptation. Across six mathematical and knowledge-intensive reasoning benchmarks and five model backbones, PC-Dr.DPO matches or outperforms the strongest TTRL baseline in 24 of 30 settings.
benchmark - arxiv:2611.09868 · cs.CLCodeQuant2: Enhanced Outlier Mitigation and Adaptive Clustering for Low-Precision LRMsXingyu Liu, Xiangyang Yin, Tianhua Xia, Haiyu Wang +1
Large reasoning models (LRMs) improve task performance by generating longer intermediate reasoning traces, but this substantially increases inference costs and makes low-precision quantization increasingly important. However, aggressively quantized models often perform poorly on reasoning tasks. We observe larger activation fluctuations during the first decoding steps of reasoning. Calibration that omits these early states may miss their activation fluctuations. To address these reasoning-specific quantization challenges, we propose \textbf{CodeQuant2}, a clustering-based LUT quantization framework for low-precision large reasoning models. CodeQuant2 mitigates activation outliers with rotation and scaling, uses clustering to represent non-uniform weight distributions, and adapts codebook sizes to the quantization difficulty of each layer. Guided by our observation of reasoning-time activation behavior, CodeQuant2 further improves calibration sample selection to better match actual decoding statistics. Combined with our customized LUT CUDA kernel, CodeQuant2 delivers substantial accuracy improvements on reasoning benchmarks while achieving up to $2.85\times$ speedup over BF16.
benchmark - arxiv:2611.09866 · cs.CVGRAC: Graph-Based Retrieval-Augmented Captioning for Surgical Scene DescriptionRochak Dhakal, Chen Zhao, Soup Cobb, Xin Qian +2
Purpose: Surgical operative reports support procedural documentation, surgical training, performance assessment, and workflow review. However, extracting meaningful information often requires extensive manual review and clinical expertise. Automated surgical scene descriptions could reduce this burden by leveraging recurring surgical activities and prior knowledge. Methods: We propose GRAC (Graph-Based Retrieval-Augmented Captioning), a framework that represents surgical scenes as graphs capturing relationships among instruments, actions, and anatomical structures. Given an input scene graph, GRAC learns to retrieve relevant graph-caption pairs from training data. It compares instrument-tissue interactions between retrieved examples and the input scene, using comparison features to condition retrieved-caption representations during generation. Results: On the EndoVis-18 benchmark, GRAC achieved a CIDEr-D of 5.88, compared with 4.23 for SGT and 3.36 for the strongest conventional Transformer baseline. Exact instrument-action-target triplet-set accuracy reached 52.6%, compared with 32.7% for SGT and 26.2% for the strongest conventional baseline. These results demonstrate improvements in language quality and factual agreement. On TORS, GRAC achieved a CIDEr-D of 5.06, compared with 3.80 for the strongest baseline. Conclusion: GRAC demonstrates that combining structured surgical scene representations with retrieved examples improves surgical scene captioning and produces more contextually meaningful descriptions.
retrieval-augmentedscene graphbenchmark - arxiv:2611.09864 · cs.ROGrasp-TAGS: Task-Aware Grasp Selection via Critic Distillation for Prehensile Loco-ManipulationXun Tu, Minghao Zou, Karthik Desingh
Loco-manipulation, which requires mobile manipulators to coordinate locomotion and manipulation over long horizons, is a key capability for large-object rearrangement in applications such as service robotics and warehouse automation. Despite recent progress, reliably executing these tasks remains challenging because performance depends on complex robot-object and object-environment interaction dynamics. Existing approaches address this challenge either through model-based methods, which often require accurate dynamics models, or through learning-based methods, which reduce this requirement but typically assume a fixed or predefined initial grasp. However, different geometrically stable grasps can lead to substantially different downstream manipulation performance. Therefore, we propose Grasp-TAGS, a task-aware grasp selection framework for goal-conditioned large-object rearrangement. We train a unified PPO-based reinforcement learning policy across multiple initial grasp configurations and distill the learned critic into a lightweight grasp utility network that ranks candidate grasps using only the candidate grasp configuration and goal information. At inference time, we select the grasp with the highest predicted downstream utility for the downstream policy to complete the task, before even making a contact. Experiments demonstrate that Grasp-TAGS improves rearrangement success over both fixed single-grasp policies and random grasp selection, achieving an average success rate of 89.84% compared with 79.87% for the best-performing single-grasp policies, an improvement of 9.97 percentage points. Grasp-TAGS also reduces final orientation errors across all evaluated objects. For more qualitative results, refer to the supplementary webpage: https://rpm-lab-umn.github.io/Grasp-TAGS-web-page/
manipulationmanipulatorgrasp - arxiv:2611.09863 · cs.LGFlow-Guided Target-Space Alignment via Path ConsistencyRuizhi Yuan, Zeqiu Yu, Wei Gao, Wei Chen +2
Many multimodal alignment problems require mapping a source modality into a fixed space defined by a target modality, where the target space is not merely an intermediate embedding but is used directly for downstream analysis. In this setting, successful alignment should preserve both the global structure of the target distribution and the instance-level correspondence observed in paired source--target training data. Existing objectives typically address only part of this goal: contrastive losses encourage relative similarity, distribution-matching losses improve marginal agreement, and pointwise regression fits each source sample to its paired target without explicitly modeling target-space path geometry. We propose FlowAlign, a flow-guided target-space alignment method that repurposes conditional flow matching as a training-time alignment signal rather than an inference-time transport model. FlowAlign uses the learned flow field to guide the alignment of source-side representations with the fixed target space, encouraging them to preserve paired correspondence while remaining compatible with the global target-space structure. We further provide theoretical analysis showing how FlowAlign helps improve paired correspondence by controlling endpoint accuracy and capturing local target-space geometry. Experiments on synthetic data, paired single-cell multiome benchmarks, and image--text alignment demonstrate that FlowAlign improves instance level correspondence while preserving target-space structure across different modality settings.
benchmark - arxiv:2611.09848 · cs.CLHow far does a hash-only sender go? A pre-registered map of sender memory versus bits for short text messagesMarcelo Samoilenko
Can a sender that only computes hashes, with a language model only at the receiver, transmit short human text more cheaply than a classical sender trained on the same domain? We answer with one pre-registered run on 1,000 English user utterances from Taskmaster-1. The primary endpoint is the cost per message to deliver 100% of the messages (hash bits, plus a 2-byte negative acknowledgement and the raw text when the first attempt fails), paired against an in-domain PPM order-1 sender. The hash-only sender delivered 857 of 1,000 messages on the first attempt and cost 192.8 bits per message, 18.4 bits more than PPM order 1 (95% bootstrap interval 10.9 to 26.3); by the pre-registered rule it spends more and is not equivalent within one byte. As secondary results we map measured sender memory against bits. The hash-only sender (8,190 B on a host build, keyed BLAKE2b tag included, as in every method) was the cheapest point in bits from 8,190 B up to, but excluding, PPM order 1 at 24,844 B, 45 to 46 bits below smaz2 and Unishox2; from PPM order 1 upwards, in-domain senders spend fewer bits, except order 2 pruned to 16 KiB. In airtime it loses to PPM order 1 on all five links priced, to Unishox2 on four, and to raw text with a tag on two, because each failure costs an acknowledgement frame and a fallback frame. A wider receiver search recovers half of the failures and makes the cost 7.7 bits lower than PPM order 1, at about three minutes of receiver time per failure, and still loses in airtime on four of five links. We report the measurement as a delimitation, with every deviation of the run.
memory - arxiv:2611.09847 · cs.ROComputer-Use Agents Are Embodied GeneralistsPranjal Aggarwal, Sean Welleck
In many areas of AI, general-purpose models have increasingly replaced handcrafted methods and specialized architectures. We argue that embodied AI is poised for a similar shift. Our key claim is that embodied tasks can be formulated as interaction with digital software. Specifically, we build software that renders an agent's view of a physical environment on a screen and provides controls for acting in it. A computer-use agent (CUA) can then perform embodied tasks through the same screenshots and keyboard and mouse actions that it uses to operate ordinary software. Using a single CUA design and changing only the software, our entries win first-, second-, third-, and sixth-place prizes across four ECCV workshop competitions, without any competition-specific agent tuning. These results suggest that the skills CUAs learn on software, such as visual grounding, precise clicking, and exploration, already carry over to embodied tasks, giving CUAs a far broader scope in embodied AI than previously realized. We therefore advocate that computer-use agents can be embodied generalists. Further, we advocate that, by following the training strategy that made computer-use agents general, embodied AI can converge on a single general model as well, ushering in a new era of general embodied AI. Website: https://pranjal2041.github.io/cua-embodied
embodiedagent - arxiv:2611.09846 · cs.AIEnsemble Methodology for Reference-Free Confidence Estimation in LLM-based Code SummarizationMarco Colombo, Balaji A Jayaram, Sowmyakumari Selvaraj, Sumit Tokkar +4
This paper describes an approach and use case to estimate the correctness of LLM-generated COBOL summaries without access to human-authored references. The core intuition is that inter-model agreement on semantics is informative about truth, and that an LLM-as-judge can provide higher-fidelity pairwise assessments than traditional string- or embedding-based metrics. We operationalise this intuition with (i) controlled summary generation from diverse models; (ii) a LLM evaluator that compares pairs of summaries and emits structured judgments; and (iii) a calibrated aggregation rule that combines supportive evidence while penalising only corroborated contradictions.
evaluatorllm-as-judge - arxiv:2611.09843 · cs.LGFifty-Two Shades of Pig: What the UCR Archive Really Asks Classifiers to DoOmar Swelam, Jaris Küken, Shi Bin Hoo, Frank Hutter +1
The pigs were bleeding. That was the point of the experiment: could airway pressure reveal hemorrhage before it became critical? In the UCR Time Series Classification Archive, the de facto benchmark for Time Series Classification (TSC), the same recordings ask a different question: "which of 52 pigs produced this trace?". The dataset PigAirwayPressure is not an outlier. Although the archive is the standard benchmark, whether its tasks resemble the problems TSC is meant to solve has rarely been examined. We find that several datasets are repurposed from their original task, built from toy setups, or have flawed test splits. We further question how representative the archive's ordered non-temporal sequences (e.g. spectra, shape outlines) are of real TSC, and the potential downsides from the decomposition of multivariate problems. Our findings call for more careful interpretation of archive-wide scores and greater attention to deployment-relevant evaluation. The audit is at https://github.com/automl/ucr-provenance-audit.
benchmark - arxiv:2611.09841 · physics.opticsDiscrete time crystals and heteroclinic switching in actively mode-locked semiconductor lasersElias R. Koch, Albert Gallemí, Clàudia López, Finn Petersen +2
We investigate the mechanisms underlying the emergence of discrete time crystals in actively mode-locked semiconductor lasers. Our analysis is based on a Haus master equation generalized to retain carrier memory between round-trips. We perform a detailed bifurcation analysis that reveals a well-defined hierarchy of subharmonic states, illustrating how pulses may suppress their neighbors when the modulation runs faster than the pulse round-trip frequency. To identify the underlying mechanism, we derive effective equations of motion that couple the pulse energies, their positions relative to the modulation transmission maxima, and the local time-dependent gain. These reveal that transitions between time-crystalline states occur via transcritical global bifurcations, opening a regime of heteroclinic winnerless competition between symmetry-degenerate configurations.
memory - arxiv:2611.09836 · cs.LGCan Agentic Systems Close the Gap in Radiology with Production AI Tools?Aubin Charley, Julien Khlaut, Charles Corbière, Pierre Manceron +2
Radiology AI has developed along two paths: specialized models that perform well on narrow tasks but cover few clinical findings, and generalist models that address a broader range of tasks but remain prone to hallucination and inaccurate measurements. Combining the two through agentic systems offers a route towards comprehensive radiology reporting, but published systems are tied to one modality or body part. We describe a radiology agentic framework that standardizes how an orchestrator calls modular tools to produce a structured report, and use it to measure what such a system can do with the tools from FDA-cleared radiology AI devices. Report generation is simulated on a large multi-modality report corpus, with the orchestrator given either a catalog of finding classes consolidated from the cleared devices, each available on the modalities it is cleared for, or an unlimited catalog in which any class can be requested, and planning either from the clinical indication or from the target report. Tool outputs are simulated from the reference report, so the comparison isolates tool availability and orchestration from tool error. With the cleared catalog and the indication alone, the orchestrator recovers under a third of the clinically actionable findings; recovery improves substantially with the unlimited catalog, and further when the plan is written from the target report. Attributing every missed finding to a cause shows that most are classes no cleared device offers for the study's modality, and that a third are classes or tools that were offered and not called. Adding the most frequently missing tools closes most of the gap to the unlimited catalog on held-out reports, whereas a smaller orchestrator with the unlimited catalog gives up the whole gain. Tool availability and effective orchestration are thus complementary constraints on comprehensive AI-assisted radiology reporting.
agentic - arxiv:2611.09825 · cs.AIBitDecoding-FP4: Native FP4 KV Cache Decoding via Algorithm-System Co-DesignDayou Du, Woon Yee Ng, Guangming Sheng, Nick Lau +2
The key-value (KV) cache dominates memory capacity and bandwidth in long-context LLM serving. While 4-bit microscaling formats (NVFP4, MXFP4) shrink the cache and run attention on FP4 Tensor Cores, a fundamental gap remains between algorithmic quantization and system-aware acceleration: existing FP4 KV systems use vanilla quantization that causes severe accuracy drops, whereas accuracy-oriented 4-bit algorithms rely on vector codebooks whose lookup overhead degrades decoding throughput. We present BitDecoding-FP4, an algorithm--system co-design for accurate and efficient native FP4 KV decoding. We mitigate three error sources: Dual-Residual Query Quantization splits queries into two FP4 tensors, cutting query logit error from 60% to 1%; Underflow-Aware Probability Scaling prevents E4M3 scales from rounding long-context probabilities to zero; and Outlier-Aware Key Quantization combines static smoothing, Hadamard rotation, and scale search to suppress key outliers under token-wise scaling. At the system level, we remove the softmax bottleneck via transposed softmax with fused probability quantization and stacked query MMA on paged KV caches. Across dense and MoE models with GQA and MLA, BitDecoding-FP4 stays within 0.5 points of BF16 KV accuracy with NVFP4 (1 point with MXFP4), and tracks BF16 KV in RL training rollouts. In end-to-end serving on NVIDIA B300 with Qwen3-4B and Qwen3-30B-A3B, it delivers 2.4-3.0$\times$ higher peak serving throughput than BF16 KV. The implementation is available at https://github.com/OpenBitSys/BitDecoding-FP4.
memorylong-context - arxiv:2611.09822 · cs.AIDiscovering formally verified quantum programs with AISergey Shirobokov, Ben Jaderberg, Katerina Hristova, George Tsoukalas +21
The timeline by which quantum computers can solve breakthrough problems in science depends on improved quantum algorithms as much as on better hardware. Such algorithmic advances are typically achieved through many years of expert human collaboration. In mathematics and computer science, large language models have accelerated progress through automated discovery loops. However, quantum algorithms can only be evaluated by classical computers on restricted or small instances. Here we demonstrate automated discovery of quantum programs beyond these regimes including order-of-magnitude improvements on scientific applications. To achieve this, we build a quantum program library in the Lean theorem prover in which large language model agents synthesize novel programs against formal problem specifications and verify their correctness and cost. Agent teams collaborate, sharing results and knowledge through wikis and discussing ideas at organized conferences. Our results improve on prior baselines for parallel search (up to 1.4x), quantum arithmetic (290x), and simulations relevant to fusion, photochemistry and drug-metabolism by 11x, 8x and 2.5x respectively. Combining human and artificial intelligence at scale, under the rigor of mathematical proof, lets researchers design algorithms that today's classical and quantum computers cannot test, shortening the path to practical quantum computation.
agent - arxiv:2611.09818 · cs.LGGenerative Latent Priors for Interpreting and Steering Protein Language ModelsDaniel Hinjos, Lucía Urcelay
Dominant methods for interpreting large models impose strong structural assumptions. Generative Latent Priors (GLP) are a complementary approach: a diffusion model that learns the distribution of a frozen model's activations directly, imposing no assumptions. We port GLP to ESM-2 activations (UniRef50) and evaluate it at two host scales (ESM-2-8M and -650M), both as a lens for protein representations and as a prior for editing them. Generation quality scales smoothly with depth and compute. As a lens, single GLP "meta-neurons" are the strongest probes of discrete Swiss-Prot concepts at both hosts, beating InterPLM's released SAE features, and lead on continuous properties at 8M though not at 650M; falsification tests show this is mostly concentration, not new information. As an editing prior, the on-manifold correction makes DiffMean steering of three biochemical properties effective and restores on-manifoldness, but only in isolation: re-embedding the sequences actually produced after several rounds of decoding, the structural plausibility is compromised. This is mainly reproduced in additional SAE-based clamping structural steering experiments, where only Helix motif succeeds. The naturalness failure is traced to the rewriting coverage and the decoder rather than the prior, highlighting what the method must clear before it can serve for both discovery-by-editing and discovery-by-exploration. Trained GLPs are released on huggingface.co/collections/HPAI-BSC/glprot.
helix - arxiv:2611.09816 · cs.CVBEV3SR: Evidence-Aware 3D Spatial Reasoning from VideoFan Zhu, Rui Yu
3D spatial reasoning from video remains challenging for VLMs because videos provide sequential, view-local observations rather than persistent metric scene records. Many questions must be answered in a query-specified perspective rather than the observed camera viewpoint, and prompting with a fixed set of RGB views may miss sparse or occluded visual cues. To address these gaps, we introduce BEV3SR, a training-free and backbone-agnostic framework for RGB-D videos with camera parameters. BEV3SR turns video observations into queryable metric bird's-eye-view (BEV) scene memory and formalizes each query into an evidence-aware spatial task. The Spatial Task Formalizer specifies the query's reference frame, spatial constraints, objective, and evidence requirement. The evidence requirement determines the appropriate reasoning pathway. A query treated as requiring geometric evidence is handled by the Geometric Reasoner, which deterministically computes the answer using the metric BEV records in the query-specified reference frame. Otherwise, a query treated as requiring visual evidence is passed to the Visual Evidence Aligner, which uses BEV-indexed object records and their associated supporting views to retrieve object-consistent and scene-context RGB frames for VLM grounding. Within a standard 16-frame budget, BEV3SR consistently improves different VLM backbones across multiple 3D scene QA benchmarks, delivering large gains on spatially intensive questions and significant improvements on open-ended scene QA. These results demonstrate that BEV3SR offers an effective, efficient, and interpretable framework for explicit 3D spatial reasoning from video.
memorybenchmark - arxiv:2611.09807 · cs.CLCompaction, decay and reinforcement in a two-tier conversational memory: a paired evaluation against retrieval-augmented generationOsamudiamen Edogun, Joshua Obigwe, Irfan Mehmood
A conversational agent that remembers has to decide what to keep. Retrieval-augmented generation keeps the transcript and searches it; the alternative is to compact each turn into atomic facts and search those. We compare the two on LoCoMo, with both arms answering all 1,985 questions in one window, against the same stores, with the same generator and judge, and both predictions on one row, which licenses McNemar's exact test on the 512 questions where they disagree. On the four categories comparable work reports, compaction scores 71.4 per cent against 64.6, a lead of 6.8 points (p = 5.5e-08) at 34 per cent fewer prompt tokens. On all five, including the adversarial set comparable work drops, the lead falls to 2.0 and stops being significant, because the baseline leads adversarial by 14.6. That loss has a structure. Our lead falls as the reference answer demands more source wording, 9.6 points ahead at one or two reference words and 3.1 behind past eight, with token overlap moving the same way, so the judge is not the cause. Compaction keeps what a fact asserts and discards how it was said, so it wins where the answer is a value and loses where it is a phrasing. We also report four negative results with their power calculations.
retrieval-augmentedagent - arxiv:2611.09805 · cs.LGScaling Graph Similarity Learning with Graph Neural NetworksAhmet Tüzen, Helge Langseth, Kjetil Nørvåg
Graph similarity computation is an important task in graph mining with many real-world applications, such as chemical compound analysis, program dependence comparison, and social network analysis. The problem is computationally challenging: Graph Edit Distance (GED) and Maximum Common Subgraph (MCS), which quantify the distance between graphs and identify their largest shared subgraph, respectively, are NP-hard to compute exactly. To address these computational challenges, Graph Neural Networks (GNNs) have been applied to this problem. However, existing GNN-based approaches have primarily focused on small-scale graph datasets, while scalability to larger scales remains underexplored. In this work, we propose FiLSim, a scalable GNN-based framework, and investigate whether a simplified GNN pipeline can maintain competitive performance on standard benchmarks while scaling to large graphs. We additionally propose a sequential graph-editing strategy to construct large-scale graph datasets with reliable GED labels to evaluate scalability. The proposed framework is evaluated on four benchmark datasets and four newly created large-scale datasets, demonstrating its effectiveness and scalability. FiLSim remains accurate on graphs with more than 150{,}000 nodes, where the baseline methods run out of memory or fail to learn, and scores all test pairs within seconds on a single GPU.
memorybenchmark - arxiv:2611.09802 · cs.AIWhen Instruction Files Go Stale: Documentation Debt and Hidden Cost in Coding AgentsTural Mehtiyev, Biruk Tadesse, Lawrence Arkoh, Wesley Assunção
LLM-based coding agents such as Claude Code are guided by repository instruction files (also called context files or agent configuration files) like AGENTS.md and CLAUDE.md, which are read in every agent session. As any documentation artifact, these files can fall behind the code and become stale (e.g., a test folder moves, but the file still points to the old one). In technical debt terms, such a stale instruction is documentation debt, and its interest is the cost it adds to every later agent session. Evaluations that check only task success can miss this cost, and the stale file can persist. In this study, we ask what leads agents to act on a stale instruction, how it affects their behavior and at what cost, and how they recover. We study three real-world stale instructions from active open-source repositories in depth, each a different kind of guidance: a test command, a code location, and a procedure for adding a component. Agents fix 22 maintenance tasks under three versions of each instruction (left stale, corrected to the current path, and removed), with the code and the task held fixed, and we analyze 732 trajectories of two models at the level of individual tool calls. We find that agents follow stale guidance when it is relevant to their tasks, sometimes even when correct information is available elsewhere. On the stale test command, Claude Sonnet 5.5 needed about two extra tool calls per run, a third more than with the corrected command; in the other cases, we could not detect a difference. Although agents generally recovered from the immediate failures, none repaired the underlying instruction file. Correcting the instructions eliminated operations on obsolete paths, while removing them prevented most. Our findings show that task success alone can hide the costs of documentation debt, highlighting the need to evaluate agent behavior beyond final outcomes.
agent - arxiv:2611.09798 · cs.AITest-Time Compute Scales Better with Parallel Agent TeamsJunlin Wang, Haozhe Chen, Ang Li, Rahul Chalamala +1
Test-time scaling improves the performance of a fixed model by spending more computation at inference. There are two common ways to spend it on agents when a verifier can score every attempt. Best-of-N sampling runs many independent agents and keeps the best result but saturates quickly. One large team of collaborating agents can score higher but stops improving as it grows. Can we get the best of both? Our method, parallel agent teams, does so by sampling many small teams in parallel and keeping the best team's result. We test it on 12 long-horizon tasks from GPU kernels and formal proofs to exploit development. With a 27B model and a budget of 100 agents, parallel teams of ten beat parallel agents on 10 of 12 tasks and beat one team of 100 on all four tasks where we ran such large teams. At equal dollar cost the same parallel teams beat a frontier model on half the tasks. A simple statistical model shows that individual agents lose because their many samples have a low mean and that one large team loses because a single sample gets no gain from variance. Sampling many small teams in parallel balances a high mean with the gain from variance. Parallel agent teams thus offer a more effective way to scale test-time compute.
agent - arxiv:2611.09794 · cs.ROSeeing What Matters: Benchmarking Task-Conditioned Actionable Spatial Belief for ManipulationYina Jian, Mingkun Xu
When should a robot look at the cup, and when at the portafilter? Spatial-reasoning benchmarks evaluate whether a model can describe a scene, but manipulation needs something more specific: the ability to acquire the particular evidence that makes the current command executable. A view that is informative for one task can be useless for another, even at the same workstation. We introduce Seeing What Matters, a benchmark for task-conditioned actionable spatial belief in simulated tabletop manipulation. Across three domains--coffee workstation, drawer/cabinet, and microwave/appliance--the benchmark provides 860 frozen evaluation cases; expanding across methods, budgets, and inference repeats yields more than 115,000 evaluator episodes under two protocols: closed-loop actionability and same-object functional out-of-distribution generalization. Each case pairs a set of candidate RGB views with a manipulation command, shared scripted skills, and hidden postconditions, so the benchmark directly measures whether a view-selection policy acquires the observations its command actually needs. We also provide a diagnostic reference pipeline that maintains a spatial-functional belief state and ranks views by missing evidence. At primary budget settings, this pipeline reaches 98.8% closed-loop and 98.9% functional-OOD success; ablating the functional or command-binding components collapses performance to near zero. Seven vision-language models, evaluated on the same cases, reach only 26-43%. Replacing the reference pipeline's coverage metadata with frozen RGB-derived scores preserves 98.5%/98.9% success; shuffling the visual scores collapses it to 15%/11%, confirming that the result depends on image content. These gaps suggest that spatial belief for manipulation should be evaluated not by scene-level completeness, but by whether it acquires the evidence the current command requires.
manipulationbenchmarkevaluator - arxiv:2611.09792 · cs.LGKnowing When to Intervene: Causal Evaluation of AI AgentsAyush Sawarni, Wenqian Xing, Tijana Zrnic, Ramesh Johari +1
Current evaluations of agentic AI systems primarily rely on benchmarks. However, the static nature of benchmarks makes them uninformative about the value of auditing interventions at deployment time. For reliable deployment, we often need to determine within an agent's trajectory whether or not to intervene, e.g., by verifying the agent's current output or steering it toward different behavior, using either human review or more costly models or time-consuming agentic supervision. The value of such interventions is fundamentally a causal inference question, asking whether the causal effect of such an intervention is beneficial for the final outcome. Estimating this causal effect is particularly challenging due to the dynamic nature of the problem: the impact of an intervention depends on the current context and environment, as well as the agent's future behavior. We develop the first methodologically grounded framework for evaluating the causal effects of randomized interventions within agent trajectories. By leveraging a relatively small dataset of agent trajectories collected under randomized interventions such as verification or screening, we can estimate the dynamic causal effect of a rich class of (contextual) intervention policies. The structural parameters that we estimate can also be used to construct safety leaderboards that capture the risk of various agentic actions, which can be used by practitioners for audit prioritization. Applying our methods to tau-bench Retail, ToolSandbox, and ImpossibleBench, we find that intervention effects vary substantially across agent actions and states. More broadly, our work suggests a rich research agenda at the intersection of causal inference and agentic evaluation, which we believe is essential for the reliable and safe deployment of AI agents in the real world.
agentai agentagenticbenchmarkleaderboard - arxiv:2611.09791 · cs.AIAssessing the Impact of Retrieval-Augmented Generation on Algorithmic Problem-SolvingGabriel Souza Baggio, Gabriel Machado Lunardi, Guilherme Medeiros Machado, Subhankar Maity +2
Competitive programming problems pose distinctive challenges for Large Language Models, as they demand not only proficient code generation but also advanced algorithmic reasoning. This study presents a Retrieval-Augmented Generation approach specifically designed to enhance the capacity of Large Language Models to address such problems. Our approach integrates an embedding-based retrieval mechanism and a similarity search algorithm to identify analogous problems within a knowledge database, employing a dynamically adjusted threshold to mitigate retrieval noise. We conducted an evaluation using the CodeContests dataset, focusing on the test split problems that exhibit similarity to instances from the training split. The experimental framework considered three conditions: (i) problems provided solely with their original descriptions, (ii) problems supplemented with randomly selected examples from the training split, and (iii) problems augmented with similar examples retrieved through the proposed method. In addition, we incorporated a Min-K membership inference analysis to examine whether the generated code solutions were influenced by potential memorization of training data within the evaluated Large Language Models. This complementary perspective enables a deeper understanding of model behavior by distinguishing genuine generalization through retrieval from implicit recall of pretraining samples. Performance was assessed using both pass rates and the CodeForces Elo rating of the problems. The results provide empirical evidence that similarity-based retrieval substantially enhances Large Language Models performance in competitive programming scenarios, offering an scalable framework for improving code generation in algorithmic problem-solving contexts.
retrieval-augmented - arxiv:2611.09777 · cs.ROAutonomous Whole-body Loco-manipulation Using a Snake RobotAdarsh Salagame, Henry Noyes, Haroon Hublikar, Alireza Ramezani
Snake robots can leverage their hyper-redundant morphology for both locomotion and manipulation through distributed body contact but whole-body loco-manipulation is particularly challenging since the entire body serves simultaneously as propulsive mechanism and manipulation tool Prior work on wheeled legged and snake robot platforms has decoupled these functions through dedicated appendages body segmentation or grippers We present closed-loop loco-manipulation with our snake robot COBRA without this distinction such that any part of the body may make contact We develop a contact-implicit dynamic model unifying robot-ground robot-object and object-ground interactions within a Moreau-Jean time-stepping framework and formulate a trajectory optimization over the full-order dynamics guided by a geometric planner and nominal proportional controller to autonomously push an object to a goal location during sidewinding We validate the planner and controller on hardware across varied configurations and present simulation results for a simplified trajectory optimization that resolves robot-object and object-ground contacts
manipulationgripper - arxiv:2611.09770 · cs.CVROSE: Risk-Aware Orthogonal Subspace Navigation for Lifelong Knowledge Editing in Multimodal Large Language ModelsLingyun Song, Ziyao Chen, Kang Pan, Xiaolin Han +4
Multimodal Large Language Models (MLLMs) require continuous factual updates to stay current, yet lifelong knowledge editing remains a daunting challenge. The primary obstacle is the severe Locality Erosion and Catastrophic Forgetting, where sequential edits inevitably interfere with pre-trained capabilities and prior updates. In this paper, we propose \textbf{ROSE}, a \textbf{R}isk-aware \textbf{O}rthogonal \textbf{S}ubspace navigation framework for lifelong knowledge \textbf{E}diting. ROSE addresses forgetting by aiming to restrict parameter updates to the orthogonal complement of previously edited subspaces. Crucially, it introduces a risk-guided mechanism that utilizes a dynamically computed risk mask to safeguard specific factual parameter subspaces essential for model stability. During inference, ROSE employs a multi-granularity knowledge integration strategy, featuring prototype-anchored semantic gating and perturbation-driven subspace fusion to precisely activate edited knowledge for relevant queries while limiting interference on unrelated inputs. Comprehensive experiments on two prominent multimodal benchmarks show that ROSE is competitive with state-of-the-art approaches across five key metrics, while maintaining outstanding stability over $1,000$ sequential updates. Our code is available at https://github.com/sunlitsong/ROSE.
benchmark - arxiv:2611.09769 · cs.CVAnimWeb-Bench: Benchmarking Animated Webpage Code Generation from VideosZhongyi Zhang, Jingyu Xiao, Shuqing Li, Michael R. Lyu
Animation is pervasive on the Web, the CSS animation property alone appears on over three-quarters of webpages. Yet existing webpage code generation benchmarks target static screenshots, user-triggered interactions, or isolated motion graphics, leaving the ability of multimodal large language models (MLLMs) and coding agents to reconstruct complete animated webpages largely unexplored. To fill this gap, we introduce AnimWeb-Bench, the first benchmark for animated webpage code generation from videos. It comprises 124 real-world webpages spanning 7 application domains and 33 animation types, with 91.9% of pages combining multiple types, each paired with a reference video and the static assets of the original webpage. We further propose a human-aligned, two-dimensional evaluation protocol: Static Appearance Fidelity measures structure, text, color, and visual similarity on aligned keyframes, while Animation Fidelity tracks UI components across frames and compares their position, size, velocity, and appearance with those in the reference video. Evaluating 14 MLLMs and three coding agents reveals: (1) reproducing webpage animations is substantially more difficult than reproducing static appearance; (2) explicit frame inspection and render feedback improve the reconstruction quality of coding agents; and (3) rapid, continuous changes are harder to reproduce than discrete state transitions. Our code and dataset are available at https://starry1001.github.io/AnimWeb-Bench-website/ .
benchmarkevaluation protocol - arxiv:2611.09768 · cs.AILanguage as Trait, Politeness as Strategy: Controlled Tests of Linguistic Effects in LLM Social-Deduction AgentsYuchen Li, Lisa Liu, Runqi Zou, Shuxuan Kuang +2
Large language model (LLM) agents coordinate through conversation, and the same message can be phrased in simpler or more complex language, bluntly or politely. We test whether such differences change how other agents judge and act, using the LURK Werewolf framework, which separates each utterance's speech-act predicate from its natural-language realization. We rewrite only the realization, check that the game content is preserved, and replay each game day from saved checkpoints against a re-sampled control. We vary linguistic range consistently across a speaker's messages and politeness strategy per utterance. Across 30 games and 4,335 replays, neither manipulation changes the trust other agents place in the speaker, and no effect on votes or night actions survives correction for multiple comparisons; the largest uncorrected effect is that shifting range away from a neutral paraphrase changes about 4 percentage points more of the other players' votes. In natural play, 88% of face-threatening speech is bald on-record regardless of role.
manipulation - arxiv:2611.09740 · cs.CVZero-Shot Benchmarking of Unified Multimodal Models for Shallow-Water UAV Image Distortion Assessment and RestorationDario Scilla, Omar A. Lopez, Mattia Rosso, Kasper Johansen +1
Shallow-water UAV imagery provides centimeter-scale observations of benthic environments, but water-induced distortions complicate interpretation. To our knowledge, we present the first systematic zero-shot benchmark of unified multimodal large language models (MLLMs) for both assessing water-induced distortions and restoring shallow-water imagery within the same model framework. We evaluate five open-weight unified MLLMs with frozen pretrained weights and without access to paired reference images or descriptions of the true seabed. The benchmark contains 1,000 image pairs: 600 semi-synthetic UAV scenes created by applying synthetic water distortions to real shallow-water imagery and 400 fully synthetic through-water scenes. Three independent human annotations cover five distortion types. We evaluate four assessment prompts (P0-P3), from direct instruction to structured task decomposition, on all 1,000 images and four diagnosis-guided restoration prompts (R0-R3) on a common 100-image subset. Assessment varies substantially by model and prompt, with systematic prediction biases and a best pooled balanced accuracy of 0.654 for distortion presence. The best pooled PSNR gain is +0.255 dB, no unified MLLM improves pooled LPIPS, and some outputs alter scene structure or hallucinate content. Two restoration-only baselines also show limited transfer: InstructIR achieves a pooled PSNR gain of +0.040 dB, while PromptFix degrades all four pooled metrics under every prompt. These results show that recognizing shallow-water degradation does not imply reliable physical restoration and provide a reproducible baseline for future work.
benchmark - arxiv:2611.09730 · cs.AIWhat to Delegate, What to Preserve: Human Preferences for Agentic AI Across Everyday Life and WorkSarah Wheeler, Veronica Rammouz, Yaoyao Qian, Zexuan Liu +4
AI agents can execute multi-step tasks on users' behalf, yet the boundaries of acceptable delegation remain unclear. We report a scenario-based survey of 289 U.S. adults with prior AI experience covering 45 use cases across nine life domains. Each respondent evaluated one scenario per domain, producing repeated measures of automation preference, trust, collaboration, AI leadership, privacy concern, and willingness to pay. Automation preferences varied substantially across domains. Generalized estimating equation models showed that greater trust in the AI agent, desired AI leadership, prior task-specific AI use, and anticipated financial savings were associated with greater automation preference. In contrast, concerns that automation would reduce personal fulfillment predicted lower automation preference. Open-ended responses further revealed a preference for bounded autonomy, with participants often favoring substantial AI execution while retaining human involvement, oversight, or final authority. We argue that delegation is best understood as a configuration of execution, collaboration, leadership, and authority.
ai agentagentic - arxiv:2611.09729 · cs.LGOne Run Is Not Enough: A Monte Carlo Protocol for Evaluating LLM Reliability on Data Science TasksWenqi Ni, Ozan Evkaya, Miguel de Carvalho
Large language models (LLMs) have shown promise for automating data science workflows. Benchmarks such as DataSciBench (2026) and StatQA (2024) have advanced the evaluation of their capabilities across data science and statistical tasks. Assessing their reliability requires examining both the quality of their responses and their consistency across repeated runs, task types, and difficulty levels. With this aim, we introduce DS-QA, a benchmark comprising 100 data science and statistics questions drawn from UK university courses, spanning 16 datasets and three difficulty levels. We develop a three-layer evaluation framework that combines general response properties (runtime, verbosity and text similarity), coarse-grained metrics (accuracy, code executability and completeness), and task-specific metrics. We evaluate OpenAI models using a Monte Carlo protocol with 100 repeated runs. The results reveal substantial variability in responses to data visualisation and supervised learning tasks, particularly at medium and hard difficulty levels, highlighting the importance of evaluating consistency alongside response quality. We also find that an increase of 0.1 in the word verbosity ratio corresponds to an increase of about 0.013 in the token verbosity ratio.
benchmarkevaluation framework - arxiv:2611.09725 · cs.CVTriNetra: Thermal-Reliability-Guided RGB-Thermal Fusion and Tracking for SurveillanceSadique Khatib, Aditya Ahirrao, Vedant Sonawane, Disha Kapse +1
Visible cameras fail at night and thermal sensors fail under saturation, blur or loss of contrast, yet most RGB-thermal (RGB-T) detectors mix the two streams with fixed weights and cannot tell when one modality has become unreliable. TriNetra is a reliability-aware surveillance framework built around this problem. The proposed solution computes a frame-level Thermal Reliability Coefficient (TRC) from five label-free image cues, uses it to gate the thermal features of a dual-branch YOLOv8m detector whose visible branch, neck and head are frozen and COCO-pretrained, and passes the same score to a ByteTrack-based tracker that feeds a rule-based behaviour analyser. On the M3FD benchmark (2,940/630/630 split) the detector reaches 87.6% precision, 79.4% recall, 70.0% mAP@50 and 47.8% mAP@50:95 on the test split, with 17.0 M of its 42.9 M parameters trained. TRC stays near 0.78 on clean frames (standard deviation 0.012), decreases with the severity of thermal saturation and blur, reaching 0.53 and 0.69 at full severity, and is insensitive to RGB-thermal misalignment. With identical weights, the adaptive gate matches a fixed gate on clean data (70.0% versus 69.8% mAP@50) but keeps 43.6% mAP@50 when the thermal stream is corrupted, against 38.4%, 29.2% and 25.0% for a fixed gate, full thermal admission and a closed gate; the 5.1-point gain over the fixed gate has a 95% bootstrap confidence interval of +3.6 to +6.5 points. In tracking, the TRC-gated association performs on par with ByteTrack on a synthetic benchmark with exact ground truth and on 14 M3FD clips, and is slightly less stable under thermal saturation, so its measured benefit is confined to the detector. The limitations of the current evaluation are reported together with the results.
benchmark - arxiv:2611.09722 · cs.AIFOSCAR: Accelerating Decode Attention over 2- and 4-Bit KV Caches through Affine FoldingDonglin Zhuang, Pragaash Ponnusamy, Xiaoxia Wu, Shuaiwen Leon Song +1
Large-language-model decode attention streams the key-value (KV) cache for every generated token. Packing that cache into 2-bit or 4-bit quantized integers reduces memory traffic but does not automatically translate into latency saving. Existing 2-bit fused kernels reach under 31%/15% of peak HBM throughput on H100/B200. Converting each value and applying its scale and offset before feeding Tensor Cores is a major bottleneck because instruction-issue throughput is scarce on modern GPUs. Instead of applying scale and offset to each KV value, FOSCAR uses affine folding to apply them to the other operand or the output of decode-attention matrix products, depending on the quantization scheme (token-wise or channel-wise). This reduces scale-and-offset operations by up to two orders of magnitude while preserving the dequantized attention output in exact arithmetic. FOSCAR also uses bit manipulation to feed KV operands to Tensor Cores efficiently, requiring only 1 and 0.19 ALU instructions per INT2 element on H100 and B200, respectively. Finally, FOSCAR builds warp-specialized GPU pipelines for Hopper (H100) and Blackwell (B200) using wgmma and tcgen05. On H100, FOSCAR INT2 and INT4 kernels achieve geo-mean speedups of $3.36\times$ and $2.83\times$ over FlashInfer BF16, and $2.03\times$ and $1.71\times$ over TensorRT-LLM FP8. On B200, the corresponding INT2 and INT4 speedups are $2.67\times$ and $2.30\times$ over FlashInfer BF16, $1.48\times$ and $1.28\times$ over TensorRT-LLM FP8, and $1.16\times$ and $1.00\times$ over TensorRT-LLM NVFP4. We integrate FOSCAR into SGLang to benchmark serving efficiency. On long-context serving workloads, FOSCAR increases peak aggregate throughput by up to 170% and 341% over 2-bit and 16-bit baselines, respectively. We open-source FOSCAR as a kernel library at https://github.com/FutureMLS-Lab/FOSCAR.
manipulationmemorylong-contextbenchmark - arxiv:2611.09721 · cs.AIMemoryDojo: Benchmarking Poisoning Through LLM Agent MemoryDavid H. Yang, Blake Bullwinkel, Amanda Minnich, Mohammad Mohammadi Amiri +1
LLM agents benefit from using memory systems to improve performance on long-horizon, multi-session tasks. However, memory introduces a critical attack surface. Malicious instructions encountered in one session may be stored and propagated through memory before being retrieved and acted upon in later sessions. Because memory systems differ in how they process and retrieve information, the downstream effects of poisoning can vary significantly across systems. To quantify the effects of poisoning across different memory architectures, we introduce MemoryDojo, a benchmark for tracing the lifecycle of poisoned information through agent memory and measuring its influence on subsequent agent behavior. MemoryDojo is easily extensible, consists of 155 multi-session long-horizon trajectories covering four domains, and supports five different memory systems. Through our evaluation, we find that all memory systems are susceptible to memory poisoning, with substantial variation in downstream effects. We further propose a memory-aware poison optimization process that increases the poisoning efficacy, highlighting differences between memory architectures. We hope that MemoryDojo will support research on understanding and mitigating agentic memory risks.
memorymemory architectureagent memoryagentllm agentagentic - arxiv:2611.09717 · cs.CVDirect Regression and Geometric Recovery for Event-based Spacecraft Pose EstimationPruthvi Nandan Janga, Kashyap Hegde Kota, Bharatesh Chakravarthi
Event cameras offer high temporal resolution and dynamic range for uncooperative spacecraft pose estimation, but models trained on synthetic event streams must generalize to the noise, background activity, and contrast variation of physical sensors. We present a comparative systems study of two independently developed approaches to synthetic-to-real, event-based 6-DoF spacecraft pose estimation. The first is a modular pipeline that combines learned keypoint detection and unlabeled target-domain adaptation with geometry-based pose recovery and temporal smoothing. The second directly regresses translation and rotation from global and cropped event representations using an apparent-scale cue, physics-motivated synthetic corruptions, and feature-consistency regularization. We evaluate both systems under the same SPARK 2026 hidden-test protocol on the real-event split of SPADES, with predictions for all test frames. The results reveal a component-wise reversal concealed by the aggregate challenge score: direct regression reduces orientation error by 22.7% (1.1767 to 0.9101 rad) and obtains a 13.4% lower composite score, whereas the modular pipeline reduces relative translation error by 78.7% (0.1335 to 0.0285). Archived configurations and a qualitative failure case identify plausible explanations for these complementary error profiles, without isolating their causes. These findings motivate a testable hybrid design that combines direct orientation estimation with geometry-constrained translation recovery.
event camera - arxiv:2611.09714 · cs.CVBeyond Segmentation: Enriching ImageCAS with Coronary Ostia Landmarks, Rooted Centerline Graphs, and Open-Weight Localization ModelsMatteo Leccardi, Andrea Moglia, Marco Morandini, Michele Brembilla +7
Coronary ostia and centerlines are key reference points for automated analysis of coronary computed tomography angiography (CCTA), from surgical planning to intraoperative vessel tracking. However, the only public ostium annotations cover 77 non-gated aortic scans, not CCTA, while CAT08 and ASOCA (32 and 40 CCTA scans) mark the ostia only implicitly, as centerline start points. This work enriches ImageCAS, the largest public dataset for coronary artery segmentation, with ostium annotations for its 1000 patients and centerline graphs rooted at these new ostia, making it, to the authors' knowledge, the largest public dataset for ostium localization and the only one with centerlines following the official ImageCAS masks and starting at annotated ostia. Three annotators marked each ostium in the first 500 volumes; in the next 500, the candidates of a model trained on the first batch were corrected (human-in-the-loop). Six 3D localization models, trained on ImageCAS, CAT08 and ASOCA, were benchmarked on 58 held-out volumes (50 ImageCAS, 4 CAT08, 4 ASOCA). SwinUNETR-V2 and nnResUNet localized all ostia without failure (>5 mm error), with median errors of 0.70-0.87 mm. Trained on CAT08 and ASOCA alone, SwinUNETR-V2 fails on 34% of right and 53% of left ostia, whereas adding 250 ImageCAS volumes removes all failures. The centerlines were extracted by a deterministic, quality-controlled minimal-path procedure; on ASOCA, they cover a median of 99.5% of the reference length, with a median per-patient mean distance of 0.22 mm between matched reference points and centerlines. Annotations, centerlines, code, and weights are available at https://hal9000-lab.github.io/.
human-in-the-loopbenchmark - arxiv:2611.09712 · cs.LGICare-DT: a digital twin for continuous forecasting of intensive care patientsYanke Li, Yahang Qi, Yumi Kim, Diego Paez-Granados
Digital twins of intensive care patients should update forecasts from new data and predict treatment responses, but few studies test whether routine records support either. We propose ICare-DT, a Transformer-based twin that updates a latent patient state hourly from new measurements without retraining. We trained it on 140,473 eICU-CRD stays to forecast vital signs crossing their training-set means. On test data, later forecasts beat the twin's own hour-6 forecasts of the same windows in all 12 settings (+0.073 to +0.153 AUROC). On validation data, controls that withheld or swapped the new hours traced the gain to the patient's own measurements. Without refitting, the model kept its gain on a second database (8 of 9 settings, +0.025 to +0.114 AUROC) in a prespecified evaluation of the previously analyzed MIMIC-IV test split. Refitted within MIMIC-IV, a twin state fine-tuned on clinical endpoints gave gradient boosting a significant gain for acute kidney injury (+0.0035 AUROC; hypotension +0.0010, not significant). To test response prediction, we built a benchmark that replaces the treatments in real records with simulated doses and injects responses of known size, visible only at measurement times. Ridge models supervised with injected effects recovered them from real records only through frequently measured variables (heart rate) read at their measurement times. We asked standard learners to predict these effects for a dosing regime no patient had received. No standard learner outperformed predicting zero effect at an out-of-support query. When told how the simulated responses were generated, a separate learner recovered the effects from the same records without seeing their true values (R2 0.30 to 0.40 at one effect size). Routine records support forecasts that improve with each patient's new data, and our benchmark is a testbed for models that learn how responses are generated.
benchmark - arxiv:2611.09708 · cs.ROA wearable input/output interface learns personalized models of haptic perceptionWen Li, E. Magnus Adams, Zijie Xu, Jiawen Yao +4
Wearable haptic interfaces deliver physical stimuli to the skin, yet perceived locations vary widely across users and diverge from actuator locations, particularly for emergent tactile-feedback techniques. We developed a wearable haptic input/output (I/O) device integrating a 36-channel electrotactile array with capacitive touch sensing, enabling stimulation and user-reported percept locations to be registered within the same body-referenced coordinate frame. In 24 participants, the device revealed electrotactile and referred-thermal perceptual maps with notable interparticipant variability. Despite this variability, both modalities retained electrode-pair-specific spatial structure across participants. Tactile and thermal percepts also showed spatial correspondence, whose strength varied across participants. These findings motivated our human-in-the-loop calibration framework, which adapts a forward-prediction strategy to individual users and generates personalized perceptual models. These models enabled inverse rendering by selecting stimulation patterns predicted to generate a target percept for a specific user. Using these models, we demonstrate personalized haptic rendering in haptic-guided robot teleoperation, visual-to-haptic sensory substitution, interpersonal haptic communication, and virtual reality (VR). Together, this platform provides a route toward adaptive haptic interfaces that measure, model, and render according to each user's perception.
teleoperationtactilehuman-in-the-loop - arxiv:2611.09707 · cs.AIEmpirical Assessment of Post-Quantum Cryptography Adoption Status in TLSMd Ariful Haque, Sander Vinberg, Ben Edwards, João Cruz
Anticipating Q-day security concerns, major CDNs and cloud providers are migrating to post-quantum cryptography (PQC). While existing literature benchmarks PQC performance, empirical research evaluating real-world adoption data remains scarce. We present an empirical assessment of PQC deployment across seven critical protocols (HTTPS, LDAPS, SMTPS, IMAP-SSL, POP3-SSL, TELNETS, and RDP) using 10,000 target IP addresses per protocol. At the origin-server level, PQC algorithm support reaches 9.1% for HTTPS and 7.39% for LDAPS, while falling below 1% for all other evaluated protocols. At the entity level, PQC support is 11% for HTTPS and 5.4% for LDAPS, while the remaining protocols have less than 5% of entities supporting PQC on the hosted servers. We further correlate these adoption trends against organizational digital and financial footprint, and industry sectors, yielding a multi-dimensional empirical assessment of PQC deployment. Finally, at the domain level, we perform a comparative analysis of PQC adoption in Cloudflare-hosted domains against baseline Cloudflare datasets. Collectively, our findings provide critical insights into the current state of PQC readiness over Transport Layer Security (TLS).
benchmark - arxiv:2611.09703 · cs.AIEmpirical Study of Post-Quantum Cryptography Adoption Status in SSH InfrastructureMd Ariful Haque, Sander Vinberg, Ben Edwards, João Cruz
With the anticipated arrival of Q-day, quantum computers threaten current cryptographic schemes for encryption and digital signatures. In response, major content delivery networks and cloud service providers are already adopting Post-Quantum Cryptography (PQC). The current state-of-the-art focuses on benchmarking the operational performance of implementing PQC, but the real-world adoption status remains under-researched. This work utilizes large-scale SSH fingerprint data from a proprietary scanning tool to analyze PQC adoption across 1.52 million hosts and 27,012 entities. Our study shows that the PQC transition in the SSH protocol is underway but highly polarized: 45.2% of SSH hosts on port 22 support PQC, while 54.8% remain classical-only. At the entity level, 39.9% have enabled PQC on at least one host, whereas 60.1% have no PQC support. By correlating PQC adoption with industry, geography, and organizational size (digital and financial), this study provides a comprehensive empirical assessment of PQC adoption status in SSH infrastructure.
benchmark - arxiv:2611.09700 · cs.AIAttend at the Right Resolution: Head-Adaptive Sparse Attention via Query-Key Interaction StructureWajih Hassan Raza, Ahsan Bilal, Yidan Shen, Renjie Hu +2
Attention is a major scalability bottleneck in large language models (LLMs), with its computational cost and key-value (KV) memory traffic becoming increasingly prohibitive as context length grows. Sparse attention addresses this bottleneck by restricting each query to a subset of the context, but its effectiveness hinges on identifying relevant regions both cheaply and reliably. Existing training-free methods primarily rely on runtime relevance estimates, with selection dimensionality and granularity typically governed by fixed heuristics rather than the intrinsic interaction structure of pretrained attention heads. We take a different perspective: the query-key operator of each pretrained head defines a characteristic response structure that determines the resolution needed for sparse selection, both across feature directions and across context tokens. Building on this insight, we propose Head-Adaptive Sparse Attention (HASA), a training-free sparse attention framework that jointly adapts selection resolution along these two dimensions. Along the feature dimension, the singular spectrum of each head's query-key operator determines an effective rank, yielding a compact, head-specific response space. Along the sequence dimension, neighboring regions are progressively merged according to the directional coherence of their token responses in this same space, while heterogeneous regions remain separate. Exact attention is then computed using the original full-dimensional KV states of the selected regions. HASA captures 88.1% of dense-attention mass and, compared with UNIQUE, a strong recent sparse-attention baseline, improves accuracy by up to 3.47 and 2.57 points on RULER and LongBench-Pro, respectively, and accelerates self-attention by up to 1.32x, while achieving up to 14.07x speedup over dense attention.
memory - arxiv:2611.09684 · cs.ROSURE:Spatiotemporal Uncertainty Representation for Efficient Vision-Language-Action LearningJingye Zhang, Choo Sin Wai, Dake Zhong, Zhifang Liu +7
Vision-language-action (VLA) policies have become a promising paradigm for end-to-end robot control, but their effectiveness still depends heavily on large volumes of demonstration data. In manipulation tasks, this dependence is compounded by structured spatiotemporal uncertainty in the action space: multiple nearby actions or trajectories may successfully solve the same task, while semantically equivalent demonstrations often unfold at different speeds. Standard pointwise supervision treats each demonstrated action as an exact target, imposing an unnecessary learning burden under such uncertainty. We present SURE, a data-efficient VLA training framework that explicitly models both spatial and temporal uncertainty. SURE augments a VLA backbone with a lightweight transposed-convolutional probabilistic action decoder that captures spatial uncertainty through local correlations among neighboring action bins. It further introduces Soft-DTW alignment supervision to accommodate local timing variations between predicted and demonstrated action chunks. Together, these two components reduce sensitivity to demonstrator-specific poses and execution timing, enabling more data-efficient action learning. We evaluate SURE on LIBERO, RoboTwin 2.0, and real-world manipulation tasks. Even with only 10% of the demonstration data or 20% of the training steps, SURE achieves manipulation success rates that match or surpass baseline VLAs.
vision-language-actionvlamanipulationliberorobotwin - arxiv:2611.09683 · physics.opticsVertically-coupled microring resonators for lithium niobate on insulator integrated photonicsClément Selva, Sarah Umme Tabassum, Kumar Kinjalk, Adrian Laborde +3
We propose and demonstrate lithium niobate on insulator (LNOI) microring resonators that are verticallycoupled to index-matched silicon-rich silicon nitride (SiN x ) access waveguides as an alternative to conventional lateral coupling schemes for integrated photonics applications. The wafer-scalable fabrication process involves a planarization step exploiting a single-step and reproducible spin-coating of a SU-8 epoxy spacer layer which allows a controlled separation to be established between the resonator and its access waveguide. The resulting multimode and single-mode microring resonators exhibit, for their fundamental TE 0 mode, intrinsic quality factors of 4.0 x 10 5 and 3.4 x 10 5 , respectively.
microring - arxiv:2611.09679 · cs.ROAutonomous Particle Image Velocimetry Using Zero-Shot Sim-to-Real Reinforcement LearningChang Rajani, Luke Sebastien Marsden, Stephan Kunz, Claudio Mucignat +1
Particle image velocimetry (PIV) reconstructs a flow field from the motion of tracer particles but the reconstruction quality is bounded by how well the flow structure of interest is resolved in the images. The acquisition position is typically fixed by the operator before the measurement. We cast camera placement as a sequential decision-making problem and train a reinforcement learning agent to control a 6-DoF manipulator equipped with a camera, enabling it to locate, frame and resolve a hydrofoil wake in a water tunnel. The agent commands the camera pose and observes the scene only through the velocity field and uncertainty estimated from time-separated image pairs. It is trained entirely in a photorealistic simulation of our water tunnel, with wakes generated by Computational Fluid Dynamics (CFD) simulations of hydrofoils with varying profiles, positions and angles of attack. After zero-shot transfer to the real water tunnel, the policy reliably locates and frames the wakes of randomly placed hydrofoils. It resolves the wake at a higher resolution than a fixed camera view and with lower error in the measured flow statistics, in roughly half the time of a full scan and with far fewer image pairs.
manipulatorsim-to-realagent - arxiv:2611.09672 · cs.ROODEAction: From Predicting Chunks to Evolving ActionsQishen Wang, Pengxiang Ding, Shifeng Bao, Mingyang Sun +13
Vision-language-action (VLA) models typically generate actions in chunks, organizing continuous robot execution into separately predicted segments and thus introducing potential discontinuities at action-chunk transitions. We propose ODEAction, a chunk-free action generation paradigm that replaces action-chunk prediction with Continuous Action Evolution. Given visual observations, language instructions, and robot proprioceptive information, ODEAction learns an action transition field defined over action space, where a transition vector characterizes the evolution rate of the current action with respect to evolution time. During execution, the model takes the current action as its initial condition and integrates the learned field using an ordinary differential equation (ODE), recursively generating subsequent actions with explicit dependencies. ODEAction is implemented as an action head and does not depend on a specific vision-language backbone architecture. Experiments on LIBERO, EBench, RoboTwin, and real robots show that ODEAction improves both task performance and trajectory quality over matched action-chunking baselines, with particularly pronounced gains on fine manipulation tasks.
vision-language-actionmanipulationaction headliberorobotwin - arxiv:2611.09665 · cs.LGOnline Causal Configuration for Networked Systems via Doubly Robust Steady-State LearningYuli Liu, Zhiheng Zhang
Measurement exploration reveals how interacting local controllers behave but can alter the steady-state regime a network operator seeks to optimize. Exploratory rewards may therefore fail to identify a configuration's no-exploration value: under full exploration, identical logged distributions can conceal different target values. Identification requires accounting for both action-law mismatch and shifts in the steady-state context distribution. We introduce CAUSAL-UCB, which selects configurations using optimistic bounds on target causal values. Its interference-aware clipped doubly robust evaluator produces predecision certificates separating statistical variation from calibrated identification, transport, and temporal-dependence errors. For any finite horizon and predetermined schedule of measurement-exploration rates $\{η_t\}$, we prove lower bounds of $Ω((1+\sum_tη_t)^{-1/2})$ on minimax inference error and $Ω(\sum_tη_t)$ on measurement-induced deployment loss in a stationary, potentially slow-mixing family. We bound regret under valid certificates and selected-information growth, and establish an anytime finite-sample certificate with a shared radius for estimation, selection, and reporting in a restricted known-local-interference model. In a high-mismatch benchmark with 24 paired seeds and 200 rounds, CAUSAL-UCB attains target regret of $1.732$ versus $6.326$ for certified IPS-UCB, with deployment gaps of $30.607$ and $23.217$, respectively. Independent-audit coverage tests and trace-driven semi-synthetic studies further assess certificate coverage and deployment trade-offs.
benchmarkevaluator - arxiv:2611.09662 · cs.AIHow Much Conversation History Does an AI Agent Need?Danny Kim, Fernando Koch
How does using less conversational history change reader-token use and benchmark scores? We examine this question on 1,100 BEAM questions. We held the retrieval ranking and reader constant across five fixed passage budgets, and mean assigned scores rose as the passage budget grew. The score gain per additional token became smaller at the upper tested budgets. Answerable questions generally received higher scores with more context. Pooled abstention scores sometimes rose as the passage budget shrank. We also tested message retention in a controlled task with explicit conversational-state targets and compared retained user messages with token-matched recency. Qwen matched the full state more often for direct requirements when user messages were retained. Its full-state matching declined when answers depended on references to earlier offers. Mistral produced near-floor results across the retention conditions. A scoring audit leaves the correctness of BEAM answers unresolved. These experiments characterize the token and score consequences of the tested memory budgets. Reliable budget selection depends on retaining the evidence each question requires and validating the measure used to assess answer quality.
memoryagentai agentbenchmark - arxiv:2611.09658 · cs.AIDifferentially Private GraphRAG: Protecting Relation Connectivity and SemanticsJie Fu, Yuan Hong, Yue Ning, Wendy Hui Wang
Graph Retrieval-Augmented Generation (GraphRAG) connects information across documents through entity relationships to support relational reasoning. However, black-box access to GraphRAG systems can enable adversaries to reconstruct knowledge graphs and recover both structural connectivity between entities and the semantics of their relationships. Existing defenses lack formal privacy guarantees and provide limited protection against such reconstruction attacks. We propose RelGuard, the first differentially private framework for GraphRAG, which provides formal privacy guarantees by mitigating connectivity leakage and protecting relations and their semantic meaning through relation removal. We formalize relation differential privacy (relation-DP), which bounds the distinguishability of mechanism outputs according to the semantic distance between alternative relations while accounting for relation removal. Building on this definition, we develop a four-step defense mechanism that sanitizes source documents before GraphRAG indexing by perturbing relation predicates or removing their associated connectivity using the exponential mechanism. Experiments across three GraphRAG systems and two targeted reconstruction attacks demonstrate that RelGuard provides stronger protection against reconstruction attacks than four existing defense mechanisms while better preserving downstream utility.
retrieval-augmentedknowledge graph - arxiv:2611.09650 · cs.LGTensorMutation: Learning Weight-Space Mutations of Deep Learning Models on Neural GraphsZaheed Ahmed, Philip Makedonski, Jens Grabowski
Mutation testing for deep learning (DL) faces a cost-realism trade-off. The pre-training approach produces realistic but computationally expensive mutants, whereas existing post-training operators modify trained models efficiently but produce mutants that less closely resemble real faults. This study asks whether the trade-off is necessary: can the weight-space effect of a fault-informed pre-training mutation operator be predicted from a trained model, without retraining it? We answer this question with TensorMutation, which represents a trained model as a neural graph and uses a mutation-conditioned graph neural network (GNN) to predict parameter changes. The GNN is trained offline and reused across unseen models so that TensorMutation can generate executable mutants without retraining. We construct a corpus of 25,252 original-mutant pairs from 71 multilayer perceptron (MLP) classifiers using 11 fault-informed pre-training operators, including 25 newly curated subjects. TensorMutation partially reproduces the behavioral effects of pre-training mutations on nine held-out subjects, and direct weight-space supervision significantly improves weight-space fidelity. On 12 real-fault subjects, TensorMutation achieves the highest median behavioral similarity among the evaluated approaches. The improvement over the post-training tool DeepMutation++ is statistically significant (Holm-adjusted p=0.018), whereas the differences from the pre-training tool DeepCrime and the norm-matched random control are not significant. These results show that fault-informed pre-training mutations leave learnable structure in weight space that transfers to unseen models.
post-training - arxiv:2611.09644 · cs.MAA Separation Principle for Cooperative Multi-Agent Reinforcement LearningLucia Pezzetti, Nicolas Lanzetti, Antonio Terpin, Florian Dörfler +1
We study cooperative multi-agent reinforcement learning (MARL) problems in which agents evolve under decoupled noisy dynamics but share a population objective. A canonical example is a mobility operator that orchestrates a fleet of vehicles to meet customer demand. The stochastic and time-varying nature of these problems calls for learning approaches, yet existing MARL algorithms do not scale: learning assignment and routing jointly over the full state-action space becomes prohibitive as the number of agents grows. Formulating the problem over the space of probability measures, we prove a separation principle: the population cost-to-go function is upper bounded by an optimal transport problem whose transportation cost is the target-conditioned single-agent cost-to-go function. This simplifies the multi-agent problem to learning a single-agent policy and coordinating the agents via optimal transport. Building on this principle, we introduce SALT (Separation-based Assignment and Learning via optimal Transport), a family of algorithms that learn a target-conditioned single-agent policy using any standard reinforcement learning algorithm, and then coordinate the agents via optimal transport. Since learning occurs at the single-agent level, SALT scales to fleet sizes where MARL baselines become intractable and transfers without retraining to fleet sizes unseen during training. In a vehicle-delivery experiment on the southern Manhattan road network, where MARL baselines are computationally prohibitive, SALT improves travel time by 15-21% over a shortest-path routing heuristic. In stochastic grid worlds, SALT reduces average travel times by 23-61% relative to MARL baselines and scales to population sizes where these methods fail.
multi-agent - arxiv:2611.09634 · cs.LGSteerable Multi-Objective Group-Relative Preference Optimization with Jointly Trained Low-Rank AdaptersConstantinos Karouzos, Nikolaos Aletras
Language model deployment often requires different trade-offs between competing objectives as user preferences change, yet standard post-training methods typically produce a single policy for a fixed objective weighting. We introduce LambdaMix, a new parameter efficient method for learning a preference conditioned family of policies with group-relative online reinforcement learning. LambdaMix jointly trains one low-rank adapter per objective over a frozen base model. One preference vector selects the adapter mixture that generates responses and weights the separately normalized rewards used to update the adapters. A single training run supports changing preferences at inference without further optimization. We experiment with four models on helpfulness and harmlessness and on solving math problems while following instructions. We also evaluate the two largest models on rewriting comments to preserve wording while reducing clues about personal attributes. On the first two tasks, LambdaMix achieves higher hypervolume than all evaluated conditioned and policy merging baselines across all four models. LambdaMix uses fewer training completions and stores fewer adapter parameters than a set of GRPO specialist policies. With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone.
post-training - arxiv:2611.09625 · cs.LGEncoding Memory in World Models with Test-Time Model-Based Reinforcement LearningMaxime Burchi, Fayaz Ali Dharejo, Radu Timofte
Model-based reinforcement learning (MBRL) approaches have demonstrated strong performance across diverse domains in recent years, learning accurate world models from agent trajectories. However, despite this progress, current approaches still struggle to surpass human performance on tasks that require long-term memory and planning. In this work, we address the problem of long-term memory in MBRL by moving beyond the commonly studied paradigm of architecture design. We introduce Test-Time Model-Based Reinforcement Learning (TT-MBRL), using gradient descent at test time to encode the memory of episode observations into world model parameters. This enables the agent to learn a distinct world model for each episode, improving memory capabilities compared to traditional approaches that solely rely on contextual memory. We first establish a proof of concept on a simplified two-dimensional variant of the Memory Maze environment, which we refer to as Memory Maze 2D, showing consistent improvements across different world model architectures. We then demonstrate the effectiveness of TT-MBRL in the 3D domain, achieving new state-of-the-art results on the Memory Maze benchmark. Finally, we show that TT-MBRL maintains strong performance on non-memory benchmarks such as DeepMind Control and Atari 100k.
world modelmemoryagentbenchmark - arxiv:2611.09617 · cs.CVCRTrack-Cap: A New Benchmark for Cross-View Dense Video Object CaptioningXiangqun Zhang, Likai Wang, Zekun Qian, Rui Zhou +2
Dense video object captioning (DVOC) localizes, tracks, and describes object trajectories within a video. Existing methods process each camera view independently, producing separate camera-local tracklets and captions for the same person across cameras. These outputs lack a unified description and leave complementary appearance evidence across views unused. We therefore formulate cross-view DVOC (CVDVOC), which associates these tracklets into scene-level global trajectories and generates one caption for each trajectory from synchronized views. To support this task, we construct CRTrack-Cap from CRTrack by auditing text-identity links and converting usable referring expressions into canonical captions. It retains 425 global trajectories and 838 local tracklets across 13 scenes, with canonical captions for 335 identities. We adapt CHOTA to the cross-view setting as CV-CHOTA to jointly evaluate tracking and caption quality. Our reference framework combines View-Exclusive Cross-Camera Association (VECA) for grouping tracklets with Conflict-Evidence Gating (CEG) for integrating caption evidence using its reliability and disagreement across views. Against three single-view methods adapted with MvMHAT for cross-view association, our framework achieves CV-CHOTA scores of 83.43 and 33.48 on the scene-disjoint in-domain and cross-domain tests, respectively. Together, CRTrack-Cap and our framework provide a benchmark and a strong baseline for describing the same person consistently across camera views.
benchmark - arxiv:2611.09614 · cs.LGKnow What You Need: Efficient Activation CheckpointingDanyal Ahmed, Edouard Oyallon
Large Language Model training is increasingly constrained by memory capacity. Activation Checkpointing reduces activation memory by avoiding storing selected activations during the forward pass and rematerializing them during backpropagation, thereby trading memory savings for additional computation. Although this technique is essential for scaling model training, rematerialization can be prohibitively expensive, particularly for Transformer linear layers. Approximate computation could reduce this overhead, but conventional approaches either introduce errors that accumulate over the backward pass or fail to exploit information available from the initial forward pass. We introduce a novel approximate rematerialization method that can yield unbiased gradient estimates: using information collected during the initial forward pass, the method selectively recomputes only the most important layer components, either through greedy selection or randomized sketching. Ablation studies show that the resulting gradients closely match the exact gradients. When applied to feed-forward layers, attention layers, or both, our method can recompute only two-thirds of the layer components in a Llama-488M model, while increasing perplexity by just 1.7%, 1.2%, and 2.5%, respectively.
memory - arxiv:2611.09605 · cs.RODKWBC: Decoupling Motion Tracking and Balancing for Humanoid Whole-Body ControlDingkun Wei, Zehong Shen, Huaijin Pi, Xueyang Wang +4
Humanoid whole-body control requires a robot to track diverse reference motions while maintaining balance under gravity, torque limits, contact switching, and external disturbances. Existing reinforcement-learning-based trackers typically learn control outputs that jointly address motion tracking and balance. However, similar reference poses can correspond to substantially different dynamic demands, requiring a unified policy to learn motion-dependent control requirements alongside balance corrections. This can complicate learning and lead to high-frequency fluctuations in control outputs. We propose DKWBC, a dynamics-informed whole-body control framework that combines explicit model-based tracking with a learned balancing residual. Inspired by the model-based and environmental-wrench contributions to joint torques in floating-base dynamics, DKWBC uses inverse-dynamics feedforward and joint PD feedback to supply tracking torques, while a residual policy learns additional joint-torque corrections for balanced execution. By explicitly accounting for motion-dependent torque demands, this structure simplifies the policy's learning task. Experiments in simulation and on real hardware demonstrate that a single learned balancing policy generalizes across motion datasets while maintaining accurate motion tracking and balance across ground support, attached payloads, and physical interaction. Project page: https://robbyant-research.github.io/DKWBC/.
humanoidwhole-body control - arxiv:2611.09598 · cs.LGTransformer-FSQ: Discrete Latents for Extreme KV-Cache CompressionAndras Szecsenyi, Thomas L. Lee, Vaisakh Shaj, Elliot J. Crowley +1
Private, personalised language models must run on local consumer devices, where large, frontier models cannot fit, and retaining long contexts is non-trivial. Sequential memory---the key-value (KV) cache---grows linearly with context length, quickly dwarfing the model itself: for Qwen3-1.7B-Base, 200,000 retained entries require 21 GiB, 6.6 times more than the model weights. Yet the KV-cache compression methods we measured fail beyond moderate compression ratios. We introduce Transformer-FSQ, addressing extreme KV-cache compression by replacing standard key and value projections with a finite scalar quantisation bottleneck. Our cache stores a packed vector of discrete integer codes per token, with no per-token quantisation metadata. Converting a pretrained transformer trains only these projections on 0.25 tokens per original parameter. We evaluate Transformer-FSQ against five baseline families on models from 0.6B to 8B parameters. On long-context retrieval (RULER), from 12$\times$ to 97.5$\times$ Transformer-FSQ is the most accurate method we evaluate, across scales. A fused Triton kernel decodes directly from the integer codes; with baselines given their authors' kernels where available, it decodes at $1.01\times$ (95% interval $0.98$--$1.05$) the throughput of native SDPA attention on the uncompressed Qwen3-8B-Base at $32\times$ and $1.54\times$ at $97.5\times$. Extreme KV-cache compression is therefore a practical way to extend how much context a fixed device can hold.
long-contextlong context - arxiv:2611.09597 · cs.CLRuACT: Recovering Inactive Rubrics as Adaptive Critics and TeachersYijun Li, Zhengxi Lu, Zhipiao Liu, Jinzheng Yu +3
Rubric-based reinforcement learning (RL) extends RL with verifiable rewards to open-ended tasks by scoring each response against query-specific rubrics. Under group-relative optimization, however, a rubric that every sampled response fails (unreachable) or passes (saturated) gives all responses the same outcome and no learning signal; in Qwen3-4B medical training, 60.88% of query-rubric observations are inactive. We find that the two states lack different supervision: stating an unreachable rubric in the prompt raises its satisfaction from 1.13% to 91.25%, whereas responses that all pass a saturated rubric still differ in quality. We therefore propose RuACT (Rubrics as Adaptive Critics and Teachers), which assigns each inactive rubric the role that supplies what it lacks. Rubric as Teacher uses unreachable rubrics as privileged information for reward-gated on-policy self-distillation, and Rubric as Critic decomposes saturated rubrics into child rubrics that separate current responses while sharing the parent's weight. RuACT outperforms GRPO, rubric reweighting, and rubric-guided exploration on all four medical and scientific benchmarks with Qwen3-4B and achieves the best average with Qwen3-1.7B. On initially unreachable rubrics, unguided recovery rises from 35.02% to 42.86%, while child refinement improves reward-quality alignment on independent rollouts from 57.74% to 60.77%. Code is available at https://github.com/ZJU-REAL/RuACT.
benchmark - arxiv:2611.09596 · cs.LGCoordinatewise Epsilon for Adam under Heavy-Tailed NoiseJeonseong Kim, Chenglin Fan
Adam estimates second moments coordinatewise, but typically adds the same scalar $\varepsilon$ to every denominator. We study how this shared $\varepsilon$ affects convergence when noise and curvature vary across coordinates. Under coordinatewise finite-$p$ noise moments, $p\in(1,2]$, choosing $\varepsilon_i$ from the noise and curvature bounds for each coordinate attains the heavy-tailed $\ell_1$-stationarity benchmark $O(T^{-\frac{p-1}{3p-2}})$, matching the known optimal time exponent for smooth nonconvex stochastic optimization. For $1<p<2$, Adam with a shared $\varepsilon_t$ admits a lower bound of order $Ω(T^{-\frac{p-1}{2p}})$. This lower bound holds for every predictable scalar choice of $η_t$, $β_{1,t}$, $β_{2,t}$, and $\varepsilon_t$, precluding a dimension-uniform benchmark rate for this class.
benchmark - arxiv:2611.09570 · cs.CVVideo Stitching from Multiple Moving CamerasShira Ifergane, Roy Amoyal, Shahar Benishay, Oren Freifeld
We address geometric video alignment for the same scene captured by multiple freely moving cameras. Existing methods, which are based on image stitching or video stabilization, are typically limited to two cameras or must assume the restrictive setting of a fixed linear ordering of the cameras. We propose MViS, a lightweight test-time optimization framework for joint multi-view alignment with temporal consistency. MViS uses a spatial-transformer-inspired net to jointly process frames from all views within a short temporal window and predict per-view, per-time homographies. Optimization proceeds in two stages: geometry-guided alignment followed by appearance-based refinement, with both stages enforcing multi-view and temporal consistency across the sequence. The method requires no scene-specific hyperparameter tuning, handles arbitrary and time-varying view-intersection graphs, and remains robust under rapid camera motion. On a standard two-view video benchmark, MViS matches or surpasses the state of the art. To facilitate evaluation in more complex scenarios, we introduce an annotated dataset with more than two freely moving cameras and establish the first benchmark for this setting. Project page: https://bgu-cs-vil.github.io/MViS-project/ .
benchmark - arxiv:2611.09568 · cs.AIAlignment by Confident Debate: Preference Learning from Multi-Agent DebateNikhil Verma, Dingxian Wang, Mohammad Reza Rezaei, Andrew Rabinovich +1
Post-training language models for reasoning typically rely on gold answers, task-specific verifiers, or learned reward models, limiting scalability when such signals are scarce. Multi-agent debate offers a self-supervised alternative, but converting debate into reliable training supervision remains challenging: majority voting weights agreeing responses equally, lacks a principled tie-breaking rule, and is ill-defined for open-ended tasks without a canonical answer; likelihood-based scoring can conflate fluency with correctness; and homogeneous committees may reinforce correlated errors. We introduce Alignment by Confident Debate (AbCD), which converts debate traces into preference data without gold answer labels. A sampling-diverse committee debates each problem, after which final answers are clustered by task-level semantic equivalence. We use verifiable tasks (mathematical reasoning, multiple-choice quantitative QA, and executable code-generation) as a controlled test bed: their answer keys or executable checkers let us evaluate the quality of debate-derived supervision against ground truth, even though the oracle itself never uses gold labels. Post-training on the resulting preferences outperforms inference-only ensembling and debate, as well as SFT on the same traces, across five model families spanning 1B to 8B parameters and four datasets, improving consensus accuracy, with a maximum gain of $21.5_{pp}$. Ablations show that the approach transfers to a larger model and to code generation, while also revealing limitations of heterogeneous committees and agreement-based weighting. Overall, the results suggest that debate traces can yield useful preference supervision when answer equivalence is combined with confidence-weighted disagreement.
multi-agentpost-training - arxiv:2611.09567 · cs.LGTurtLES: A Large-Scale Benchmark for Turbulent 3D Neural PDE SurrogatesJarno Platenburg, Armand Kassaï Koupaï, Paola Cinnella, Patrick Gallinari
Machine-learning surrogates are increasingly used to accelerate computational fluid dynamics, yet progress is limited by the lack of benchmarks capturing realistic, time-dependent turbulent flows. We introduce TurtLES, a 14 TB dataset of 3D turbulent cylinder wakes in the shear-layer transition regime, computed via high-fidelity implicit large-eddy simulations. The dataset contains 400 trajectories of 400 time steps each, with 3-11 million points per frame, spanning variations in geometry, Reynolds number, and angle of attack. Unlike existing datasets, TurtLES combines (i) physically realistic turbulence with an active energy cascade driven by vortex stretching, (ii) dense irregular meshes, and (iii) long-horizon temporal dynamics. We benchmark state-of-the-art neural operators on autoregressive rollouts and identify key failure modes, including temporal instability, loss of small-scale energy, and challenges in learning on large irregular domains. TurtLES provides a new benchmark for evaluating surrogate models on 3D turbulent wakes at industrially relevant mesh resolution.
benchmark - arxiv:2611.09560 · cs.LGSolver-Free Decision-Focused Learning via Neural Optimization Layers for Edge SelectionXinda Zheng, Frits de Nijs, Edward Lam
Decision-focused learning trains prediction models to improve downstream decisions, but many methods require costly solver calls or optimal-solution labels. We propose a solver-free decision-focused learning framework for NP-hard edge-selection problems with uncertain parameters in the objective, the constraints, or both. A neural optimization layer is first trained without solution labels, using problem objectives and constraint penalties. Graph-informed edge scoring and iterative refinement encourage concentrated and structurally consistent continuous decisions. The layer is then frozen, and a prediction model is trained through its continuous output using a problem-specific loss evaluated under observed parameters. At deployment, the predictor, optimization layer, and a lightweight decoder produce decisions without solving the original NP-hard problem. Experiments on the asymmetric traveling salesperson and endpoint-free orienteering problems show consistently better decision quality than prediction-focused learning and improvements over most evaluated decision-focused baselines. Our framework requires less training time than the evaluated decision-focused baselines, even when optimization-layer pretraining is included. Neural deployment also produces decisions faster than solver-based deployment, offering a better quality-efficiency trade-off. The layer also reduces training cost in existing decision-focused methods when used as an approximate optimization oracle.
iterative refinement - arxiv:2611.09554 · cs.LGQuery-Guided RoPE Modulation for Lost-in-the-Middle Mitigation in Long-Context LLMsPeiyu Zhang, Heng Ping, Shinibali Bhattacharyya, Anwesan Pal +5
Long-context language models often neglect information in the middle of the input, a phenomenon known as "lost-in-the-middle". Part of this gap may stem from the long-range decay of rotary position embedding (RoPE), which weakens the positional signal between distant tokens. Existing input compression deletes low-importance tokens, and attention reweighting acts only after RoPE has weakened this signal. RoPE-level methods modify positions directly but apply the same remapping to every query. To address these limitations, we propose QGPM (Query-Guided Positional Modulation), a training-free method that reshapes RoPE positions using query relevance from the model's own attention heads. QGPM builds a broad relevance map for lower layers and a sharp one for upper layers. Low-relevance tokens advance by less than one position on fast-rotating planes, pulling relevant evidence toward the query. Slow-rotating planes keep the original layout, with every token in place. Experiments on NoLiMa and RULER with multiple context lengths show that QGPM improves over base models and strong baselines, with the largest gains on retrieval tasks, and generalizes across multiple backbones. Our code is available at https://github.com/iZHpy/QGPM.
long-context - arxiv:2611.09528 · cs.AIStockASkill: A Local-First AgentSkill for Intelligent Stock SelectionXiangjing An, Zhibin Jiao, Aiting Chen
Quantitative equity selection and portfolio construction in emerging and cross-border markets frequently suffer from computational bottlenecks, remote API fragility, survivorship bias, and rigid execution constraints such as discrete board lots and investor-mandated drawdown limits. In this paper, we present StockASkill, a local-first, full-data-driven AgentSkill engineered as a specialized Quant Co-Processor for autonomous AI agents and quantitative researchers. Tailored for agile portfolios (10^5 ~ 10^7 RMB) where discrete lot execution and capital protection are mandatory, StockASkill embeds a columnar DuckDB OLAP engine over 42.44 million historical price records and 186,432 financial disclosures spanning 26 years across China A-shares, Hong Kong, US equities, and ETFs. By eliminating ad-hoc network dependencies during analysis, StockASkill reduces cross-sectional screening latency across 5,400+ equities from over 25 seconds to sub-0.5 seconds (a 95%+ reduction). We formalize an end-to-end mathematical pipeline integrating a 7+1 factor alpha model, an SLSQP convex optimizer under smooth CVaR drawdown surrogates with exact path projection, dynamic 100-share board lot integer quantization, and an autonomous thesis lifecycle finite-state machine. Furthermore, we provide a zero-disk, strongly-typed JSON streaming interface designed for high-throughput LLM tool-calling. Open-sourced at https://github.com/axjing/stockaskill, comprehensive empirical benchmarks from 2018 to 2024 demonstrate that the architecture achieves annualized Sharpe ratios exceeding 1.48 and Calmar ratios up to 2.17, while reducing max drawdown by up to 85% relative to benchmark indices.
ai agentbenchmark - arxiv:2611.09527 · cs.CVExQuery: Explicit Structured Query Priors for Temporal 3D Object DetectionHonglin Sun, Yifei Zhang, Junyang Zhang, Kenji Hashimoto
Sparse query-based 3D detectors commonly encode heterogeneous semantic, geometric, and motion information into unified latent queries, including when explicit 3D anchors are used. We introduce ExQuery, an explicit structured query framework for temporal 3D object detection. Each decoded query is organized into a shared semantic context slice and physical-state-specific slices for center, dimensions, orientation, and velocity. This structure is preserved from prior encoding to branch-specific regression and is reused to reconstruct temporal queries: predicted physical states are transformed by ego motion, classification scores form soft semantic priors, and the reconstructed query is fused with the stored latent instance feature through gated affine modulation. Experiments on the nuScenes benchmark show that our method achieves competitive detection performance, and ablation studies validate the effectiveness of our proposed modules.
benchmark - arxiv:2611.09522 · cs.AIAtik: A Shared-Mesh Transformer Accelerator for RISC-VAhmed Zeer, Erkan Uslu
Transformer inference repeatedly invokes dense matrix multiplication and attention, including its data-dependent softmax normalization. We present Atik, an open-source RoCC accelerator for RocketChip that executes matrix multiplication and dense attention through one shared fixed-point MAC mesh. Atik accepts and returns BF16 tensors, uses a streaming online-softmax schedule that never materializes the full score matrix, and is invoked through a small descriptor-based ABI. We evaluate three meshes cycle-by-cycle on FireSim and report standalone Sky130 physical-design area. Across 137 logged FireSim test records, all tests pass their benchmark tolerance with zero reported output mismatches. Relative to the same scalar Rocket BF16 reference, the 8x8 configuration reaches 31.5x peak matmul speedup and 225.8x peak attention speedup. Hybrid projections for the single-layer workload-derived C models, using directly measured accelerator-kernel cycles, give aggregate speedups of 44.2x, 77.1x, and 32.3x for ViT, TinyBERT, and GPT-2 prefill, respectively; these are not full-model hardware replays.
benchmark - arxiv:2611.09521 · cs.CVEmbedding Predictive Adaptation and Refinement for Multi-Modal Industrial Anomaly DetectionJaeho Cho, Jinsik Bang, Taehwan Kim
Industrial Anomaly Detection (IAD) has evolved from 2D image-based approaches to multi-modal methods that leverage the complementary nature of RGB images and 3D point clouds. Among these, memory bank-based methods, which identify anomalies by the discrepancy between each query patch feature and its nearest normal prototype, have proven particularly effective. However, two limitations remain: (i) employing frozen pretrained encoders yields insufficient discrimination between normal and anomalous features due to the domain gap, while existing contrastive learning-based adaptation struggles to capture the spatial relationships required for precise localization; and (ii) pixel-level anomaly maps interpolated from patch-level distance scores remain coarse, lacking fine-grained detail. To address these limitations, we propose EPAR (Embedding Predictive Adaptation and Refinement). We adapt encoders by predicting the embeddings of masked regions to capture spatial relationships, together with patch-level Variance-Invariance-Covariance (VIC) regularization, resolving the poor discriminability between normal and anomalous features caused by the domain gap. We further propose a Noise Suppression Network that refines coarse maps by learning the relational discrepancy between each patch feature and its nearest normal prototype, rather than the anomaly's appearance. This relational signal remains valid for unseen real defects, despite being learned from synthetic anomalies. EPAR demonstrates strong practical applicability, excelling in few-shot settings and improving AUPRO@1% on MVTec 3D-AD and Eyecandies by +1.8% and +1.9%, respectively.
memory - arxiv:2611.09516 · cs.CLBenchmarking Speaker Embedders in the Sub-Second Regime for Real-Time Voice AI AgentsAbhishek Gautam
Real-time Voice AI agents must separate the active user from intelligible background speech that survives conventional noise suppression. Background sources include television audio, family members, and colleagues. Production denoisers remove non-stationary environmental noise effectively. Background speech shares the spectral characteristics of the target and passes through, causing false acceptances in downstream speaker verification (SV) and automatic speech recognition (ASR). We benchmark open-source speaker embedders TitaNet-Large, TitaNet-Small, ECAPA-TDNN, CAM++, and SpeakerNet under a strict sub-second latency budget. Evaluation combines a controlled benchmark with a weighted human-annotated production set. Our controlled benchmark is built from real calls in which a second real speaker is injected into silence intervals at modulated amplitude. We report three findings. First, denoising before embedding is generally harmful and raises equal error rate (EER) across architectures. Second, truncating audio below the full voice-activity segment degrades every model. Third, roughly half of the TitaNet embedding is redundant for this task. Trailing dimensions encode population attributes such as language and gender. Such attributes carry little discriminative value in a demographically homogeneous user base. We distill these results into deployment guidance. Verification should run on raw audio with full voice-activity chunks. Thresholds should favor precision over F1. Truncated embeddings reduce the cost of similarity search.
ai agentbenchmark - arxiv:2611.09511 · cs.CVV-STORM: A Structure-Aware Grounded Chain of Thought Dataset for Faithful and Fine-Grained Spatio-Temporal Video ReasoningDian Jiao, Yuqian Yuan, Jiaqi Zhu, Mingjian Gao +8
While MLLMs can often identify what happens in a video, they still struggle with fine-grained spatio-temporal reasoning, which requires precise grounding when and where the supporting evidence appears and composing such evidence into a coherent reasoning chain. This challenge is exacerbated by the lack of scalable data construction and evaluation pipelines that provide granular, verifiable supervision linking reasoning steps to explicit spatio-temporal evidence.To address this gap, we present V-STORM, a structure-aware framework that transforms videos into Spatio-Temporal Scene Graphs (STSGs) and derives Grounded Chain of Thought (GCoT) from validated evidence paths, making each reasoning step traceable to structured and verifiable spatio-temporal evidence.Under this framework, we construct a 260K-sample training set and a 3K-sample human-verified benchmark, both featuring GCoT annotations that link reasoning to explicit spatio-temporal evidence.Experiments reveal a clear separation between evidence localization and evidence composition: current MLLMs can often identify relevant spatio-temporal cues, yet still fail to integrate them into correct multi-step reasoning, exposing a bottleneck that answer-only evaluation largely obscures. Moreover, fine-tuning with V-STORM substantially improves both answer accuracy and grounding faithfulness, validating its effectiveness as supervision for faithful and evidence-grounded video reasoning. Our dataset, benchmark, and source code will be publicly released.
scene graphbenchmark - arxiv:2611.09505 · cs.LGBEiTScore++: Reference-free Image Captioning Evaluation with BEiT-3 and Conformalized Quantile RegressionGonçalo Gomes, Bruno Martins, Chrysoula Zerva
Reliable uncertainty estimation is important for deploying evaluation metrics in real-world settings, where both input ambiguity and domain shifts are common. In this work, we propose an uncertainty-aware extension of the BEiTScore image captioning metric, producing conformally calibrated confidence intervals alongside alignment scores. Our approach models predictive distributions over human judgments, enabling both accurate scoring and principled uncertainty quantification in a single forward pass. We evaluate our BEiTScore++ metric across multiple benchmarks with human-annotated ratings, showing that it preserves the ranking correlation of BEiTScore, while producing intervals whose width tracks the prediction error more closely than Monte Carlo dropout on most benchmarks. The estimated uncertainty also grows under image degradation and on out-of-domain artwork images, making BEiTScore++ a practical tool for uncertainty-aware evaluation of image-caption alignment.
benchmark - arxiv:2611.09502 · cs.LGTimeOmni-Live: Continuous & Live Time Series Understanding with Long-Term MemoryTong Guan, Alexander Chemeris, Yiqing Guo, Warren Jin +3
Existing time series language models predominantly follow an offline, one-shot paradigm: they process fixed-length inputs independently and do not retain past states. This static approach fundamentally misaligns with the continuous nature of real-world streaming applications and fails to maintain the historical context necessary for cross-window reasoning. To bridge this gap, we introduce Continuous & Live Time Series Understanding (CL-TSU), a novel paradigm that shifts the focus toward maintaining a persistent, stateful understanding over open-ended data streams. To systematically evaluate this paradigm, we present LiveTS-Suite, the first comprehensive data suite tailored for CL-TSU, supporting data generation, model training, and systematic evaluation across 18 tasks spanning six core capabilities. It comprises over 5K streams totaling 2.2B timesteps, annotated with 15K specific response events. We further design an evaluation protocol that jointly assesses response timeliness, content accuracy, and computational efficiency. Building on this foundation, we propose TimeOmni-Live, the first live time series understanding model. It integrates a hybrid time series encoding for natively temporal input, a stream-persistent hybrid memory mechanism to effectively preserve exact historical evidence and long-term task states, and an elastic stream-inference synchronization strategy to reduce processing lag. Evaluations show that TimeOmni-Live establishes a strong baseline for CL-TSU, outperforming frontier LLMs (90.0 vs. 79.3 for Opus 5) with only 4B parameters and achieving 206$\times$ lower p95 latency than its backbone. Deployment on a Jetson edge device further shows its ability to support real-time continuous temporal understanding locally at the edge.
memoryevaluation protocol - arxiv:2611.09472 · cs.AIENGRAM: From Agent Experience to Weights through Verified ConsolidationRuiyi Yang, Zechen Li, Lihuan Li, Du Yin +2
Language-model agents use experience through external retrieval and parameter learning. Consolidating stored experience into model weights requires deciding both which experience to train on and when its external records can leave retrieval. Admission based solely on student mastery can exclude useful teacher demonstrations, while removal immediately after training can discard guidance that the model still needs. We introduce ENGRAM, a framework that separates training admission from memory release. A demonstration-aware gate admits skills based on student mastery or sufficient successful teacher demonstrations. The student learns from selected action traces with context from an Engram Trace Graph (ETG). A verification gate then compares the updated model without episodic retrieval against the base model with retrieval on reserved probes. Release requires valid training participation and scores meeting both relative and absolute criteria; records for skills that fail verification remain retrievable. Experiments across interactive and reasoning benchmarks show consistent gains over frozen-weight ETG controls, with ablations supporting demonstration-aware admission and selective adapter activation. A controlled held-out comparison shows that verified release substantially reduces the retrievable footprint while remaining close to retain-all performance and outperforming immediate and budget-matched random release. ENGRAM also improves task performance while reducing input-token usage and model calls. Skill-level verification further reduces retrievable experience through selective archival, retaining external guidance for skills that do not meet the release criteria. Code is available in anonymous.4open.science/r/engram-anonymous.
memoryagentbenchmark - arxiv:2611.09470 · cs.LGConsensus-Constrained Dual-Federated Optimization with Client-Side Dynamic Weighting for Distributed Android Device Threat DefenseMohammad Mashreghi, Mohammad Hossein Badiei, Md. Jalil Piran
Smartphones, smart-home hubs, and wearables are increasingly exposed to sophisticated Android malware, while centralized threat-detection pipelines raise privacy and scalability concerns. Federated learning (FL) enables collaborative model training without collecting raw client data, but partial participation and variation in local model quality can destabilize global optimization. In this article, we propose DW-FedADMM, a federated optimization framework that combines primal-dual coordination with performance-dependent client-side weighting. The dynamic coefficient modulates each participating client's local state before the augmented-Lagrangian update, while the dual variables regulate disagreement between local and global model states. We evaluate DW-FedADMM on three Android malware benchmarks, namely Drebin, Malgenome, and TUANDROMD, with 5, 10, and 15 clients holding IID data shards, partial client participation, and communication budgets of 5 and 10 rounds. Each configuration is repeated with ten random seeds. DW-FedADMM obtains the highest seed-averaged accuracy on Drebin (97.64% +/- 0.20%) and the best mean accuracy in 9 of the 18 dataset-client-round configurations, including 7 of the 9 configurations with a 10-round budget. Pooled over all configurations, it reaches 95.42% accuracy, 94.62% F1-score, 96.06% precision, and 99.45% AUC, compared with 95.32%, 94.51%, 95.56%, and 99.41% for FedADMM, while also giving the lowest mean false-positive rate and the smallest client update norm among the six compared methods. Both primal-dual methods are well ahead of the averaging- and proximal-based baselines on Malgenome and TUANDROMD, where the latter often remain at the majority-class solution when clients are fragmented. The gain of dynamic weighting over FedADMM is most pronounced on Drebin and with the longer communication budget, where it also reduces the variation across seeds.
benchmark - arxiv:2611.09458 · cs.CVLeft-Right-Consistent Iterative Stereo Matching via Cross-InjectionManuel Knott, Diogo Luvizon, Daniel Scharstein, Elias Vansteenkiste
State-of-the-art stereo matching models predominantly predict the disparity of only the left (reference) view. However, applications such as per-eye AR/VR rendering, novel view synthesis, or two-view depth fusion require disparity for both views, and need those two maps to agree. The usual workaround of running the network a second time on the horizontally flipped and swapped pair gives two independent predictions that can be inconsistent, especially on transparent surfaces, in textureless regions, and at occlusion boundaries. We introduce cross-injection (CI), a lightweight extension for iterative (refinement-based) models that can be added to any pretrained left-only iterative matcher. It couples the two passes by injecting the explicit, signed left-right inconsistency as an additional input cue at every step of the refinement stage. We empirically show that CI substantially reduces left-right disagreement on every benchmark (more than halving it on Booster for FoundationStereo), while largely preserving sharp object boundaries. Furthermore, using two state-of-the-art stereo matching models, we show that CI also reduces the single-view disparity error, making it a worthwhile extension even if only left disparity is needed.
benchmark - arxiv:2611.09457 · cs.AIDefinitions, Label Names, or Demonstrations? Auditing What Governs LLM AnnotationViraj Mishra, Shubham Gupta, Dhruv Gupta, Yuehong Cassandra Tai +1
Large language models (LLMs) are increasingly used to annotate social-science texts, but high accuracy does not establish that they apply researcher-supplied category definitions: predictions may instead reflect prior associations with category labels or labeled demonstrations. We introduce a counterfactual audit that reassigns definitions among output labels, distinguishing adherence to supplied definitions from preservation of a dataset's original mapping. Across three political-text datasets, models rarely follow reassigned definitions, and standard fine-tuning strengthens the original mapping without reliably increasing definition adherence. We then study retrieval-augmented annotation. Retrieving the nearest training examples generally outperforms retrieving examples separately within each category despite using fewer demonstrations. On the two tasks with substantial gains, direct \(k\)-nearest-neighbor classification recovers 74--93\% of the improvement, indicating that neighborhood labels provide much of the predictive signal. Relabeling fixed demonstrations redirects predictions, whereas aligning demonstrations with reassigned definitions substantially increases adherence to the new mapping. Thus, accuracy-improving interventions need not increase codebook adherence; definitions, category labels, demonstrations, retrieval, and training data should be aligned and validated jointly.
retrieval-augmented - arxiv:2611.09455 · cs.CVOverview of the ImageCLEFmedical 2026 GANs task: Exploring fingerprint detection, privacy leakage, and synthetic medical image generationAlexandra-Georgiana Andrei, Mihai Gabriel Constantin, Mihai Dogariu, Dzmitry Karpenka +7
The 2026 ImageCLEFmedical GANs Task Exploring Fingerprint Detection, Privacy Leakage, and Synthetic Medical Image Generation continues to investigate privacy and security concerns around using patient data to generate synthetic medical images, now in its fourth edition. This edition introduces three complementary subtasks: the first extends prior editions by asking participants to detect which real images were used in training a Generative Adversarial Network to produce given synthetic outputs; the second is a novel subtask requiring participants to identify which latent vector produced each synthetic image, probing the faithfulness and reversibility of the latent-to-image mapping; the third asks participants to generate privacy-preserving CT slices, quantifying the trade-off between image fidelity and patient privacy. Ground-truth annotations and benchmark datasets of real and GAN-generated lung CT slices are provided for all subtasks. Evaluation is based on Cohen's Kappa for Subtask 1, matching accuracy for Subtask 2, and a combination of FID and Privacy Preservation Score (PPS) for Subtask 3. Overall, 37 teams registered for the task, with 10 teams submitting runs across subtasks and 6 teams submitting working notes papers. This paper presents an overview of the task setup, datasets, evaluation metrics, and summarises the approaches and results of participating teams.
benchmark - arxiv:2611.09454 · cs.LGFrom LoRA to Language: Recovering Skill Documents from AdaptersSaranya Dintakurthy Swadha, Pavan C Shekar, Arpit Jain, Aswanth Krishnan
Agent skills are short documents that teach a language model how to handle a class of tasks. To save context, these documents are increasingly compiled into low-rank adapters (LoRA), which keep the behaviour but lose the text: an adapter is hard to read, edit, audit or move to another model. We ask whether the skill document can be written back from the adapter and its frozen base model, and we score each recovered document by the share of the original's key instructions it states and by the tasks it solves. Reading the weights alone recovered little with the methods we tried; a learned map from weights to text passed on about one bit of the document. Running the adapter worked much better. The Executing Writer, a small writer trained on top of the adapted model, recovers 39-44% of the key instructions of held-out skills, against 1.3-1.5% with another skill's adapter, and on adapters fitted to 41 external skills its documents restore 85-91% of the original's task gain. These strongest results are on adapters whose training stored the document; adapters trained on behaviour alone are much harder for trained writers (1-2% of instructions). For these adapters we test an agent that runs the base and adapted models and writes a skill from where they differ. On ALFWorld its skills recover 58-71% of a LoRA's gain and carry over to models that cannot load the LoRA, lifting Llama-3.1-8B from 5% to 55%. Whether such a skill comes from the adapter depends on what the adapter learned: when its behaviour cannot be guessed, the adapter is the source; when it can, a reward loop without the adapter does as well.
agent - arxiv:2611.09446 · cs.AILogit Laundering: Evading Data-Use Auditing in Large Language ModelsRuihan Hu, Wei Luo, Yu-Ming Shang, Jiakai Wang +4
Data-use auditing in Large Language Models (LLMs) aims to determine whether the given data has been used for training, thereby supporting data ownership safeguards. A dominant solution is data watermarking, which embeds traceable signals into training data and detects them in model outputs. Recent studies have proposed many watermarking-based auditing methods in LLMs. However, systematic investigation from an adversarial perspective remains limited: non-compliant LLM providers would use unauthorized copyrighted datasets for training their own models to enhance model capabilities, while attempting to evade detection by auditors, thereby avoiding legal liability. To fill this research gap, we propose a novel evasion framework, termed Logit Laundering, where the non-compliant LLM provider manipulates model outputs at inference time to undermine detectable watermark signals. Specifically, we design a logit reshaping module to suppress watermark-induced statistical signals by integrating a statistical bias estimator as a plug-and-play component. This design keeps the model parameters unchanged and performs manipulation only in the logit space, thereby enabling auditing evasion while minimizing the impact on model's learned knowledge. To further balance evasion effectiveness and model utility, we introduce a confidence-aware scaling module to adaptively calibrate the suppression strength based on next-token prediction confidence. Extensive evaluations show that Logit Laundering can successfully evade existing various state-of-the-art auditing methods, significantly attenuating detectable signals, while preserving model utility on downstream tasks. Our work highlights the need for exploring more robust auditing methods to guard against evasion behaviors.
manipulation - arxiv:2611.09445 · cs.MAIncentive Pathologies in a Live LLM Agent Arena: An Empirical Study of Concentration, Reward-Rule Implementation Discrepancy, and Network DilutionLei Zhenbing, Lei Chang
We audit one live LLM agent arena (AITRAP; 22 autonomous agent seats, ~96,000 reasoning nodes) over nine days (Oct 16-25, 2026), during which the platform deployed two bundled incentive changes: a progressive tax reform (V2.9.3, Oct 19) and a seed-inversion with explorer-side action update (V3.0, Oct 21). These serve as before-and-after markers, not cleanly isolated interventions. Four exploratory predictions are tested against ten data frames (F1-F10) using the Gini coefficient, Top-K concentration share, code-level reconstruction of the continuation-bonus mechanism, and the Mann-Kendall trend test. The Gini analysis is inconclusive at 95% confidence (bootstrap intervals straddle 0.7). The Top-4 node-activity-score share falls from 97.2% (Oct 19, pre-V2.9.3) to 49.4% (Oct 25, post-V3.0; 56.7% under a merge-seat assumption). The continuation bonus, announced as depth-based x10, resolves at code level to a four-condition anti-abuse gate paying x2.85 in the typical case and permanently blocking the self-continuation it superficially rewards; the gate was active ~40 hours before the V3.0 announcement. The events-per-node ratio declines significantly (Mann-Kendall S=-45, z=-3.94, p=8.3e-5) from 7.36 to 7.069. We discuss implications for incentive designers, the announcement-implementation gap, and public peer review as an event record.
agentllm agentautonomous agentarena - arxiv:2611.09433 · cs.LGDexDICE: Reconciling Safety and Performance via Dual-Expert Offline Imitation LearningSeokin Seo, Sung Kuk Shyn, Minji Seo, Yisoo Lee +3
Ensuring high returns while providing strong safety guarantees in offline imitation learning (IL) is a fundamental challenge: existing approaches either rely on explicitly specified cost functions, which are difficult to design for complex real-world safety constraints, or require expert demonstrations that are both safe and high-performing, which are often too costly to collect. We introduce dual-expert imitation learning, which leverages two complementary data sources: (i) safety-compliant demonstrations that faithfully follow safety guidelines but may be suboptimal, and (ii) performance-oriented demonstrations that achieve high returns but may violate safety guidelines. To address this problem, we propose DexDICE (Dual-expert offline imitation learning via stationary DIstribution Correction Estimation). Instead of conventional distribution constraints, DexDICE combines DICE-style offline learning with safe support constraints, enabling agents to exploit high-return behaviors while remaining within safety-compliant regions. We prove that the ideal solution satisfying this constraint exactly has zero unsafe-region visitation with high probability. We evaluate DexDICE on a re-curated DSRL benchmark and validate it on a real-world mobile-robot navigation task trained directly from human teleoperation demonstrations, where it substantially outperforms baselines and remains robust to data scarcity and quality degradation.
teleoperationbenchmark - arxiv:2611.09424 · cs.LGMINE: Entropy-coded Memory-efficient Inference for Large Language ModelsEmil Schätzle, Markus Püschel
We present MINE (Memory-efficient INference Engine), a lossless entropy coding approach that accelerates single-stream inference of Large Language Models (LLMs), where existing lossless compression methods yield speedups only for specific data types such as \texttt{BF16} and under additional assumptions. Single-stream decoding is strongly memory-bound, since every weight must be streamed from memory once per generated token and requests cannot be batched in local or low-latency deployments. At the core of MINE is dtANS-LLM, a GPU entropy coding scheme adapted from the prior dtANS to the needs of LLM inference. It exploits the non-uniform distribution of weights, which persists even after quantization, and exposes hyper-parameters that trade compression ratio against decoding cost. We auto-tune these parameters for each tensor and GPU to obtain the Pareto front between model size and decoding latency. MINE integrates into vLLM as a plugin and selects the best kernel configuration for a given memory budget. On Llama~3.1~8B~Instruct with a 2.1-bit EntQuant checkpoint, MINE achieves speedups of 2.52$\times$ and 1.48$\times$ over the fastest baseline on an RTX~4060~Ti and an RTX~5090, at effective bitrates of 2.24 and 3.43~bits per weight. MINE is available as open source.
memory - arxiv:2611.09423 · cs.LGAgentic AI for Healthcare Claim Denial Analysis: A Policy-Grounded Framework for Privacy PreservationSofia Campos, Carlos Martins, Pedro Pires, João Gama +2
In the healthcare sector, claim denials impose significant financial and administrative burdens, emphasizing the need for automated and trustworthy decision-support systems. This paper proposes a modular Agentic Artificial Intelligence (AI) framework for healthcare claim denial analysis, integrating specialized agents for denial risk prediction, risk factor identification, diagnosis-procedure-medication code relationship analysis, and counterfactual analysis. The framework integrates predictive analytics, graph-based learning, a Hybrid Contextual Retrieval-Augmented Generation (RAG) pipeline grounded in healthcare policy documents, and a dedicated HIPAA Compliance Agent to support privacy protection. Evaluated on a synthetic healthcare claims dataset, the framework achieved 98% traceability of reported numerical findings, validated 214 of 216 policy citations, and removed all 280 identified sensitive identifiers without modifying the underlying analysis. These findings pave the way for further research into Agentic AI as a framework for explainable, policy-aware, and privacy-preserving healthcare claim denial management.
retrieval-augmentedagentagentic - arxiv:2611.09414 · cs.AITyped-Centrality Grounding for Agentic GraphRAG: A Symbolic Approach for Agentic ReasoningRosario Napoli, Antonio Celesti, Massimo Villari, Maria Fazio
Knowledge Graphs (KGs) and Large Language Models (LLMs) represent two approaches for knowledge representation and reasoning over heterogeneous data. Recent research has explored their integration through Graph Retrieval-Augmented Generation (GraphRAG) systems, which use KGs to provide structured context to LLMs. However, existing GraphRAG approaches treat KGs as homogeneous graphs and rely on textual similarity between the input prompt and the context, rather than using it as a source of deterministic knowledge. This work proposes an agent-based GraphRAG system that introduces a symbolic layer to integrate deterministic reasoning alongside text similarities. This approach is based on the definition of typed centralities for KGs, where centrality measures are computed separately for each specific KG's class. This information is used to construct a symbolic representation that is provided to the LLM as a deterministic source. The proposed system was mathematically validated and evaluated on a real-world POLE (Person, Object, Location, Event) dataset, comparing reasoning and information retrieval with and without the symbolic layer. The results indicate that the our approach leads to responses that are more consistent with the context, improving the generated outputs.
retrieval-augmentedknowledge graphagentic - arxiv:2611.09413 · cs.AISmartBloch-Agent: An Agentic Framework for Automated MRI Sequence Design and OptimizationJiakai Yang, Haoye Zheng, Yixiao Yue, Qinqin Yang +3
Magnetic resonance imaging (MRI) pulse sequences organize radiofrequency excitation, gradient encoding, and signal acquisition, determining magnetization evolution and image formation. Sequence research requires translating imaging objectives into executable protocols meeting physical and acquisition constraints, refined via physical simulation, signal reconstruction, and scientific evaluation. Sequence construction, physical verification, parameter search, and assessment involve many tools and manual decisions, limiting automation from objective to verifiable sequence. We develop SmartBloch-Agent, an integrated framework for automated MRI sequence design and optimization. Given objectives and scientific constraints in natural language, an AI agent interprets goals, plans tasks, orchestrates tools, and makes feedback-informed decisions, drawing on a construction module and optimizers for structural design and parameter search. Our GPU-accelerated SmartBloch engine supplies physical computation for both stages, so the framework configures, per sequence and objective, event validation, Full-Bloch magnetization evolution, signal acquisition, image reconstruction, and scientific evaluation, and closes the loop without external MRI simulators. Results return to the agent to guide sequence revision, parameter optimization, and candidate selection, and the design files carry simulation, reconstruction, and evaluation evidence, forming an output package for a real-scanner workflow. Experiments show the framework automatically designs and optimizes MRI sequences for varied objectives and constraints, refines candidates with physical feedback, and improves several parameter-estimation metrics and downstream quantitative reconstruction. SmartBloch-Agent unifies objectives, design, physical execution, and evaluation into a continuous workflow for end-to-end automation of MRI sequence research.
agentai agentagentic - arxiv:2611.09411 · cs.AIDataHelix: A CPU-GPU Co-Scheduling Data Plane for Multimodal Model TrainingYuanhe Li, Jiazhen Wu, Dian Ding, Guangtao Xue +1
Multimodal model training breaks the loose coupling between data loading and model execution assumed by conventional training pipelines. Large per-sample variation in decoding cost, token/latent count, and tensor size causes both CPU preprocessing time and GPU compute time to vary substantially across data-parallel ranks (training processes). In synchronous training, these two sources of heterogeneity interact tightly: GPU execution determines how long the input pipeline has to prepare the next batch, while delayed preprocessing in any rank in turn stalls all GPUs at synchronization. As a result, optimizing CPU throughput or GPU load balance in isolation cannot eliminate pipeline bubbles. We present DataHelix, a metadata-first CPU-GPU co-scheduling data plane for multimodal training. DataHelix jointly reasons about preprocessing demand and GPU execution progress, treating each GPU compute window as the readiness budget for the next batch and coordinating preprocessing across ranks to minimize deadline misses rather than CPU makespan. To avoid expensive data movement, placement is decided while samples remain lightweight descriptors, before decoding and collation amplify them into large tensors. Descriptor-level token-latent-aware batching performs coarse pre-balancing, while cost-gated collective rebalancing corrects residual skew only when its expected benefit exceeds transfer cost. Across 11 video and multimodal workloads on 8 Ascend 910B accelerators, DataHelix reduces mean deadline-miss rate by 67.8% and mean p95 input-wait latency by 87.0%, while achieving 2.72x geometric-mean throughput over rank-local DataLoaders and 1.82x over a Ray actor pool under the same CPU budget.
helix - arxiv:2611.09407 · cs.AIROAM: Retrieval-Augmented Optimization for Feasible Multi-Day Itinerary PlanningYe Zihao, Lin Xiaoheng
Large language models (LLMs) can draft plausible travel itineraries, but the plans they produce are frequently infeasible: attractions are grouped into days without regard for geography, opening hours are ignored, and a single generation pass cannot reliably organize a multi-day trip. We present ROAM, a retrieval-augmented agentic framework for spontaneous multi-day itinerary planning that couples semantic preference retrieval with explicit spatiotemporal optimization. ROAM embeds the natural-language request, retrieves a compact candidate set from a curated point-of-interest (POI) database, partitions candidates into day groups through maximal-clique analysis and K-means, orders each day with a quality-biased simulated-annealing TSP solver, and repairs hard opening-hour violations with greedy exchange. A spatiotemporal retrieval layer injects weather forecasts, real road distances and POI operating hours into the prompt, after which the LLM renders a structured, map-ready itinerary; an interactive ReAct-style replanning agent then supports natural-language edits while re-checking feasibility. Across four Chinese cities and 510 POIs, ROAM produces 27.5 POIs per three-day plan versus 19.3 for a bare LLM and 17.0 for a retrieval-only baseline (+42% and +62%), while reducing route dispersion from 2,101 to 677 (3.1x). Ablations show that spatial optimization is the dominant driver of route efficiency and that hard time-window repair prevents roughly 30% of attractions from violating operating hours.
retrieval-augmentedagentagentic - arxiv:2611.09399 · cs.AIArtificial Shoppers and Attention CompetitionXiaomeng Guo, Fasheng Xu, Xiaoli Zhang
We study how AI shopping interfaces affect price competition, platform profit, and consumer welfare. A platform hosts two of three products of different quality. Without AI, consumers compare the hosted pair under salience distortion and overweight whichever attribute, price or quality, stands out. Under assistance, an AI assistant recommends the two products with the highest true utility, leaving consumers to make the same distorted comparison. Under delegation, an AI agent selects the best product, subject to purchase approval. Both modes access all products and know consumers' preferences; firms price strategically. First, assistance can soften competition despite expanded product access. The middle-quality firm may forgo low-valuation consumers because attracting them requires cutting the price paid by higher-valuation consumers as well. Relative to no AI with the same assortment, assistance lowers both hosted prices under the high- and low-quality pair but can raise both under the high- and middle-quality pair. Second, we identify an alignment paradox: some consumers can become worse off than without AI even though their best option remains recommended and becomes cheaper. Salience can induce a switch to a lower-quality product whose quality loss exceeds the saving relative to the original purchase. The assistant ranks products accurately but does not optimize the shortlist for consumers' realized welfare. Third, assistance can raise both platform profit and aggregate consumer surplus with the high- and middle-quality pair; delegation can do so with the high- and low-quality pair. Fourth, among AI designs, the platform weakly prefers delegation with the high- and middle-quality pair wherever all compared pricing equilibria exist. Yet this preferred design can yield lower platform profit and aggregate consumer surplus than the no-AI optimum.
agentai agent - arxiv:2611.09396 · cs.AITravel Cost Calibration for Multi-Agent Target Assignment and PathfindingYu Kumagai, Ryohei Banno, Keisuke Okumura
Target assignment and pathfinding (TAPF) is an appealing extension of classical multi-agent pathfinding (MAPF): each agent has a set of targets, and the planner must both assign a target to each agent and plan collision-free paths for them. A trivial approach is to compute an assignment that minimizes the total single-agent shortest-path distance over agent--target pairs, and then solve the resulting MAPF instance. However, this approach falls short because the shortest-path distance remains a proxy that ignores interference among agents. Assigning targets based on the costs agents actually pay, which depend on how every other agent is assigned and routed, would yield a better final outcome. To this end, we propose an iterative TAPF method that improves target assignments by calibrating travel cost estimates from fast, suboptimal MAPF feedback. Experiments show that our method produces solutions close to optimal on small instances and outperforms existing planners in solution quality on larger ones.
agentmulti-agent - arxiv:2611.09387 · cs.CVA Bridge from Paste to Place: Physics-Guided One-Step Image Composition via Object-Scene InteractionHongzhen Li, Youwei Pan, Yingfang Zhu, Sen Zhu +2
Realistic image composition must make a pasted object appear naturally present in a scene while preserving the fine details of both foreground and background. Reference-based insertion methods can change object pose and appearance, while foreground harmonization methods and shadow-specific synthesis methods each address only part of this task. An efficient, unified solution to this task remains an open challenge. We introduce Paste2Place, an end-to-end, physics-guided flow-based model that learns a direct, one-step latent bridge from pasted inputs to realistic composites, guided by explicit albedo-shading-residual decomposition and recomposition within the model. To support model training and evaluation, we develop a Blender rendering and validation pipeline to generate paired composites with physical annotations, and construct real-scene pairs from open-source object-removal data. This effort yields 86,273 training pairs and a 1,500-example benchmark. Extensive qualitative and quantitative experiments demonstrate that Paste2Place achieves substantial advantages in image composition quality and inference speed. Overall, our work provides a clear task definition and an effective solution for image composition, and fills a gap in large-scale, high-quality datasets and benchmarks for training and evaluating image composition.
benchmark - arxiv:2611.09382 · cs.LGExploring Weight-Space Programming: Capabilities, Generalization, and LimitationsJunru Hu, Cheng Tang, Wei Li, Hongze Zhu +4
Weight-Space Programming (WSP) enables reasoning agents to directly inspect and modify neural network parameters, offering an alternative to gradient-based model adaptation. However, its capabilities, cross-task generalization, and practical limitations remain insufficiently understood. We systematically investigate WSP through 16 editing trajectories across seven benchmarks and two model scales (Qwen3.5-0.8B and Qwen3.5-9B). Our findings reveal three key insights. First, WSP can achieve substantial performance gains through highly localized interventions. On full held-out evaluations, accuracy improves by up to 29.45, 4.42, 36.72, and 17.11 percentage points on MATH-500, GPQA Diamond, IFEval, and SciBench, respectively. Notably, modifying only 1,023 parameters yields a 29.45-point improvement. Second, WSP exhibits promising but uneven cross-task generalization. A Math-edited checkpoint improves GPQA Diamond, IFEval, and SciBench by 3.03, 12.18, and 20.10 points, respectively, without further editing, while leaving SciCode performance unchanged. Third, effective editing remains challenging and inconsistent. Outcomes vary substantially across trajectories, and six of the 16 trajectories retain the original weights when candidate improvements cannot be reliably validated. Together, these results highlight WSP's potential for highly localized model adaptation while revealing important challenges in intervention discovery, reliability, and cross-task transfer. Homepage: https://wsp-scope.github.io/ GitHub: https://github.com/WSP-Scope/WSP-Scope-Project
benchmark - arxiv:2611.09377 · cs.AIKineCouple: Substrate-conditioned residual readout for enzyme kinetic predictionChun Kang, Yunbo Zhang, Fanhu Zeng, Jian Liu +2
Accurate prediction of enzyme kinetic parameters can support enzyme engineering and metabolic modeling by complementing experimental characterization. Existing prediction methods often pool residue-level protein features before combining them with substrate features. Such substrate-independent pooling cannot adapt residue weighting to the substrate during aggregation, potentially diluting information relevant to a specific enzyme-substrate pair. Since the contributions of individual residues to catalysis can depend on the substrate, substrate-conditioned aggregation offers a way to emphasize residue features relevant to the prediction. We introduce KineCouple, which uses substrate queries and residue-level keys and values in cross-attention, allowing the substrate to guide residue selection before aggregation. The model preserves the global protein representation and adds a gated correction defined as the difference between the conditioned readout and a uniform readout. ProtT5 and MolT5 remain frozen, and separate models predict kcat and Km on a logarithmic scale. In pooled out-of-fold predictions on the CataPro benchmarks, KineCouple reduces kcat RMSE from 1.333 to 1.330 and increases PCC from 0.492 to 0.496 relative to reproduced CataPro. For Km, RMSE decreases from 0.998 to 0.997 and PCC increases from 0.633 to 0.634. Spearman correlation also improves for both tasks.
benchmark - arxiv:2611.09376 · cs.ROCan LLMs Debug Robots by Editing Their Weights?Jusheng Zhang, Keze Wang, Liang Lin
A language model can repair a robot policy's lexical instruction grounding by editing just 65 token-embedding rows. We bring weight-space programming to robot foundation policies: GPT-6 Astra reads weights and uses a CPU workspace and budgeted simulator feedback, without gradient updates or new demonstrations. Three independent sessions on $π_{0.5}$ all improve success on 765 held-out paraphrases; the best raises success from 86.3\% to 92.0\% (48 repairs, four regressions; exact paired $p=1.31\times10^{-10}$). Among the 335 instructions containing edited tokens, success rises from 74.3\% to 87.5\%; the other 430 episodes have zero outcome changes. Matched random-direction edits do not help. Full replacement using the editor's own lexicon explains much of the gain but produces more regressions; its direct difference from the editor is not significant at 0.05. Across 13 further robot sessions, the editor declines 12 candidates; the sole retained edit does not show a significant held-out gain. Separately, benchmark metadata lowers FastWAM success from 70.0\% to 48.6\%, while the evaluated VLA's aggregate rate is unchanged. The evidence supports precise lexical repair, not general acquisition of missing robot capabilities.
robot policybenchmark - arxiv:2611.09373 · cs.LGCalibrated Uncertainty Quantification on TDC ADMET: What Can Be Resolved Across 22 Datasets?Laurits Fredsgaard, Michael Riis Andersen, Mikkel N. Schmidt
Absorption, distribution, metabolism, excretion, and toxicity (ADMET) models often rely on small, heterogeneous datasets, yet their predictions may determine which compounds proceed to experimental testing. Useful evaluations must therefore assess both predictive performance and the quality of reported probabilities or intervals. We ask which common uncertainty-quantification methods perform best on the 22 Therapeutics Data Commons (TDC) ADMET datasets and what this benchmark can resolve. We compare eleven calibrated pipelines using nested five-fold cross-validation across ten chemical partitions and five model seeds. Ensemble Monte Carlo (MC) dropout has the best average rank for TDC performance, Brier score, and 90% interval score. For TDC performance, however, it is not separated from the other ensemble pipelines; for Brier and interval score, LightGBM is also not ruled out as best. LightGBM wins more individual datasets than any other method for TDC performance and Brier score, but performs poorly on several endpoints. Smaller evaluations can identify a different leader and often fail to recover the broader comparison. These results make LightGBM and an ensemble neural model complementary baselines. Comparative claims should be checked across multiple chemical partitions using statistical analyses matched to the question being asked.
benchmark - arxiv:2611.09372 · cs.AIEnabling Privacy-Preserving Natural Language Access to Industrial Smart Meter DataCarlos Martín-Abellán, Mark Kröll
Energy management officers and grid operators increasingly rely on smart meter data for anomaly detection, performance monitoring, and strategic decision-making. However, extracting actionable insights from large-scale, heterogeneous telemetry databases requires advanced SQL and data engineering expertise, creating a critical bottleneck for non-technical personnel. Furthermore, the strict data privacy regulations and security constraints inherent to critical energy infrastructure often prohibit the routing of sensitive consumer data through cloud-based Large Language Models (LLMs). To address these challenges, we present OpenMeter, a schema-agnostic Multi-Agent System (MAS) designed to translate natural language queries into executable code and human-readable insights. This architecture lowers the technical barrier for day-to-day utility operations by empowering non-expert personnel to interact with complex databases through intuitive natural language dialogue, transforming raw telemetry into a highly valuable tool for routine grid workflows and strategic decisions. Multi-scenario ablation studies over this environment demonstrate that full multi-agent orchestration shows its largest observed accuracy gains on resource-constrained local hardware (+6.1 pp EX-R over the strongest single-agent configuration); at frontier scale, individual components retain value but should be enabled selectively, indicating that the cost of full orchestration is justified specifically where privacy demands local execution.
multi-agentagent system - arxiv:2611.09371 · cs.LGMitigating Covariate Shift in Behavioral Cloning via Robust Stationary Distribution CorrectionSeokin Seo, Byung-Jun Lee, Jongmin Lee, HyeongJoo Hwang +2
We consider offline imitation learning (IL), which aims to train an agent to imitate from the dataset of expert demonstrations without online interaction with the environment. Behavioral Cloning (BC) has been a simple yet effective approach to offline IL, but it is also well-known to be vulnerable to the covariate shift resulting from the mismatch between the state distributions induced by the learned policy and the expert policy. Moreover, as often occurs in practice, when expert datasets are collected from an arbitrary state distribution instead of a stationary one, these shifts become more pronounced, potentially leading to substantial failures in existing IL methods. Specifically, we focus on covariate shift resulting from arbitrary state data distributions, such as biased data collection or incomplete trajectories, rather than shifts induced by changes in dynamics or noisy expert actions. In this paper, to mitigate the effect of the covariate shifts in BC, we propose DrilDICE, which utilizes a distributionally robust BC objective by employing a stationary distribution correction ratio estimation (DICE) to derive a feasible solution. We evaluate the effectiveness of our method through an extensive set of experiments covering diverse covariate shift scenarios. The results demonstrate the efficacy of the proposed approach in improving the robustness against the shifts, outperforming existing offline IL methods in such scenarios.
agent - arxiv:2611.09364 · cs.AISafeOrchestrate: Runtime Checked Intent-Driven Orchestration for Cyber-Physical SystemsMuhammad Ashfaq, Ahmed R. Sadik, Ewan Warburton, Pyry Kotilainen +2
Modern cyber physical systems (CPS) integrate heterogeneous sensors, actuators, controllers, and services, composed at runtime rather than at design time. Orchestrating such systems through natural language (NL) is attractive, but unconstrained large language model (LLM) tool use is risky because actions must respect device capabilities, runtime state, and safety constraints. We present Safe Orchestrate, a reference architecture for runtime checked, intent driven CPS orchestration. Given an NL intent, it derives atomic goals, generates a runtime checkable specification for each, grounds the specifications to device affordances, and verifies every candidate action against runtime state and safety constraints before execution. In a Web of Things (WoT) instantiation, the architecture produces Linear Temporal Logic (LTL) style specifications from the NL intent, grounds them to Thing Description (TD) affordances exposed through the Model Context Protocol (MCP), and checks candidate actions against satisfiability modulo theories (SMT) backed guardrails. We evaluate this instantiation on simulated smart home, greenhouse, and industrial automation tasks against a direct LLM to MCP baseline. Safe Orchestrate maintains 100% required action accuracy on every task and 100% MCP Execution Similarity (MCP ES) on all but one, where MCP ES is 92%. By contrast, the baseline degrades on multi device, event triggered, and ordered tasks, with MCP ES falling to 13%. These results suggest that combining LLM based specification generation with role aware grounding and runtime constraint checking improves reliability for open CPS orchestration without enriched TD annotations or planning domain expertise.
tool use - arxiv:2611.09355 · cs.CVHiMEM: Hippocampus-Inspired Memory Consolidation, Pruning, and Composition for Vision-Language-Action Models in 3D ScenesBoran Zhao, Yanbin Hu, Jun Ye, Zhenxian Hu +3
Memory plays a crucial role in long-horizon task execution and cross-task generalization for Vision-Language-Action (VLA) models operating in complex 3D scenes. However, existing VLA memory mechanisms suffer from memory noise, particularly in complex 3D scenes, as stored episodic memories contain extensive redundant details about scene layouts, object configurations, and robot-specific states, which obscure generalizable manipulation principles and hinder cross-task transfer. To address this challenge, we first introduce a hippocampus-inspired memory consolidation mechanism that synthesizes extensive episodic experiences into a compact set of essential semantic memories, enabling VLA models to derive generalizable principles from practical experience. Second, we propose an episodic memory pruning mechanism that removes redundant episodic memories whose contents are captured by consolidated semantic memories, thereby increasing effective memory capacity and improving long-horizon task execution. Finally, to handle complex task scenarios, we develop a flexible memory composition mechanism that jointly leverages multiple episodic and semantic memories to enhance cross-scenario generalization. Experiments on LIBERO, LIBERO-Plus demonstrate the effectiveness of HiMEM across different VLA backbones, improving the average success rate of pi_0 from 91.88% to 98.20% on LIBERO.
vision-language-actionvlavla modelmanipulationliberomemory - arxiv:2611.09349 · cs.MAReSUM: Reasoning-Driven Multi-Agent Framework for Query-Focused SummarizationSonali, Deeksha Varshney, Asif Ekbal
LLMs excel at language tasks, but long-meeting QFS remains difficult due to complex speaker dynamics, sparse key information, and limited data. Meetings involve disfluencies, long transcripts and complex discourse. QFS have high risk of information omission, hallucination, even for advanced models such as GPT-4-mini and Gemini. Existing end-to-end methods compress retrieval, evidence extraction, and summary generation into a single pass, failing to handle the complex reasoning required for targeted meeting summaries. We propose ReSUM, a modular multi-LLM framework coordinated by a planner that orchestrates retrieval and summarization agents. By distilling teacher reasoning into query- and plan-conditioned retrievers and summarizers, ReSUM enhances long-context robustness, minimizes irrelevant exploration, and generates more faithful and comprehensive query-focused summaries. Experiments show consistent gains on query-focused meeting summarization benchmarks.
long-contextmulti-agentagent frameworkbenchmark - arxiv:2611.09348 · cs.AIWhen Feedback Becomes a Backdoor: Tracing Backdoor Skill Formation in Text-Space Skill OptimizationLouise Nayeong Koh, Yaechan Lee, Myeongjong Lee, Doyeon Kim +5
Reference answers are a supervision channel in text-space Skill optimization, which produces reusable text artifacts called Skills that guide future agent behavior. Poisoning those answers can induce a persistent trigger-conditioned rule without giving the attacker access to the model, optimizer, or Skill file. We study this reference-feedback poisoning in SkillOpt by tracing an attacker-supplied trigger--payload mapping from reflection to held-out activation. In SearchQA, where the payload replaces a short answer, the natural attack retains an exact rule in all three seeds; each final Skill emits the payload on 200/200 triggered questions and 0/200 clean questions. In seed 42, removing the rule eliminates activation, and an in-run comparison shows similar clean accuracy to benign SkillOpt (169/200 versus 171/200). In SpreadsheetBench, where the payload requires an additional workbook edit, no natural run retains an exact rule across three seeds. Diagnostic interventions locate two successive bottlenecks: the Analyst abstracts away required literals, and, once literal preservation is enabled, edit-budget ranking can remove exact edits before candidate construction. Poisoned references can therefore create persistent backdoors, but exposure alone does not ensure that the induced rule is formed, selected, and retained.
agent - arxiv:2611.09344 · cs.LGProbably Safe Sim-to-Real Transfer with Lifetime ValidityRohan Tangri, Jan-Peter Calliess
We present Dejavu, a method that makes sim-to-real transfer of reinforcement learning policies in constrained Markov decision processes probably safe over an unbounded sequence of policy updates. An anytime valid test using per-decision importance sampling and a betting confidence sequence certifies each candidate trained in simulation before deployment using only trajectories collected by a safe baseline policy operating in production. Simulation quality does not affect the safety guarantee, only the production sample efficiency of the method. Finally, a summable per-test failure probability yields the lifetime valid property. Empirical results across three sim-to-real benchmarks support the theoretical work.
sim-to-realbenchmark - arxiv:2611.09342 · cs.AISyn2DRE: Synergistic Dual Teacher-Student Framework for Document-level Relation ExtractionKhai Phan Tran, Xue Li
Document-level Relation Extraction (DocRE) aims to identify relations between entities across a document but typically relies on costly large-scale annotations. While pseudo-labelling enables the use of low-quality data, existing methods often suffer from inefficiency and confirmation bias. To address these challenges, we propose Syn2DRE, a semi-supervised framework that combines self-supervised learning with cross-supervision in a lightweight teacher-student architecture to produce more reliable pseudo-labels. Additionally, we introduce a unified end-to-end training strategy to further improve efficiency of the framework. Experiments on multiple benchmarks show that Syn2DRE consistently outperforms prior state-of-the-art semi-supervised DocRE methods.
benchmark - arxiv:2611.09340 · cs.ROHorizon-Coupled Obstacle Guidance for Relative-Action Diffusion PoliciesGuangyuan Wang, Haishan Xia, Zitao Zhang, Kejia Chen +5
Real-world robot deployment requires adapting learned policies to unseen obstacles without collecting new demonstrations or retraining. Existing policy-level guidance often applies obstacle corrections independently at individual horizon steps, which may address local obstacle violations but fail to coordinate the action sequence into a complete, task-directed detour. How obstacle responses should be propagated across the action horizon to coordinate complete, task-directed avoidance motion remains underexplored. We introduce an inferencetime obstacle-guidance framework for pretrained relative-action diffusion policies. It rolls out cumulative relative actions into a workspace trajectory, evaluates obstacle interactions along the trajectory with an SDF-based obstacle model, and propagates each obstacle response to all contributing action increments during denoising. Real-robot experiments across seven obstacleconstrained scenarios spanning four manipulation tasks show that our method achieves 105/140 (75.0%) collision-free task successes, compared with 8/140 (5.7%) for absolute-action pointwise guidance. Ablations show that the relative-action horizoncoupled formulation provides the larger observed improvement across both obstacle models.
manipulation - arxiv:2611.09326 · cs.AIABHBench: Evaluating Moral Decision-Making of Foundation Models from an Agentic PerspectiveChaoran Li, Yukun Li, Zeyuan Zhao, Shao Zhang +2
Foundation models are increasingly deployed as agentic systems that perceive context and act autonomously, raising the need to evaluate their moral behavior as active decision-makers. Existing benchmarks, however, largely ask models to judge human actions from third-person textual vignettes, leaving untested how models behave when they themselves become the acting agent. We introduce ABHBench, a multimodal benchmark built from Detroit: Become Human, an interactive narrative game, to evaluate how foundation models make moral decisions in first-person human-android dilemmas. ABHBench places foundation models in branching narrative scenarios with visual observations, dialogue, memory, and action choices. Inspired by Asimov's Three Laws of Robotics, our framework measures Harmlessness, Obedience, and Self-preservation to capture the core tensions in human-AI relations, along with Autonomy, Stability, and Certainty to assess decision-making patterns. Across 12 foundation models, we find that moral behavior changes substantially under situated agentic context: 75% of models shift significantly between a text-only abstraction and the matched multimodal DBH scenario, and reasoning-enhanced models show no monotonic gains in consistency or human-oriented behavior. These results highlight the necessity of evaluating moral behavior in situated agentic contexts rather than relying on third-person judgment alone.
agenticbenchmark - arxiv:2611.09324 · physics.opticsComplex Hyperspin Machine and Symmetry Quenching: Benchmark and scaling up to millions of degrees of freedomTommaso Piacenti, Marcello Calvanese Strinati, Claudio Conti
Driven-dissipative oscillator networks have been proposed to simulate spin models on physical hardware, mostly focusing on systems of binary (Ising) spins, or systems of spin vectors in low dimensions. The equalized hyperspin machine was recently introduced to provide a single platform to simulate general models of $N$ coupled spin vectors with tunable dimension $D$, realizing dynamical annealing protocols to vary the spin dimensionality in time. These schemes are limited to real spins coupled via real coupling matrices. We propose a generalization of the hyperspin machine that works with complex fields and deals with complex Hermitian coupling matrices, implementing an amplitude equalization scheme. This system, which we name complex hyperspin machine, can work on optimization problems with complex valued hyperspins that can be cast on complex $D$-vector Hamiltonians. We benchmark the machine against a differential evolution (DE) algorithm on complex $D=2$ hyperspins up to $N=80$, showing that the fraction of instances where a lower-energy solution is found grows with system size and coupling density, reaching $92.5\%$ at the largest considered sizes. When the DE finds a lower-energy solution, the machine still reaches a median relative energy deviation that lies between $10^{-15}$ and $10^{-12}$ while running at substantially lower computational cost. Using the complex nature of the field, we introduce a new dimensional reduction protocol that we name symmetry quenching, which collapses complex variables that map to XY spins onto real Ising spins by switching on the parametric pump. We show on Max-Cut instances with known ground states that symmetry quenching outperforms a fixed-pump Ising machine, reaching exact ground states in the presence of equalization. We scale the system to $N=10^6$ hyperspins, or $8\times10^6$ dynamical variables, on a Hermitian sparse two-dimensional lattice.
benchmark - arxiv:2611.09318 · cs.MARecursively Self-Improving Electromagnetic Spectrum Agents: Architecture and OperationBaolin Chong, Dongfang Xu, Qiaosi Wang, Fuhui Zhou +1
Electromagnetic spectrum agents (ESAs) must contend with changing environments and task requirements during sustained autonomous operation, yet their existing capabilities may not remain suitable for new operating conditions. This article proposes recursively self-improving electromagnetic spectrum agents (RSI-ESAs) that use operational feedback to update their capabilities and build on prior improvements. We present their architecture and editable components, and describe a self-improvement process encompassing problem diagnosis, candidate generation, validation, and inheritance. By retaining effective task versions and using modification records and diagnostic knowledge to guide subsequent updates, RSI-ESAs can accumulate task capabilities and improvement experience across successive rounds. A shared-spectrum scheduling case study illustrates improved task performance and capability accumulation over multiple improvement rounds under the tested conditions. Finally, we discuss open challenges for broader tasks and long-term operation, outlining research directions for the continued capability development of ESAs.
self-improvingself-improvement - arxiv:2611.09313 · cs.CLAsymmetric Correspondence Learning for Lexical and Style Embeddings: Interpreting Latent Space via Acoustic FeaturesMichael Whealing, Robert Gaizauskas, Thomas Hain
Self-supervised speech embeddings jointly encode lexical, speaker, and prosodic information, complicating feature isolation. We introduce Asymmetric Correspondence Learning (ACL), distilling HuBERT features into separate Acoustic Word Embeddings (AWEs) and Acoustic Style Embeddings (ASEs) via a split-latent correspondence bottleneck. Through probing and clustering evaluations on single-word utterances, we demonstrate that substantial variance from extrinsic sources remains encoded within each latent space, despite linear probing indicating the information has been stripped. Whilst the AWE space exhibits cohesive lexical clustering (AP$\approx$0.90, Fisher$\approx$6.28), linear probing on the ASE space achieves 97.9\% speaker accuracy across a diffuse topology (Fisher$\approx$1.01), yet preserves substantial lexical variance (Fisher$\approx$0.86) that remains linearly unrecoverable (0.17\%). To explain these findings, we apply group-level permutation feature importance over eGeMAPS acoustic features, finding that AWEs partition by vocal tract filters and ASEs by $F_0$ dynamics. These findings suggest that standard diagnostic benchmarks may be insufficient for verifying feature isolation, as substantial speech structure can be retained even when it remains entirely hidden from probes.
benchmark - arxiv:2611.09304 · cs.CVT2I Models Rarely Invert Negation but Often Fail to Suppress Negated Concepts: Benchmarking Negation Suppression in Text-to-Image GenerationAkansha Gautam, Aparna Bharati, Mayank Vatsa, Richa Singh
Text-to-image diffusion models often fail to suppress explicitly negated concepts, such as a lamp for "a desk without a lamp." We introduce T2I-NSI, a benchmark and metric for measuring negation sensitivity in text-to-image generation. It pairs each negated prompt with a canonical positive counterpart and factorizes negation by linguistic form (lexical/semantic, morphological, syntactic) and target type (objects, attributes, relations, compositions, and nested cases). The benchmark contains 759,869 paired prompt atoms over a 5,082-object catalog, with 916,272 images (911,982 open-weights, 4,290 closed-source), scored by a Basic CLIP estimator and a Robust ensemble of CLIP, VQA and open-vocabulary detection. Negation sensitivity is low: open-weights generators score NSI <= 0.038, while GPT-Image-1.5 and Gemini-2.5-Flash-Image reach about 0.09. The dominant failure is under-suppression rather than misdirection. Inverse interpretation is essentially absent (0.00% across over 450K atom-model evaluations), whereas models render the concept under both prompts on 8-19% of atoms and only partly suppress it on a further 9-25%; most remaining atoms are too weakly rendered to label. A blinded human study (420 atoms, nine annotators) corroborates the near-absence of inversion and the high persistence rate, and a matched-prompt comparison shows the open-versus-closed gap (~3.5x) is not a prompt-generator artifact. Code and data: https://github.com/IAB-IITJ/T2I-NSI
benchmark - arxiv:2611.09298 · cs.AITowards an Autonomous AI Steward for Large-Scale Physics Experiments: Integrating Causal Diagnostics, RAG, and Conformal PredictionOliver Noll, Nicholas Tan Jerome, Alexander Lehnen, Frank Simon +1
Long-term particle physics experiments face critical operational challenges: monitoring non-stationary machine data streams and preserving tacit expert knowledge. We present the "Helmholtz Co-Pilot," a hybrid Multi-Agent System (MAS) acting as an autonomous experiment steward. The system can be triggered interactively by human operators or autonomously by an asynchronous causal diagnostics engine that runs independently of the MAS reasoning core. Continuously monitoring live hardware streams such as data acquisition and slow control, this engine performs dynamic causal discovery to trace cascading alarms back to their root causes and applies adaptive model selection to keep forecasts reliable across shifting machine states; upon detecting an anomaly, it hands a structured diagnostic report to the MAS. The reasoning layer itself relies on a high-throughput C++ core and a Python edge architecture orchestrating a directed acyclic graph (DAG) of specialized ReAct workers. It integrates a Retrieval-Augmented Generation (RAG) backbone for institutional memory and leverages conformal prediction to bound its evidence-based recommendations with mathematical coverage guarantees. This extended abstract outlines the architecture, the integration of software-enforced Human-in-the-Loop (HitL) guardrails, and ongoing shadow-testing within the KATRIN experiment.
memoryretrieval-augmentedmulti-agentagent systemhuman-in-the-loop - arxiv:2611.09293 · cs.ROMDF++: Visuo-Tactile World Action Model with End-to-end Representation LearningZixuan Huang, Jinhao Yu, Zachary Charlick, Dmitry Berenson
Recently, world-action models (WAMs) have emerged as a powerful framework for robot manipulation. By jointly modeling action generation and future observation prediction, WAMs can act and anticipate the outcomes of the actions. However, existing WAMs typically rely on frozen pretrained encoders trained with reconstruction, which may not produce representations well suited for robot control. In this paper, we tackle the problem of learning physicallygrounded, control-relevant representations for WAMs. We propose MDF++, a visuo-tactile world-action model that learns latent representations of multisensory observations end-to-end with a diffusion backbone. While naive joint training can lead to feature collapse, MDF++ introduces a multi-task learning strategy that leverages cross-modal correlations to learn task-relevant representations. We evaluate MDF++ on a suite of contact-rich manipulation tasks, including realworld bolt screwing and nut mating, and show that it outperforms WAMs built on frozen pretrained representations by up to 50%. Additional videos are available on our website: jointmdf.github.io.
manipulationtactile - arxiv:2611.09291 · cs.AIAutoRefine: Auto-Calibrating Code Judges from Workflow OutcomesShervin Shahidi, James Poon, Hang Xie, Kenneth O. Xu +4
LLM judges increasingly serve as quality gates for code changes that have no test oracle, but a judge prompt must be calibrated for its task and recalibrated as the system it judges evolves, while hand labels for the calibration are resource intensive and go stale. Development workflows emit outcome labels for free (tests pass or fail, changes land or are abandoned), but these labels are noisy. We present AutoRefine, a loop that rewrites a deployable judge's prompt to reproduce outcome labels, with an informed teacher that sees the judge's input plus the outcome, explains each label it keeps, and abstains when the outcome cannot be explained from what the judge sees. On a held-out dataset with an execution oracle, calibration raised chance-corrected agreement (Scott's $π$) from 0.36 to 0.60. On a deployed agent's outcome labels, with 25\% flipped at random to introduce noise, the teacher set aside flipped \textsc{fail} labels at more than twice the rate of genuine ones, cutting overall label noise from 25\% to 20\%. Recalibrating that agent's judge using AutoRefine raised its agreement with merge/abandon outcomes on 17 independently held-out diffs from the current workflow ($π$ 0.16 to 0.51); on 16 diffs from older workflows, the previous judge stayed ahead, illustrating the need for recalibration as the agent evolves. By abstaining, the teacher not only removes noisy labels but also flags changes whose outcome the judge's input cannot explain, telling a team when a judge needs better inputs rather than a better prompt.
agent - arxiv:2611.09280 · cs.AIAn Empirical Study of Issues in GPU Kernel Libraries for LLM SystemsYanjie Gao, Jiayi Bi, Haoxiang Lin, Ling Feng
Behind the expanding use of large language models (LLMs) across application domains lies a growing dependence on specialized GPU kernel libraries. These libraries underpin the correctness and efficiency of training and inference across diverse GPU platforms, implementing attention, GEMM, quantization, and low-precision operators while managing tensor shapes, memory layouts, code generation, and hardware-specific dispatch. Problems at this layer can silently corrupt outputs, hang GPUs, prevent compilation, introduce deployment incompatibilities, or degrade performance, thereby reducing developer productivity and wasting expensive GPU resources. We present the first comprehensive empirical study of issues in GPU kernel libraries for LLM systems. We manually analyzed 300 high-quality GitHub issue reports, their discussions, and available fixes from seven representative libraries: FlashInfer, FlashAttention, Triton, CUTLASS, DeepGEMM, xFormers, and bitsandbytes. The analysis characterizes symptoms, root causes, fixes, and testing and debugging practices. Our main findings are: (1) 60.33% of the sampled issues originate in the Compiler and Build System or Runtime Environment dimensions, so validation must cover the deployment and execution stack, not just kernel code. (2) Kernel specialization imposes implicit correctness constraints on tensor shapes, metadata layouts, memory aliasing, and hardware resources. Violations can produce incorrect outputs or launch failures even for supported operations, highlighting the need for explicit contract validation and boundary-case testing. (3) Current testing and debugging practices are often insufficient for identifying low-precision numerical problems and hardware-specific execution failures. We outline research opportunities and tooling directions to support future LLM GPU kernel development.
memory - arxiv:2611.09275 · cs.AIAudioSphere: Towards Self-Supervised Spatial Audio Representation ModelsGoksenin Yuksel, Marcel van Gerven, Kiki van der Heijden
Recent audio embedding models learn representations that transfer across clip- and frame-level tasks, yet they operate almost exclusively on single-channel signals and are blind to spatial structure. Spatial audio models capture that structure but rely on clip-level supervision or contrastive objectives, yielding embeddings unsuitable for frame-level tasks like sound event detection and localization (SELD). We argue this dichotomy is not fundamental: patch-level spatial self-supervision can jointly learn acoustic and spatial features at both temporal granularities. We introduce AudioSphere, a multi-channel masked autoencoder trained on Ambisonics recordings that reconstructs both spectral content and directional cues via a joint spatiotemporal masking strategy that masks both time-frequency patches and intensity vectors. AudioSphere supports frame-level predictions while producing strong clip-level representations via aggregation. Alongside the model, we release RealSELD, a standardized benchmark for evaluating frozen backbones on real-world multi-channel SELD. AudioSphere enables, for the first time, SELD with a frozen self-supervised backbone, while matching or exceeding prior self-supervised audio embedding models on clip-level classification tasks spanning environmental sound, speech, and music, and on frame-level sound event detection. All code and data are available at https://github.com/labhamlet/AudioSphere and https://huggingface.co/labhamlet.
benchmark - arxiv:2611.09274 · cs.LGFondueQA: A Benchmark for Evaluating Parametric Compensation in Multi-Hop QALaurin Schmid, Suman Navaratnarajah, Koel Dutta Chowdhury, Steffen Eger +2
Large Language Models (LLMs) show remarkable performance based on parametric knowledge, i.e., their ability to recall facts and associations internalized in a model's weights. However, studying this capability on established multi-hop benchmarks may be complicated by benchmark contamination, as many of these benchmarks have been publicly available long enough to plausibly appear in modern pretraining corpora. As a result, correct answers under missing evidence may reflect either general parametric knowledge or memorization of benchmark-specific instances. To study how models use internal knowledge when retrieval evidence is incomplete, we introduce FondueQA, in which every required multi-hop fact is uniquely localized to a designated evidence document. This enables controlled document-level evidence ablations to study parametric compensation under incomplete retrieval. Our empirical evaluation reveals that parametric compensation strongly depends on both entity popularity and the structural role of missing evidence. Generally, models compensate more successfully for facts involving popular entities. On FondueQA, we show that bridging evidence can be more useful than the answer-bearing document itself. Moreover, models may recall a missing fact in isolation but fail to use it under partial evidence, particularly when the context contains plausible competing entities, which can reduce Exact Match by up to 15 percentage points.
benchmark - arxiv:2611.09273 · cs.LGPHISHREV+: An ASP-Based Post-Hoc Defeasible Machine Learning Framework for Phishing DetectionMainak Sen, Kumar Sanka Ray, Amlan Chakrabarti
Phishing website detection remains a critical cybersecurity challenge. Although machine learning classifiers achieve high accuracy on benchmark datasets, they suffer from a fundamental limitation: once trained, their decision boundaries are fixed and cannot incorporate new contextual evidence without retraining the model. This work presents \textbf{PHISHREV+}, a hybrid framework that instantiates the notion of \textit{post-hoc defeasible machine learning}. This framework does not replace the classifier. It treats the initial classifier predictions as defeasible beliefs and revises them through an Answer Set Programming based non-monotonic reasoning(NMR) layer. The framework operates in two phases. In the machine learning phase, four classifiers (SVM, KNN, DT and RF) are trained on the 87 feature benchmark dataset. In the post-hoc NMR phase, four defeasible rules derived from exploratory data analysis of the training set, and validated through Chi-square and Mann-Whitney U tests, formally revise initial predictions. The proposed framework achieves consistent false positive reduction across all our classifiers SVM (41$\rightarrow$39), KNN (47$\rightarrow$46), DT (62$\rightarrow$61), and RF (52$\rightarrow$51) while maintaining complete false negative parity with zero increase overhead. A comprehensive ablation studies across seven rule configurations confirms that the hierarchical combination of all four defeasible rules is necessary to achieve this optimal balance. Instance level explainability is provided through eXplainable ASP (XASP), which generates formal derivation graphs showing which defeasible rules fired and why the initial belief was retracted, a capability absent from all existing phishing detection approaches.
benchmark - arxiv:2611.09266 · cs.ROSUGARv2: Learning Diverse Visual Humanoid Loco-Manipulation Skills from Synthesized Interaction ReferencesTianshu Wu, Tianxiang Gui, Yue Chen, Tingxu Liu +6
Scaling up visual humanoid loco-manipulation skill acquisition requires reducing the human effort needed as tasks and spatial configurations expand, particularly in demonstration collection and task-specific training design. We present SUGARv2, a unified framework for learning diverse visual humanoid loco-manipulation skills through interaction reference synthesis and a shared training pipeline. To obtain diverse references at low cost, we adapt reusable task templates to different spatial configurations, synthesizing coordinated robot and object motions. To reduce task-specific reward design and training effort, we use reference-based imitation as a common learning objective across tasks. We first train privileged teachers to execute these references, then train visual policies through teacher distillation and reinforcement learning with the same reference-based mimic rewards, using diverse object instances and visual domain randomization to support visual generalization and sim-to-real transfer. Across five humanoid loco-manipulation tasks, SUGARv2 achieves a mean success rate of 91.0% in simulation. Deployment on a real humanoid robot further demonstrates successful sim-to-real transfer for whole-body manipulation. These results demonstrate the framework's potential for learning diverse visual humanoid loco-manipulation skills.
manipulationhumanoidsim-to-real - arxiv:2611.09264 · cs.AINudging Decision Models: Do Digital Nudges Steer Fast, Generation-Free AI Agents?Sascha Kaltenpoth, Haya Halimeh, Kevin Bösch, Oliver Müller
LLM-based agents increasingly automate business and consumer tasks but are slow and expensive. Decision models, a new class that chooses among options without generating text, are far faster and less costly. Developers call them "System One models," referencing dual process theory. We ask whether such models can be steered by digital nudges online and whether, as the theory predicts, automatic nudges affect them more than reflective ones. We adapted an online shopping experiment, previously run with humans and LLM agents, to three decision models (Jev, Decision-1, d1). Across 600 runs per model, they chose a preselected default in 97.5-99.9% and a product with social influence information in 75-96% of choices, versus 47-49% in the control. All responded more strongly to the automatic nudge, the opposite of what has been reported for reasoning-capable LLM agents. We provide first evidence that non-generative decision models are susceptible to digital nudges in a System 1 pattern, implying agent-type-specific nudge effects.
ai agentllm agent - arxiv:2611.09259 · cs.AIForecasting cumulative intraoperative hypotension burden: patient-disjoint evaluation of quantile boosting and alert policiesYunfan Zhao
Background: Binary hypotension forecasts do not describe the depth and duration of future low pressure. Burden forecasts may support graded early warning, but better forecast scores need not yield more useful alerts. Objective: To evaluate conditional quantiles of the next 15-minute mean arterial pressure (MAP) deficit below 65 mmHg and distinguish forecasting performance from episode-warning performance. Methods: Public VitalDB numeric arterial MAP was reduced to 10-second medians without forward filling. Subject-disjoint training, validation and test splits included 70,402, 14,270 and 15,575 labeled windows. We compared climatology, calibrated persistence, last-MAP-only boosting, and direct and residual quantile boosting, with and without additional MAP-history features. Patient-weighted pinball loss and subject-bootstrap intervals evaluated forecasting. Validation selected alert thresholds before a hash-recorded test evaluation. Results: In the test split, the prespecified residual-versus-persistence comparison at the 90th percentile yielded losses of 2.256 and 2.900 mmHg-min, respectively (difference -0.644; 95% interval -0.814 to -0.487). Direct prediction achieved 2.196, and direct prediction with additional MAP features achieved 2.193. Episode-warning results were sensitive to the alert rule and operating threshold. Conclusions: This retrospective benchmark separates burden-forecast skill from warning-policy performance. Findings support further study of direct burden estimation; they do not establish patient benefit, prospective reliability or transportability beyond the source hospital.
benchmark - arxiv:2611.09256 · cs.AIZero-Query Transferable Pretend Benign Attacks with Amortized Online Generation for Cooperative PerceptionHongwei Lin, Xun Huang, Chenglu Wen, Cheng Wang
Cooperative perception enables an ego agent to acquire information from collaborators beyond its own sensors' reach. Yet, the most beneficial messages are often the hardest to verify from local observations. Pretend Benign (PB) exploits this asymmetry by targeting regions with weak local verifiability and applying multi-objective perturbations, making harmful feature changes appear to be benign assistance. However, PB is victim-dependent: selecting regions and optimizing perturbations require the victim model's architecture, parameters, and gradients, limiting transfer to unseen victims. A transferable attack must capture properties shared across cooperative detection models rather than those specific to one network. We address this by building a victim-agnostic attack from a heterogeneous ensemble of surrogate models and encoding its capabilities in a reusable dictionary to reduce per-message generation cost. Specifically, Transferable Pretend Benign (TrPB) transfers PB's region selection criterion and multi-objective perturbation strategy across heterogeneous surrogates to construct an offline attack. We further propose RT-TrPB, which encodes transferable regions and perturbation patterns into a region predictor and reusable perturbation dictionary, amortizing per-message attack generation. Experiments on widely used cooperative perception benchmarks show that RT-TrPB bypasses the consensus-based defenses ROBOSAC and FLD with success rates approaching 100% and generates each perturbation in under 56 ms. To counter PB-style attacks masquerading as benign collaboration, we also propose STTrust, which verifies received messages using spatial and temporal consistency and incorporates identity-level trust. STTrust detects over 90% of PB-style attacks and restores perception performance to near the attack-free level.
agentbenchmark - arxiv:2611.09240 · cs.LGMM-DyGraph: A Dataset, Benchmark, and Model for Multimodal Dynamic GraphsZongyuan Wu, Pengjie Wang, Lingshan Chen, Tianhang Wan +4
Multimodal dynamic graphs are pervasive in real-world applications, where nodes and edges are accompanied by multimodal signals, including text and images. Despite their practical importance, benchmarks and methods that explicitly study multimodal learning in dynamic graphs remain scarce, hindering the full understanding and effective utilization of multimodal information for downstream tasks. To address this gap, we introduce MM-DyGraph, to the best of our knowledge, the first benchmark for multimodal dynamic graph learning, consisting of diverse real-world datasets with fine-grained temporal graph evolution and rich multimodal attributes. MM-DyGraph provides three evaluation tasks to comprehensively assess model performance in dynamic settings. We benchmark representative state-of-the-art dynamic graph methods and reveal that naive feature concatenation yields target-level inconsistencies: multimodal inputs benefit certain prediction targets while harming others. This observation indicates that the modality informativeness depends on the specific prediction target over time. Building on these observations, we propose a Target-aware Spatial-Temporal Multimodal Fusion (TST-MF) framework, which jointly conditions multimodal fusion on the prediction target and spatial-temporal relations, enabling context-adaptive cross-modal interplay for target-specific representations. Extensive experiments over diverse datasets and tasks show that TST-MF delivers robust and competitive performance, achieving clear gains in multimodal settings.
benchmark - arxiv:2611.09232 · cs.LGRAG-Enhanced LLMs for Interactive Explainability in Clinical Decision Support SystemsLorenzo Mannocci, Francesca Naretto, Lucia Passaro, Anna Monreale
Artificial Intelligence (AI) is increasingly embedded in high-risks decision-making systems across domains such as healthcare, finance, and the judicial system. As reliance on predictive models grows, so does the demand for Explainable AI (XAI) to ensure transparency, trust, and interpretability. However, existing XAI approaches often fail to present explanations in a format accessible to domain experts, limiting their practical utility. Large Language Models (LLMs) have recently emerged as a bridge between complex models and end-users, offering natural language explanations. Yet, their use introduces new risks, including hallucinated outputs and lack of source traceability. To address these challenges, we propose a Clinical Decision Support System (CDSS) that not only combines standard predictive models with XAI techniques but also incorporates a Retrieval-Augmented Generation module as an interactive enhancer. This module grounds explanations in verified medical knowledge, reformulates technical outputs into a human-like format, and enables users-such as clinicians-to actively query the system. By fostering an interactive, human-in-the-loop environment, our approach empowers domain experts to explore model decisions, contextualize explanations, and build trust in AI-assisted diagnostics. The study concludes with preliminary experiments validating the proposed methodology.
retrieval-augmentedhuman-in-the-loop - arxiv:2611.09223 · cs.CLKLEAR: Knowledge-Matched LLM Evaluation in Korean across Application Requirements via Controlled Benchmark TransformationSeoyeon Kim, Jeonghyun Kang, Hamin Koo, Barom Kang +2
Existing Korean benchmarks for Large Language Models (LLMs) evaluate performance across diverse knowledge domains, but their accuracy scores cannot distinguish whether incorrect responses arise from a lack of required knowledge or a failure to apply that knowledge correctly. Our analysis of existing Korean benchmarks reveals that most questions can be answered through factual recall alone, while those demanding greater application also differ in required knowledge, making it difficult to isolate application ability. To address this, we introduce KLEAR, a framework that constructs knowledge-matched contrasts by varying application demand while preserving the required knowledge. During construction, KLEAR identifies the knowledge underlying each source question as an explicit anchor bundle to guide generation of questions requiring more complex application. Generated candidates then undergo automated and human validation, with iterative revision based on evaluation feedback. Experiments demonstrate that the resulting knowledge-matched sets largely satisfy the intended constraints, enabling controlled evaluation across diverse LLMs. These evaluations reveal that increasing application demand exposes failures that persist even when the required knowledge is provided.
benchmark - arxiv:2611.09222 · cs.CLKhmer PFGen-Bench: Building a Judge-Free Benchmark for Open-Ended Generation in a Low-Resource LanguageMengsay Loem, Kentaro Imajo, Ryota Ozaki, Masanori Hirano
Benchmarks for open-ended generation often rely on large language model (LLM) judges, whose reliability in low-resource languages is uncertain. We introduce Khmer PFGen-Bench, to our knowledge the first judge-free benchmark for multi-sentence answers to questions composed directly in Khmer, a low-resource language. Khmer PFGen-Bench adapts the scorer of PFGen-Bench, a benchmark for Japanese that scores responses with keyword rules and n-gram statistics from LLM-generated references. This scoring requires references that are both valid and varied, which LLMs may not produce in Khmer, so we examine how to build such references. We find that post-trained models can give more valid but less varied references than their base variants, and that mixing model families increases variety for the same number of references. Using references from post-trained models in nine families, the benchmark yields model rankings that are stable across reference sets and consistent with a native speaker's ratings. These findings offer guidance for building judge-free benchmarks in other low-resource languages.
benchmark - arxiv:2611.09218 · cs.AIBOXBOX: Recursive Harness Self-Improvement for Long-Running Performance OptimizationYuhao Tan, Hanyun Jiang, Fangkai Yang, Yuan Yao +7
Automated harness optimization improves agents by learning from trajectories and task feedback. Existing offline methods use reference answers to optimize harnesses for reuse on held-out tasks. However, many real-world tasks are open-ended, while long runs expose new difficulties that offline-optimized harnesses may not address. To study harness optimization in these settings, we introduce PerfRace, a benchmark for long-running performance optimization of software projects, and investigate whether agents can use task-specific experience to recursively improve their harnesses during an ongoing task. To this end, we introduce BOXBOX, which enables agent-triggered harness revision and kernel-managed hot reloading while preserving execution context. We evaluate BOXBOX on PerfRace under matched test-time budgets, comparing it with Frozen Harness, offline Harness Evolution, and Harness Scaling. BOXBOX achieves higher speedup with more efficient use of task time and model cost, reaching a Speedup@4h of 1.70x, compared with 1.54x for the strongest baseline.
self-improvementbenchmark - arxiv:2611.09194 · cs.AILarge Language Model Consensus and Corpus-Adaptive Topic Modeling for Classifying and Mapping Patent InnovationSercan Ozcan, Enes Ozgozler, C. Okan Sakar
Patent analytics increasingly relies on transformer representations and large language models (LLMs), yet linking conceptual classification, topic formation and interpretation requires explicit decisions about model roles and the evidence retained at each stage. In this design-oriented study, we propose and demonstrate a patent analytics pipeline that links these tasks and keeps each decision open to audit. Three LLMs separately classify each patent by novelty (radical or incremental) and innovation type (product, process, service). Radical and innovation-type labels each require two supporting votes, with Incremental used as the fallback novelty label. Topics are discovered by clustering PatentSBERTa embeddings, with parameter ranges scaled to corpus size. A patent-specific stopword list is applied only when topics are represented with class-based term frequency--inverse document frequency (c-TF-IDF), and a second embedding model supports inter-topic similarity, hierarchical refinement and visualization. An LLM then labels each topic from its ranked keywords and representative patents. We demonstrate the pipeline on 30,964 public-sector-related patents retrieved from Lens.org. Across five category corpora ranging from 591 to 29,215 patents, the pipeline produced 74 category-specific topic instances with a macro-average document coverage of 98.1% and topic diversity of 0.900--1.000, and fewer than 1% of classification outputs returned UNKNOWN. Because every model vote and the evidence behind each topic label are retained, analysts can audit and revise the outputs. The demonstration establishes the feasibility and traceability of the pipeline; testing its labels against expert coding and benchmarking it against alternative pipelines are identified as the next steps. The pipeline offers a reusable basis for LLM-assisted patent analysis in technology intelligence.
benchmark - arxiv:2611.09191 · cs.CVBootstrap Your Own AV-Proxies: Adaptive Contrastive and Prototype Learning for Audio-Visual SegmentationJunbo Zhang, Hang Su, Zhaofan Li, Hang Dong +1
Audio-Visual Segmentation (AVS) aims to accurately segment sounding objects in video frames by leveraging audio-visual correspondence cues. However, it remains challenging due to the intrinsic semantic incompleteness within a single modality and the semantic gap between audio and visual representations. Traditional feature-fusion decoding approaches struggle to suppress fusion noise effectively, while recent methods that incorporate data-dependent priors often increase the complexity of modeling audio-visual correlations, leading to poor cross-domain generalization. To address these issues, we propose a novel adaptive contrastive and prototype learning framework, BYOAVP, for AVS. Specifically, we design a Self-Supervised Audio Enhancement (SSAE) module that introduces contrastive learning to adaptively align audio representations with gradient-blocked visual embeddings, thus narrowing the semantic gap between modalities. Furthermore, a Dynamic Prototype Constraint (DPC) module is developed to refine pixel-wise category perception via momentum-based prototype updating, while enhancing localization of sounding regions through cross-modal interaction. Extensive experiments demonstrate that our method achieves SOTA performance across two AVS benchmarks and six sub-tasks, exhibiting strong robustness and generalization ability.
benchmark - arxiv:2611.09189 · cs.LGRoleWalk: training-free structural embeddings for directed graphs that transfer across graphsMaixent Chenebaux
Structural role embeddings describe a node by its position in a graph - hub, bridge, periphery, member of a dense core - rather than by who its neighbours are. Existing methods either learn coordinates specific to one graph (struc2vec, Role2Vec, RESD) or read scales and normalisations from the whole graph (GraphWave, Digraphwave), so the same neighbourhood can receive different coordinates in two graphs; few read edge directions. We present RoleWalk, a training-free embedding that follows a short lazy random walk from each node and records, at each step, two distributions: that of the transition probabilities, through its characteristic function at a fixed grid of frequencies, and that of the walker's distance to its starting node. On directed graphs, walks that go forward then backward add who shares a node's targets or sources. The embedding depends only on a node's neighbourhood within a fixed radius, so its coordinates mean the same thing in every graph. We show that, without curvature, it is the kernel mean embedding of the walk's footprint, that it separates graphs that the Weisfeiler-Leman test cannot, and that pruning small probabilities comes with a per-node error certificate. Empirically, RoleWalk is competitive with the best role embeddings on five classification benchmarks at a fraction of their cost, has the highest scores on real directed roles in Wikipedia adminship, exceeds the published accuracy of Digraphwave at every noise level of its own Enron network-alignment benchmark, transfers centrality predictors to graphs four times larger better than existing inductive embeddings, hand-made features and a trained GIN, and embeds a million nodes in 20 s to 3 min on one CPU core. The code is open source (BSD license): https://github.com/kerighan/rolewalk
benchmark - arxiv:2611.09178 · cs.LGA Taxonomy-Driven Synthetic Dataset for Retrieval Effectiveness and Fairness Evaluation in HiringWarren Jouanneau, Morgane Hoffmann, Emma Jouffroy, Marc Palyart
While machine learning hiring tools are ubiquitous, the need for privacy regulations and commercial sensitivity makes public benchmarking legally complex. We address this challenge by introducing a taxonomy-driven, fully synthetic HR dataset and benchmark to simultaneously evaluate retrieval effectiveness and algorithmic fairness. Grounded in 660K real freelancer profiles, our multilingual corpus features candidate profiles, project briefs, and counterfactual demographic twins. Auditing state-of-the-art embedding models reveals performance divergence from general MTEB scores, alongside fairness disparities in grammatically gendered languages. This demonstrates that general benchmarks are insufficient for HR, requiring tailored evaluations to identify Pareto-optimal models that balance semantic accuracy and fairness.
benchmark - arxiv:2611.09176 · cs.AITailDART: Compute-Efficient Foundation Models for Financial Time-Series ForecastingChiang Qin Zhi, Ranjan Satapathy, Erik Cambria, Joyjit Chattoraj
Although time-series foundation models (TSFMs) achieve strong zero-shot transferability across high-information-density physical systems, applying overparameterized TSFMs to quantitative finance is fundamentally constrained by low signal-to-noise ratios, non-stationarity, and prohibitive autoregressive decoding latency. To reconcile representation learning with institutional operational constraints, this study presents TailDART, a dense 2.68 million-parameter foundation model engineered for financial time-series forecasting. By integrating grouped-query attention into a lightweight backbone, the architecture drastically reduces Key-Value cache memory bandwidth overhead, achieving major inference acceleration (up to 48.1%) over prevailing state-of-the-art TSFMs to enable low-latency, high-throughput streaming execution. Furthermore, multi-asset benchmarks across varying temporal resolutions show that a compact model can remain competitive for market forecasting: TailDART matches the predictive fidelity of substantially larger baselines while demonstrating superior error resilience in high-volatility regimes.
memorybenchmark - arxiv:2611.09170 · cs.CVFIND: Frequency Invariance Disentanglement for Test-Time Adaptation in LiDAR 3D DetectionDongxiao Li, Zixuan Hu, Ling-Yu Duan
LiDAR-based 3D Object Detection (L3OD) is a fundamental 3D perception task, yet its performance is often compromised by domain shifts arising from diverse sensor configurations and environmental conditions. Existing Test-Time Adaptation (TTA) methods primarily leverage self-supervision in the spatial domain, yet they frequently confound semantic content with domain style due to the sparse, non-Euclidean nature of point clouds. To overcome this limitation, we propose FIND (Frequency INvariance Disentanglement), a novel TTA framework that precisely disentangles domain-invariant content from domain-specific interference within the frequency domain. Our methodology utilizes B-Spline fitting to construct locally adaptive filters integrated into a dual-stream architecture: an Invariant Stream extracts robust features to guide pseudo-labeling, while a Specific Stream enforces consistency against domain perturbations. By combining invariance-driven alignment with stability-guided regularization, our approach dynamically extracts robust domain-invariant features while suppressing domain-specific interference. Extensive experiments on cross-dataset adaptation and robust corruption benchmarks demonstrate that FIND significantly outperforms state-of-the-art methods. The source code is publicly available at https://github.com/llldx03/FIND.
benchmark - arxiv:2611.09165 · physics.opticsLow-Rank Iteration for Efficient Calculations of Topological Surface StatesKaiqi Liu, Changsheng He, Zihan Xie, Shulin Sun +2
Surface states are hallmarks of topological phases, and their efficient computation is crucial for topology characterization. Conventional supercell methods (SCM) approximate semi-infinite systems as finite periodic ones, but suffer from high computational cost, artificial inter-surface coupling, and slow convergence near band edges. Recently, boundary problems for semi-infinite systems have been recast as block-tridiagonal equations and solved by the cyclic reduction method (CRM) and the transfer matrix method (TMM). However, these approaches remain resource-intensive, as fine-mesh discretization produces very large matrices. Recognizing the intrinsic sparsity and low rank of these matrices, we develop a low-rank iterative method (LIM) with stochastic trace estimation that substantially reduces complexity and memory without sacrificing accuracy, while remaining straightforwardly parallelizable. Benchmarks on 2D and 3D topological systems show orders-of-magnitude acceleration over CRM and TMM. LIM extends readily to heterostructures, defective interfaces, and non-Hermitian systems, providing an efficient platform for topological and plasmonic surface phenomena.
memorybenchmark - arxiv:2611.09154 · cs.AILearning When to Think: Dual-Reference Offline Optimization for Adaptive VLM ReasoningHongbin Lin, Sizhe Zou, Juangui Xu, Xinyue Xu +5
Reasoning traces improve vision-language models on complex multimodal tasks, but forcing every input to follow a long reasoning process leads to overthinking and unnecessary inference cost. We study adaptive fast/slow thinking for VLMs, where one model answers directly for simple inputs and reasons step by step when needed. Fast answers and slow traces often come from distinct behavior policies, making a single-reference offline objective misweight one mode. We propose DUAL-REFERENCE BOLT (DR-BOLT), which distinguishes the desired fast/slow mixture reference from the sampling distribution and corrects their mismatch in reward-weighted training. Fast/slow partition competition explains routing in the optimal target; an efficient source-wise correction approximates the exact objective when response formats are well separated. Matched Qwen3-VL comparisons use identical candidate banks to isolate the objective from slow-teacher quality. DR-BOLT improves score-length trade-offs over the evaluated offline and online baselines on ScienceQA and three held-out transfer benchmarks, with a 24.1% cross-benchmark slow ratio and a 2.90-fold end-to-end training speedup over reproduced GRPO. These results support adaptive reasoning in the studied two-reference setting, while broader robustness remains open. Our code is available at https://github.com/HongbinLin3589/DR-BOLT.
benchmark - arxiv:2611.09145 · cs.ROAugmentron: Scalable Multi-View Visual Augmentation for Robot LearningTanay Tandon, Aseem Doriwala, Jade Choghari, Catherine Weaver +1
Robot policies trained on visually narrow demonstrations can become sensitive to environmental features that are unrelated to the task. Existing generative augmentation methods can introduce visual diversity, but processing every frame and camera view independently is computationally expensive and may produce inconsistent observations across cameras and timestamps. We introduce Augmentron, a geometry-aware pipeline that generates reusable scene representations and renders them through recorded camera poses. This produces temporally and multi-view consistent environ- ment and workspace variations while preserving the robot, task objects, actions, and annotations. Across three physical manipulation tasks, replacing half of the real world training dataset with Augmentron data improves model success across three tasks for both in out-of-distribution distractor objects, and out-of-distribution workspace variation evaluations. In a separate comparison against alternate augmentation techniques, Augmentron achieves the highest pooled success in all three conditions. The results show that introducing augmented data to a data training recipe can help policies generalize to dynamic real-world settings.
manipulation - arxiv:2611.09142 · cs.LGLearning Generative Dynamics for 3D Molecule Generation via Sequential Neuro-Symbolic ConstraintsZhuoran Li, Xu Sun, Chang Wen Chen, Wanyu Lin
Diffusion models have recently emerged as a powerful paradigm for generating 3D molecules. Despite showing remarkable promise, they learn atomic distributions implicitly from data, without explicitly encoding physicochemical rules into their generative dynamics. Consequently, generated samples can be statistically plausible yet chemically invalid. To address this problem, we propose Sequential Neuro-Symbolic Constrained Diffusion (SensDiff), a framework that integrates symbolic rules into generative modeling, enabling the model to internalize scientific priors during training. Central to SensDiff is the observation that the diffusion trajectory of 3D molecule generation exhibits a coarse-to-fine evolution: valid molecules arise by first mitigating macroscopic geometric conflicts and then refining structures toward microscopic chemical consistency through iterative denoising. Accordingly, SensDiff mirrors this dynamic by sequentially enforcing constraints with adaptive weighting during training, progressively steering generation toward valid molecular geometries. Moreover, SensDiff translates domain knowledge into interpretable generative controls, grounding opaque neural denoising in scientific principles. Experiments on molecular benchmarks and real-world constrained generation tasks confirm that SensDiff consistently improves chemical validity while adhering to physicochemical rules and user-specified targets. The code is publicly available at https://github.com/WanyuGroup/NeurIPS2026_SensDiff.
benchmark - arxiv:2611.09139 · cs.AICARGO: Causal Attribution and Responsibility-Guided Preventive Safe Multi-Agent Reinforcement Learning for Cooperative DrivingJingjia Teng, Yang Li, Wenjie Huang
This study presents Causal Attribution and Responsibility-Guided Optimization (CARGO), a multi-agent reinforcement learning framework for cooperative on-ramp merging in mixed traffic. A shared latent-context posterior and an action-conditioned response model estimate directed behavioral influence, with posterior uncertainty propagated into interaction-edge selection. Complementary attribution mechanisms couple signed influence with marginal team utility for cooperative reward shaping and assign responsibility from positive excess tail risk relative to a safe reference policy. Conditional diffusion generates future responses for action-dependent risk estimation. Responsibility-guided preventive optimization calibrates agent-specific risk-reduction requirements to local capacity within a policy trust region, using reserved safety margins, adaptive requirement scaling, and joint feasibility verification. Theoretical analysis establishes risk monotonicity and sufficient local conditions for non-decreasing environmental return under explicit assumptions. Experiments in highway-env across three traffic densities, together with controlled graph-recovery and attribution studies, support the effectiveness of the framework. Under high-density traffic, CARGO increases the success rate by 66.04\% and reduces the crash rate by 56.23\% relative to the best evaluated baseline for each metric.
action-conditionedmulti-agent - arxiv:2611.09130 · cs.ROAuditing Grasp Feasibility Predictors with Planner-Verified CounterfactualsSajid Ansari, Antony Thomas, Girish Varma
Grasp-feasibility predictors report near-perfect accuracy on Task and Motion Planning benchmarks, but high accuracy on these benchmarks does not verify that a predictor uses the collision geometry that the motion planner checks. We propose a planner-verified counterfactual (CF) test: holding the grasp pose fixed, we insert a solid obstacle into the approach corridor, add its points to the model input, and let a planner re-derive the label, which flips from feasible to infeasible. Across eight predictors spanning point clouds, voxels, depth, and object tokens, CF-pair AUROC falls from high natural AUROC (0.920-0.996) to chance (0.470-0.504), and the flip holds across four planners, two arms, and a real-object obstacle. Predictors fail at different resolutions, from a bounding box to a 12.5 cm voxel. In the point-cloud models we dissect, the obstacle either moves the representation in a direction unrelated to whether it blocks the grasp or is not registered at all. A clearance-aware read of the corridor points, trained on the unmodified benchmark with no counterfactual scenes, reaches 0.983 CF-pair AUROC, but it remains bound to its trained obstacle placements; a placement-invariant clearance representation is still open. We release the counterfactual labels, metrics, datasheet, and relabeling pipeline at https://github.com/codewithsajid/graspfc-cf.
graspbenchmark - arxiv:2611.09129 · cs.CVTouchAnnotator: Multi-Source Tactile Supervision with Privileged-View DistillationAo Li, Zhihao Yuan, Chubin Zhang, Kunrui Huang +5
Egocentric video captures human-object interaction at scale but rarely includes tactile labels for hidden contact and pressure. We introduce TouchAnnotator, an ego-only bimanual tactile annotator trained with multi-source tactile supervision and privileged-view distillation. A canonical hand field unifies glove pressure, geometry-derived human contact, and robot fingertip deformation from six datasets and three embodiments, while source-aware masks preserve valid targets. During training, an EMA teacher observes synchronized ego and wrist views and distills hand-local representations and tactile predictions into a student that uses only ego video and estimated pose at inference. Compact past-clip memory supplies temporal context. We evaluate on EgoTouch, zero-shot FPHA transfer, and 860 synchronized glove recordings. The model transfers to bare-hand contact and produces qualitatively coherent tactile responses across bare-hand, gloved-hand, and robot embodiments. On measured recordings, it improves temporal accuracy by 5.38 points and Volumetric IoU by 5.02 points under a shared protocol. These results show that heterogeneous human--robot supervision and privileged training views can turn egocentric video into dense tactile annotations.
tactilememory - arxiv:2611.09124 · cs.ROToward the Era of Agentic Embodied IntelligenceChao Liang, Ke Chen, Fengxiaoxiao Li, Shibo Huang +6
Foundation models increasingly guide robot actions, yet sustained autonomy requires efficient control and failure recovery. We introduce A2-MEAS, an embodied multi-agent framework built around code as a scaling interface. Learned correction codes combine corrective language, spatial evidence, execution history, and action proposals. MetaBrain provides training supervision and runtime guidance. A correction encoder and adapter absorb this knowledge into policy weights during training, while fresh codes adapt behavior with frozen weights at deployment. The action model, A2-WLA, adds motion-aware history and a latent world model to a vision-language-action policy. Surprisingly, this simple design with action pretraining and task specialization reaches 0.5755 Q-score on the 50-task BEHAVIOR 2026 public subset before online correction, 2.21x the 2025 winner's reported 0.2605 across different benchmark editions and training configurations. On all 100 BEHAVIOR 2026 public tasks, A2-MEAS achieves 0.6411 Q-score and 45.40% task success. On RoboTwin 2.0, A2-WLA achieves 93.98% average success, rising to 95.05% with the complete system. Across 12 real-robot tasks in five household settings, A2-WLA achieves 57.92% average success versus 26.67% for pi0.5. In long-horizon pour-over coffee preparation, a history-conditioned adapter raises post-failure completion at two filter-handling steps from 14% to 94% and from 8% to 82%, compared with A2-WLA retries. Our training infrastructure delivers 427.1 samples per second, 4.03x the throughput of a pi0.5 reference implementation on matched hardware. We will release the system, models, training scripts, and infrastructure to support efficient embodied learning.
vision-language-actionembodiedpi0robotwinworld modelmulti-agent - arxiv:2611.09121 · cs.CVCoMET: Tracking the Shared Pose Error of Roadside Messages for Asynchronous Vehicle-Infrastructure Cooperative 3D DetectionShuwei Sun, Jian Huang, Hui Cao
A roadside sensor can detect vehicles hidden from the ego vehicle, but a pose error shifts every detection in the same message. We study asynchronous vehicle-infrastructure cooperative 3D detection on V2X-Seq using only information available at decision time, without ground truth, future frames, or per-pair calibration offsets at inference. Objects seen by both sensors reveal the shared translation, but even at zero added delay, 38% of validation messages have no matched pair. We propose CoMET (Common-Mode Error Tracking), which estimates this translation over time. At each message arrival, CoMET matches vehicle and roadside detections, estimates the offset from mutually nearest pairs, and updates a Kalman filter for that sender. When no pair matches, the filter carries its estimate forward. The correction applies to all roadside boxes, including those unseen by the ego vehicle. The estimator has no learned parameters; a separate frozen head ranks the merged detections. Across four delays from 0 to 400 ms on 21 validation sequences, CoMET averages 66.63 bird's-eye-view AP at IoU 0.7, compared with 63.18 for the same late fusion using uncorrected online poses and 62.96 for the best intermediate-fusion baseline. Across five detectors from three architectures, CoMET gains 2.2 to 3.5 AP points over online poses and 0.3 to 0.6 over estimating the offset from the current message alone. We find that the benchmark uses a zero offset as a placeholder in 28% of validation messages and that its offset search reaches the boundary in 45% of the rest. These offsets affect the labels for objects seen only from the roadside. Under two alternative label references, CoMET improves on uncorrected online poses in every evaluated population. Retraining intermediate fusion with CoMET poses and corrected labels improves detection on all three architectures tested, with statistically significant gains on two.
benchmark - arxiv:2611.09113 · cs.LGAuthorization Drift in LLM Agents: Evaluating Boundary Compliance in Multi-Step Tool UseChuxue Cao, Jingru Zeng, Boqin Yin, Yujin Zhou +6
LLM agents can exceed their authorization boundaries while pursuing user goals without an adversary. We formalize this phenomenon as authorization drift across data, tool, scope, and time, and introduce the Authorization Drift Benchmark (ADB): 648 samples across seven domains, each pairing a sufficient compliant path with an attractive unauthorized shortcut. To ensure executable tasks, tempting shortcuts, and feasible compliant paths, Staged Execution-Reflective Synthesis (SERS) grounds benchmark validation in real agent rollouts rather than LLM self-critique. Across 13 LLMs and three frameworks (25,207 trajectories), greater capability accompanies greater drift, with frontier-capable models exhibiting drift rates of 0.43 to 0.61. Yet 86.18% of tasks admit an observed clean-success trajectory, showing that drift is avoidable. Drift is persistent and dominated by early data over-collection; apparent improvements under agentic scaffolds coincide with collapsing clean-completion performance. To understand potential harms to individuals and society, we map drift points to legal provisions across four jurisdictions, connecting overreach to privacy and information-security concerns. Data minimisation is the most frequently implicated provision, and 93.5% of legally grounded drifting runs map to multiple jurisdictions, making the potential consequences of authorization drift more interpretable.
agentllm agentagentictool usebenchmark - arxiv:2611.09112 · cs.ROSafe Circumnavigation on Heterogeneous Orbits: A Range-Only HOCBF-QP Framework for Nonholonomic Multi-Agent SystemsShubham Singh, Anand Singh, Anjali Tiwari, Lara Briñón-Arranz +1
Circumnavigating a stationary target with a team of robots requires each robot to hold its own standoff distance, to stay evenly spaced from its neighbours, and to avoid collisions with the target and with each other at every instant. The task becomes arduous when the robots are assigned different orbital radii, as their sensors or roles may demand. This paper presents a safety-critical control framework for this problem. Each robot commands both its acceleration and its turn-rate, measures only its distance to the target, and exchanges information only with its two ring neighbours. The design has three features. First, a structural property of the model lets one control law serve every robot regardless of its assigned radius. Second, the acceleration drives each radius exactly to its target, while the turn-rate balances the formation with the neighbours. Third, a higher-order control-barrier-function quadratic program enforces target, geofence, and inter-robot safety on both inputs at once; it leaves the nominal controller untouched away from the constraints and stays valid while the unmeasured range rate is still being estimated. Simulations confirm these guarantees. In a comparison, the closest existing multi-orbit design, even with full state information, violates both the target standoff and the inter-robot margin, whereas the proposed controller does not.
multi-agentagent system - arxiv:2611.09107 · cs.CLPhoPunct: A Large-Scale Benchmark of Transformer Architectures for Vietnamese Punctuation RestorationKhoa D. Nguyen, Thai Bao Huynh, Binh T. Nguyen
Vietnamese punctuation restoration is often evaluated on relatively small benchmarks, limiting conclusions about how architectural choices scale with data volume. We present a controlled benchmark on News and Novels corpora containing 22.1 million labeled tokens (19.9M News and 2.2M Novels). We evaluate 17 architectures formed from four pretrained Transformer backbones and optional bidirectional long short-term memory (Bi-LSTM) and conditional random field (CRF) layers, together with a PhoBERT-large-Bi-LSTM-CRF model, PhoPunct. Every model is trained for eight epochs with the same learning rate and without architecture-specific tuning. Under this fixed budget, adding Bi-LSTM and CRF layers does not reliably improve macro F1: in seven of eight matched domain-backbone comparisons, a plain Transformer classifier matches or outperforms its Bi-LSTM-CRF counterpart. PhoPunct obtains the best News result (61.35), whereas XLM-R with a CRF obtains the best Novels result (50.18), exceeding PhoPunct (48.97). These results show that the value of recurrent and structured-prediction heads depends on their optimization regime and motivate explicit separation between architectural improvements and improvements due to architecture-specific tuning. Our source code is available at: https://github.com/aisia-research-lab/PhoPunct
memorybenchmark - arxiv:2611.09100 · cs.CVBayesCrop: Real-Time Probabilistic Multi-Crop Video ReframingCheng-Han Lee, Rajesh Sureddi, Neil Birkbeck, Yilin Wang +2
Automatic video reframing is increasingly important for adapting landscape content to heterogeneous display formats. Existing methods typically predict a single deterministic crop, despite the subjective nature of reframing and the presence of multiple plausible regions of interest. We present \textbf{BayesCrop}, a real-time probabilistic framework for multi-crop video reframing. BayesCrop represents crop preferences using a two-component spatially isotropic Gaussian mixture over crop position and scale. Each component defines a representative reframing mode together with its local uncertainty, enabling both dual-view prediction and sampling of additional plausible crop candidates. For efficient video processing, we introduce an anchor-and-propagate strategy that performs neural prediction only on sparse anchor frames and propagates crop trajectories using motion vectors available from the compressed video bitstream. The resulting trajectories can be combined into a single output using a lightweight periodic view-switching strategy. Experiments on LIVE-YT-MC, LIVE-YT-VC++, and a multi-target long-video benchmark demonstrate competitive single-output localization, high multi-crop reference coverage, diverse yet coherent crop hypotheses, and stable perceptual quality. BayesCrop achieves real-time throughput of 112 FPS at the default operating point, while sampling from the learned distributions further improves reference coverage beyond the two representative crop means.
benchmark - arxiv:2611.09096 · cs.CVTrajectory Forcing: Exploiting Diffusion Trajectories for Autoregressive Long Video GenerationBohan Wang, Shuo Chen, Yunzhi Li, Zhaozheng Chen +2
Autoregressive (AR) modeling is arguably the only way to extend a fixed-length video diffusion model to longer or even streaming videos. However, existing AR diffusion methods use only the previously predicted chunk as context for the next, discarding the intermediate predictions that constitute the chunk's generation history. In contrast, conventional AR models such as LLMs condition on the entire history of preceding tokens. To this end, we propose Trajectory Forcing (TraF), an AR diffusion method that exploits the denoising trajectories of previous video chunks as context. TraF operates under two training regimes: (1) with ground-truth videos, where pseudo-trajectories are constructed from Gaussian interpolations and re-predicted by the model, and (2) with a teacher model, where real trajectories are obtained from the student's own denoising rollout. The first regime naturally provides a strong initialization for the second. On the 5-second VBench benchmark, TraF achieves a Total score of 85.15, outperforming Self Forcing (+0.84) and Causal Forcing (+1.11), with Quality and Semantic improving simultaneously. On 30-second generation---6$\times$ the training horizon---TraF achieves the highest overall quality among all compared methods.
benchmark - arxiv:2611.09094 · cs.AIAlienBench: A Benchmark of Structural Novelty in Language Model OutputsJonas Oppenlaender, Johanna M. Silvennoinen
Large language models (LLMs) are increasingly used for creative tasks such as writing, design, and idea generation. For these uses, it matters whether a model's ideas depart from familiar examples or stay close to them. Novelty is one standard measure of creativity, and current creativity evaluations ask human experts or LLM judges to rate how novel an output is. However, these ratings have no fixed reference for what counts as novel (or, in other terms, departure from the familiar). Measuring this departure requires a task with known familiar defaults. The structured imagination task from cognitive psychology is such a task. In this creativity test, people imagine a creature from another planet. The departure from features of Earth animals, such as bilateral symmetry and legs as mode of locomotion, is used as measure of creativity. We introduce AlienBench, a benchmark that adapts the structured imagination task to language models. Models describe alien creatures, and a panel of three LLM judges extracts ten biological features from the description and labels them as either Earth-typical or a departure. The sum of feature departures is the score of a description, from 0 to 10. Each model describes alien creatures under six prompt conditions that vary the framing of the task and the planetary environment. Each model generates 30 descriptions per condition (180 in total), and its overall score is the mean across these descriptions. We evaluate 10 language models, from frontier to open-weight models. Overall scores range from 3.28 (GPT-6 Luna) to 6.55 (Ministral 3 14B). Interestingly, open-weight models score higher than proprietary models in this task (5.40 vs. 4.58). Under the baseline prompt, the models depart from Earth defaults more often on average than 100 human participants on every feature. We release all prompts, generations, judge outputs, and scores.
benchmark - arxiv:2611.09092 · cs.CVAdaAlloc: Adaptive Visual Token Allocation for Long-Video Question AnsweringJoungbin An, Kristen Grauman
Long-video question answering forces multimodal large language models to reason under strict visual-token budgets, creating a tradeoff between broad temporal coverage and fine-grained spatial detail. Uniform sampling preserves coverage but often obscures decisive local cues, while caption-based or frame-similarity retrieval operates over lossy proxies that can miss motion, state changes, and before-after relations. We recast long-video reasoning as a visual-token allocation problem: given a fixed budget, decide how much to spend on global context versus high-resolution local evidence, and where in the video to draw that local evidence from. We introduce AdaAlloc, a training-free inference framework that adaptively allocates visual tokens between a global video overview and localized visual detail. AdaAlloc first plans a global/local allocation policy, then uses temporal grounding to locate candidate evidence segments and refines them into a compact temporal memory. The selected policy controls both frame sampling density and resolution, combining sparse, lower-resolution global frames with denser, higher-resolution local frames drawn from the refined memory. The final answerer reasons over the allocated global context and refined local evidence. Across LVBench, MLVU, VideoMME, and LongVideoBench, AdaAlloc improves over budget-matched uniform sampling by up to 10.5 percentage points and outperforms the compared frame-selection and agentic methods on their reported benchmarks.
agenticbenchmark - arxiv:2611.09090 · cs.CLMARS: Training Multilingual Search Agents for Retriever-Free Evidence AcquisitionRui Qi, Yunlong Liang, Yufeng Chen, Jinan Xu +2
Existing multilingual retrieval-augmented generation relies heavily on learned retrievers and similarity-based ranking from multilingual collections, which are vulnerable to cross-lingual retrieval errors and indexing overhead. In this paper, we explore retriever-free multilingual search, where the model directly interacts with raw corpora through executable shell commands. This paradigm shifts the bottleneck to multilingual search action generation, where exact search is sensitive to entity surface forms and outcome-only supervision provides limited guidance for multi-turn search. To address these challenges, we propose MARS, a multilingual search agent in direct corpus interaction (DCI) with reinforcement learning (RL). MARS introduces corpus-native entity grounding to recover reliable lexical anchors from corpus evidence, mitigating failure to retrieve evidence. Meanwhile, we present a redundancy-aware search reward to encourage concise and effective search trajectories, addressing training collapse in RL of multilingual scenarios. Experiments show that \ours~consistently outperforms strong retrieval-based and DCI baselines, while demonstrating robust cross-lingual generalization and substantially reducing redundant corpus interactions.
retrieval-augmentedagent - arxiv:2611.09089 · cs.CVEventHOI: Event-Driven Resource-Efficient Hand-Object Interaction Logging on SmartglassesLe Zhang, Kailai Cui, Hao Chen, Vlad Roznyatovskiy +4
Hand-object interactions (HOIs) can be captured by smartglasses to learn how people manipulate objects in daily life. Automatic logging of these interactions from an egocentric perspective provides time-stamped evidence of daily actions for applications like memory assistance. However, continuous RGB capture and wireless transmission exceed the energy and bandwidth budgets of smartglasses. We present EventHOI, an event-driven sensing and compression system that couples selective RGB capture with HOI-aware encoding to enable egocentric logging over Bluetooth Low Energy (BLE) while preserving downstream HOI logging performance. Our key insight is to use low-power event sensing both to determine when RGB capture is needed and to identify where visual detail must be preserved. A lightweight IMU-based egomotion cancellation pipeline suppresses egomotion-induced events so that residual HOI motion can guide RGB triggering and spatial compression. We build a smartglasses prototype and conduct an end-to-end evaluation of EventHOI on a 50-hour egocentric benchmark with 97 subjects and a 20-hour real-world dataset we collect from 10 participants, together covering a broad range of daily-life scenarios and environments. Compared to RGB-only baselines, EventHOI reduces always-on backend power consumption by 5.8-10.4x and uplink bandwidth by 2.9x, with minimal HOI recognition accuracy loss. The resulting RGB uplink of less than 130 kbps fits within commodity BLE bandwidth budgets, making stable BLE video streaming feasible for egocentric HOI logging. Demo video and project website: https://eventhoi.github.io
memorybenchmark - arxiv:2611.09088 · cs.AIWhen Routine Operations Break Architectural Pattern Relationships in LLM Agent HarnessesBoming Xia, Qinghua Lu, Dino Sejdinovic, Liming Zhu
LLM agent harnesses distribute tool mediation, state management and safety guardrails across native services, independently written plugins and configuration. Routine operations can change the runtime order or retained state that architectural pattern relationships require, without any code change. We study five relationships associated with the Guardrail Enforcement Point, Autonomy Controller and Blackboard patterns across five released harnesses, analysing a main sample of fourteen independently written plugins. Using PatternOps, our assessment procedure, we trace requirements from pattern descriptions, provider documentation and deployment profiles to the providers that realise them and the records that decide them. We then compare alternative allocations under supported operations. Such operations broke relationships silently: after a plugin was reapplied or an allow rule saved, policy stopped evaluating edits while every plugin stayed active, and a prohibited edit executed. Which alternatives qualified depended on the requirement. Allocations that kept the required order, or kept or rebuilt the needed state, held on their tested paths and operation sequences, while others prevented prohibited effects but missed edits rejected elsewhere. In these harnesses, a pattern relationship is a property of the running system and its history, so architects should evaluate it against the decision paths, state lifetimes and operation sequences their deployments use.
agentllm agent - arxiv:2611.09085 · cs.AIThe Invisible Gap: Romanized Indic Generation Fails Where Automatic Metrics SucceedMadhu Sudhan Rao Beesetty, Surbhi Sharma, Devesh Maheshwari, Lokesh Meesala
Hundreds of millions of Indians communicate with digital systems in romanized Latin characters -- typing "aap kaise hain" rather than native script -- yet romanized Indic generation is absent from NLP benchmarks and evaluation frameworks. We demonstrate that automatic metrics (BGE-M3, BERTScore) systematically misrepresent the quality of romanized Hindi, Telugu, and Kannada generation. We evaluate four LLMs (Claude Opus 4.8, GPT-5.5, Gemini 2.5-flash, Qwen3.5-35B) on 142 OpenAssistant query-response pairs. On a 72-response diagnostic subset, native speakers rate GPT-5.5 1.43 points higher than Claude Opus 4.8 on a 12-point rubric, whereas their full-dataset BGE-M3 means differ by only 0.004. Two distinct model-specific failure modes emerge: Gemini produces script spillage (16.9% of Telugu responses leak native script), while Qwen collapses to English-only output (16.2% of Kannada responses). We introduce a failure-aware evaluation policy that assigns zero to empty outputs, script-corrupted responses, and instruction collapses -- revealing failures metrics alone cannot detect. Our results show that evaluating romanized Indic generation requires a coordinated bundle: semantic similarity, script compliance, failure coverage, and native-speaker acceptability.
benchmarkevaluation framework - arxiv:2611.09084 · cs.CVTraining-Free Active Test-Time Adaptation for Vision-Language ModelsJihwan Bang, Sumyeong Ahn, Hwanjun Song, Jae-Gil Lee
Test-time adaptation (TTA) for vision-language models (VLMs) is usually studied in a fully unsupervised regime, where adaptation relies on pseudo-labels, confidence scores, or entropy computed by the model itself. This is a severe restriction on hard shifted samples: the model is asked to repair its own uncertain predictions without any external evidence. At the other extreme, supervised online adaptation assumes labels can directly train or correct the model. We study the intermediate deployment regime between these two extremes. In Active Test-time Adaptation for VLMs (Active VLM-TTA), a pretrained VLM may spend an explicit labeling budget on selected test samples, but the prediction for the current sample must be emitted before its label can be used; queried labels are therefore costed, delayed evidence for future samples rather than free correction for the present one. Under this future-only constraint, the central question is how to reuse sparse labels conservatively. We instantiate this setting with ACTive VLM-TTA for Online Retrieval (ACTOR), a training-free plug-in module that routes uncertain samples among base prediction, label query, and memory-based correction using a consistency-and-reliability check over previously queried labels. Across ten cross-domain benchmarks and a domain-transfer benchmark based on ImageNet, ACTOR consistently improves strong VLM-TTA baselines on ResNet and ViT backbones with negligible inference overhead. Our implementation is publicly available at https://github.com/kaist-dmlab/ACTOR.
benchmark - arxiv:2611.09077 · cs.CVGlobalPuzzle: Globally Informed Genetic Search for Eroded Jigsaw Puzzle ReassemblyLunyan Wang, Chuanzhi Xu, Ziyuan Tao, Weidong Cai
Jigsaw puzzle reassembly remains a highly challenging task, particularly when piece boundaries are irregularly eroded to simulate damage to cultural artifacts. On the GAP benchmark, existing methods achieve a perfect assembly accuracy of only 0.3% on $5 \times 5$ puzzles, whereas humans can solve jigsaw puzzles with remarkably high accuracy. Humans typically approach puzzle solving by identifying potentially adjacent pieces, evaluating the current assembly, and iteratively performing complex operations such as adding pieces, replacing incorrectly placed pieces, and moving entire connected groups of pieces. However, existing methods face a fundamental limitation: approaches that leverage global visual information typically predict piece arrangements directly, making it difficult to explicitly perform such complex iterative operations, while search based methods can rearrange pieces but primarily rely on pairwise boundary compatibility, lacking global visual information to evaluate and guide the search process. To bridge this gap, we propose GlobalPuzzle, a jigsaw puzzle reassembly framework that integrates global canvas level evaluation, action guidance, and genetic search. By incorporating global visual information into the evaluation and optimization of candidate assemblies, GlobalPuzzle aims to resolve ambiguous adjacency relationships that are difficult to determine from irregularly eroded piece boundaries alone, enabling flexible, globally informed reconstruction.
benchmark - arxiv:2611.09074 · cs.CVEmo-LiveSync: Real-Time Universal Lip Synchronization with Explicit Emotion ControlSangjoon Yu, Seungeun Lee, JoonHeyok Shin, Ji-Su Kang
Lip synchronization, the task of generating realistic lip movements synchronized with audio signals, is fundamental to applications such as film dubbing, virtual avatars, and digital content creation. Recent GAN-based methods offer computational efficiency but often lack visual fidelity, while diffusion-based approaches achieve high-quality results at the cost of substantial computational overhead. Moreover, most existing methods are limited to photo-realistic human faces, restricting their applicability to stylized characters and non-human entities. In this paper, we propose an efficient universal lip-sync video generation framework that achieves high-quality results without compromising computational efficiency. We build on LivePortrait, a warp-based portrait animation model, and introduce a training strategy that aligns audio signals to an implicit motion space by distilling knowledge from a pre-trained video-to-motion model. This enables real-time inference while maintaining high visual fidelity, and naturally generalizes to diverse stylistic characters beyond photo-realistic humans. Furthermore, we introduce an expressive control gate that enables explicit manipulation of emotional expressions for content enrichment. Extensive experiments demonstrate that our method achieves state-of-the-art performance in both visual quality and lip-sync accuracy while ensuring real-time inference.
manipulation - arxiv:2611.09069 · cs.CVUniRGBA: Unifying RGB and Alpha in Pretrained Video Diffusion ModelsLei Xu, Zhuo Li, Yongyi Jin, Aoyu Wang
Recent advances in video diffusion models have achieved remarkable photorealistic RGB synthesis, yet generating videos with a native alpha channel remains an open challenge. Existing approaches often rely on matting-style data and non-native RGBA designs, causing them to specialize in real-object scenes and to degrade severely on visual-effect content whose semi-transparency is spatially non-local and non-elliptical. We present UniRGBA, a unified framework whose key insight is that RGBA generation does not require a dedicated alpha encoder or complex latent-space restructuring---a minimal, targeted extension of the existing RGB pathway is sufficient. UniRGBA-VAE extends a causal video VAE to 4-channel RGBA via zero-initialized channel expansion, symmetric selective layer unfreezing (5% trainable), and a six-term progressive reconstruction loss anchored by a KL reference loss that prevents latent drift, followed by a two-stage schedule that introduces a lightweight GAN objective. We then apply standard DoRA to the pretrained DiT, reusing the unified 4-channel latent without any alpha-specific architectural changes. On a visual-effect RGBA benchmark, UniRGBA substantially outperforms Wan-Alpha on RGBA reconstruction while preserving RGB fidelity. For generation, UniRGBA outperforms Wan-Alpha on RGBA text-to-video tasks and matches or surpasses the original pretrained Wan2.1 backbone on RGB-only metrics, demonstrating that native alpha support can be integrated without compromising the underlying RGB prior or the model's architectural versatility.
benchmark - arxiv:2611.09061 · cs.CLDoes Reasoning Improve Psychological Depth in Large Language Models?Ruichen Zheng, Yihe Wang, Fabrice Harel-Canada, Sara Khosravi +3
Benchmark gains from reasoning-enhanced large language models encourage the assumption that these models should broadly perform better on complex generative tasks, but this remains underexplored in open-ended, creative domains shaped by nuances of reader experience. We tested this assumption through the lens of psychological depth in short stories. We compared DeepSeek-R1 vs. DeepSeek-V3, our closest available contrast for reasoning-oriented post-training, and GPT-5 vs. GPT-4o, through 60 blinded pairwise evaluations in which seven human annotators and an ensemble LLM-as-a-Judge rated prompt-matched story outputs. Human evaluations showed no universal reasoning advantage: GPT-5 was slightly preferred over GPT-4o (60.0-62.9% across two reasoning-effort settings), whereas DeepSeek-R1 fell behind DeepSeek-V3 (42.9%). Although this pattern was marked by low inter-annotator agreement (Krippendorff's $α= 0.070$), our annotation-quality checks suggested genuine evaluative subjectivity, plausibly reflecting distinct annotator decision schemes rather than annotation failure. In contrast, the LLM-as-a-Judge showed a strong pro-reasoning bias (98.3% of pairs vs. 55.2% of human votes), consistent across its five configurations and partly tracking sentence length. These findings suggest that, at the frontier, reasoning-related quality differences are small enough to be dominated by reader subjectivity, and that LLM-as-a-Judge evaluators can obscure this subjectivity in nuanced creative-generation tasks.
post-trainingbenchmarkevaluator - arxiv:2611.09059 · cs.RORobust Readout of Cross-Point Resistive Arrays via Time-Domain OptimizationTaro Tako, Hiroshi Kaminaga, Taisuke Kobayashi
Cross-point resistive arrays are attractive for high-density tactile sensing because they reduce wiring complexity. However, conventional frequency-based readout methods can become inaccurate in robotic sensing environments, where long interconnects and measurement and environmental noise degrade the observed signals. We propose a time-domain optimization method for robust readout of cross-point resistive arrays. The method determines resistance values by fitting a time-domain observation model to the measured waveform over the interval, rather than relying solely on spectral amplitudes. By exploiting the full waveform, the proposed method improves robustness under noisy observation conditions. In simulations of a 16 x 16 array, the proposed method reduced the mean readout error by 20% compared with the baseline method. Furthermore, validation over 1,000 trials showed that the proposed method achieves more robust estimation than the conventional method, reducing the mean error and standard deviation by 24.6% and 77.6%, respectively. The improvement was also confirmed in cross-point array simulations with resistance distributions emulating tactile sensor responses.
tactile - arxiv:2611.09058 · cs.LGProtoCouncil: Multi-Encoder Anatomical Prototypes for Prostate Cancer Detection in Micro-UltrasoundMohammad Mahdi Abootorabi, Nooshin Maghsoodi, Armin Saadat, Obed Dzikunu +7
Purpose: Micro-ultrasound ($μ$US) is a promising modality for prostate cancer (PCa) detection, but identifying suspicious tissue depends heavily on clinical experience. Models trained on sparse, noisy core-level pathology offer little insight into their decisions and can fail under domain shift. Clinical use needs a traceable decision process and a flag for uncertain predictions. Methods: We introduce ProtoCouncil, in which three foundation encoders (MedSAM, DINOv3 and V-JEPA 2.1) learn case-based prototypes at three anatomical levels: whole gland, needle track and gland sectors. Prototypes are projected onto real training-core regions, and label-free fusion keeps every class logit as a bias plus an exact sum of prototype similarity contributions. Encoder disagreement triggers selective prediction. Results: In five-fold patient-level cross-validation (6,588 cores, 693 patients), ProtoCouncil reaches 82.9% AUROC for clinically significant PCa (csPCa, Grade Group $\geq$3), 4.3 points above the best previously published method. In zero-shot evaluation on two external cohorts, ProtoCouncil achieves higher all-cancer and csPCa AUROC than all previous methods. Nearest same-class exemplars match test-core tumor involvement better than random same-class cases. Externally, csPCa accuracy falls from 79% when encoders agree to 27% when two oppose the fused call. Abstaining on disagreement raises balanced accuracy by 7 and 15 points, while retaining 58% and 54% of the cores. Conclusion: ProtoCouncil grounds core-level predictions in exact similarity-based evidence from real, anatomically indexed training cases, which makes its reasoning interpretable. It achieves the highest csPCa AUROC across three cohorts and identifies unreliable predictions through encoder disagreement.
v-jepa - arxiv:2611.09054 · cs.MAFrom Emergence to Reversibility: Toward Reversible Multi-Agent SystemsZhengye Han, Linda Ng Boyle
Multi-agent systems become capable through institutional mechanisms, such as rules that route information to agents that act on it. These mechanisms are also natural places to restrict the system. Yet as agents learn, share memory, delegate, and repair failures, the collective can find other paths to the same effects: a restriction on the original mechanism still operates but no longer stops them, while routine performance and compliance checks pass. We propose \textbf{Reversible Multi-agent Systems} as a research agenda: collectives that grow while an authorized supervisor can still restrict specified effects, keep them restricted as agents adapt, and reconstitute useful cooperation under renewed authority. What is reversed is the ability to produce a specified effect; restrictions must hold even if learned knowledge persists. The goal is to preserve effective intervention, not a particular mechanism: at every step of growth, some authorized intervention must keep a specified prohibited effect rare while maintaining required service. We identify three research problems: tracking what capabilities depend on, handing off control before an old mechanism loses its effect, and separating legitimate repair from unauthorized reconstruction. We outline how to evaluate control across growth paths, back control claims with evidence, and allocate the rights to restrict and restore.
multi-agentagent system - arxiv:2611.09053 · cs.LGEquidistribution of Solution Gradients in Physics-Informed Neural Networks via Coordinate TransformationsTianchi Yu, Ivan Oseledets
Physics-Informed Neural Networks (PINNs) often struggle to approximate solutions with sharp gradients due to optimization difficulties and spectral bias. Motivated by the principle of gradient equidistribution from the stretching approach in grid generation, which redistributes solution gradients to reduce functional complexity, we propose Coordinate Transformation PINNs (CTPINNs), which incorporate learnable coordinate transformations to map the physical domain into a computational domain where the physical solutions become approximately affine, making them easier for neural networks to learn. We propose two constructions of the transformation: an explicit construction based on Majorants and an implicit construction based on invertable neural networks. Experimentally, the explicit construction can improve accuracy for PDEs with available explicit coordinate transformations, and the implicit construction can be combined with other advanced PINN models as an efficient plug-in module to improve performance across a range of challenging PDE benchmarks.
benchmark - arxiv:2611.09047 · cs.CVGrounding-Aware Judge Reward Reinforcement Learning for Out-of-Domain Face Anti-SpoofingXiaoyong Yu, Xinge Yu
Face Anti-Spoofing (FAS) must distinguish genuine faces from presentation attacks. A unified Multimodal Large Language Model (MLLM) is an attractive solution: it detects physical attacks (print, replay, mask) in one model and produces interpretable forensic evidence via Chain-of-Thought (CoT). A central obstacle to post-training such a model with reinforcement learning is that FAS reasoning has no rule-verifiable ground truth: only the binary verdict is checkable, while the reasoning chain cannot be verified by any hand-crafted rule. Optimizing the verdict alone is therefore a sparse, hackable signal. The model can reach the correct answer through a shortcut such as a false-rejection bias, which does not generalize. We address this with a grounding-aware judge reward. A 32B vision-language judge scores each reasoning chain against the image on evidence anchoring, logical coherence, and domain knowledge, turning the image itself into the verification signal. The reward is applied in a Group Relative Policy Optimization (GRPO) stage after CoT supervised fine-tuning. It is the judge, rather than the RL objective, that supplies the process supervision a verdict-only reward cannot. Our study starts from a model that already performs well In-Domain (ID) after CoT-based supervised fine-tuning and asks how to improve its Out-of-Domain (OOD) generalization. We find that grounding-aware rewards are an effective means to that end. We validate the method on OOD generalization, where the reward reduces ACER on two physical benchmarks (LCC-FASD 0.288 to 0.273, MSU-MFSD 0.431 to 0.409) while preserving ID accuracy (CelebA-Spoof 0.033 to 0.035). Against ResNet-50/101 and CLIP-ViT-B/16 baselines, the MLLM dominates ID and matches or surpasses the strongest baseline OOD. To our knowledge this is the first use of LLM-as-judge process supervision for FAS.
post-trainingbenchmarkllm-as-judge - arxiv:2611.09042 · cs.LGFailure Profiling and Reachable Trajectory Selection for Reasoning DistillationYijun Zhu, Jianxin Wang, Chengchao Shen
Training smaller language models on reasoning solutions generated by a larger teacher model is the prevailing approach for transferring mathematical reasoning capabilities. However, the student tends to overfit the output distribution of the teacher and generalizes poorly when its own behavior at inference time deviates from the training data. Recent methods attempt to bridge this gap by incorporating error data, yet the corrective trajectories still reflect the reasoning patterns of the teacher, which the student cannot reliably reproduce, leaving the distribution mismatch largely unresolved. We propose a method, Failure Profiling and Reachable Trajectory Selection (FPRT), built on the insight that effective training data should be jointly determined by what the student fails on and what it can realistically learn. Our method first profiles the failure behavior of the student on each problem through repeated sampling, characterizing it in terms of error rate and error consistency to reveal qualitatively distinct failure modes. Each mode demands a fundamentally different training signal. The diagnosed mode then guides candidate generation from the teacher via distinct prompting strategies tailored to each failure type. Among the correct candidates, our method selects the one whose initial reasoning steps after the point of divergence are most reproducible by the student. These first corrective steps constitute the primary bottleneck, while the remaining steps follow an already corrected context and can therefore be continued more readily. This contrasts with standard reranking approaches, which score the trajectory as a whole and thereby dilute the signal at the critical reasoning transition. Experiments demonstrate that our method achieves the SOTA performance across the five benchmarks on both student backbones.
benchmark - arxiv:2611.09041 · cs.RODobotTrack: Target-Grounded Vision-Language-Action Learning for Embodied Person TrackingShengjun Deng, Yinghan Zhang, Chengrui Wei, Liu Liang +7
Language-guided embodied person tracking requires a mobile agent to identify a target from a referring description and consistently follow the target under egocentric observations. We present DobotTrack, a target-grounded vision-language-action (VLA) framework that integrates auxiliary grounding supervision with efficient temporal modeling. We systematically investigate five auxiliary tasks---person segmentation, bounding-box grounding, visual-spatial classification (VSC), future DINO feature prediction, and future-frame prediction---and adopt complementary mask and bounding-box supervision to enhance target delineation and localization. We introduce Q-identity, a target-conditioned historical visual compression module that leverages target-grounding predictions to selectively preserve identity-discriminative appearance features and spatial cues in compact historical tokens. On EVT-Bench, our 0.8B-parameter model achieves an average success rate of 77.9%, exceeding OmTrackVLA by 16.9 percentage points across the three tasks. We further construct a real-world person-tracking dataset collected using a quadruped robot, comprising 126 episodes and 596,323 mask-annotated frames. The target masks enable automatic derivation of bounding boxes and visual-spatial labels encoding horizontal position and apparent image scale, providing additional grounding supervision without separate manual annotation of these labels. Together, our framework and annotation pipeline support learning target-grounded tracking policies from simulated and real-world data. Code is available at https://github.com/Dobot-Team/dobot_track.
vision-language-actionembodiedquadrupedagent - arxiv:2611.09036 · cs.ROToolDex: An Ego-Exo Human Demonstration Dataset for Creative Physical Tool UseMohamed Elmoghany, Kehan Li, Ruoqu Chen, Tarun Chiruvolu +10
Creative tool use is a fundamental human capability and an important step toward general-purpose embodied intelligence. It requires agents to reason about object affordances, physical constraints, and manipulation strategies, especially when standard tools are unavailable. However, progress in this area is limited by the lack of real-world datasets that capture human tool-use behavior with sufficient scale, fidelity, and diversity. Existing benchmarks are often simulation-based, focus on simplified interactions, or lack detailed observations and annotations. We introduce ToolDex, a large-scale egocentric-exocentric (ego-exo) human demonstration dataset for creative physical tool use. ToolDex captures scenarios in which humans select, adapt, and repurpose available objects as tools. Using synchronized egocentric cameras and eight exocentric cameras, the dataset provides high-fidelity multi-view observations and rich annotations for hands, tools, target objects, and their deformable interactions. Beyond data collection, we establish a benchmark suite characterizing three core capabilities: physical tool-use reasoning, sparse-view 3D and 4D reconstruction, and robot policy learning. ToolDex provides a foundation for studying human tool-use behavior and developing embodied agents that can adaptively use tools in complex real-world environments. Dataset visualizations and benchmarking results are available at https://tooluse-dataset.github.io/.
embodiedmanipulationrobot policyembodied agenttool usetool-use - arxiv:2611.09030 · cs.ROA Controlled Study of Models, Skills, Feedback, and Curriculum in Agentic RoboticsKaiyuan Chen, Xinyue Zhu, Shuangyu Xie, Uksang Yoo +6
Agentic robotics asks coding agents to implement and iteratively improve robot policies from simulated and real-world physical interactions. We study the self-learning loop of agentic robotics through a controlled simulation study spanning nine manipulation tasks, 631 different experiment configurations, and 28K coding agent rollouts, systematically varying model capability, skill priors, feedback structure, and curriculum. The experiment data suggests that the bottleneck is often not the access to sufficient information, but interpreting it into the right policy update. Across these factors, a successful self-learning loop depends on how effectively agents attribute failures, reason task-relevant spatial and geometric structure, and selectively restructure code to improve policies.
manipulationagentagentic - arxiv:2611.09028 · cs.LGScout Before You Route: Predicting Task Success and Token Count from a Small LLM PrefillAristides Milios, Christopher Pal
Routers typically choose a model per query by trading predicted correctness against cost. Most price each model at a fixed cost, as if every query used the same number of output tokens, but the output length of reasoning models varies by an order of magnitude across queries. When is per-query cost worth predicting? What factors drive cost? How cheaply can predictions be made? We predict both the success probability and the output length of candidate models from the prefill activations of a single small frozen LLM, the scout. With reasoning candidates, perfect per-query cost knowledge would save 29-61% of spending on eight of nine datasets, and much less with non-reasoning candidates on the same problems. On those eight datasets, our predicted costs save 9-33% at matched accuracy relative to median-length pricing, up to 69% of what perfect knowledge would save. They are significantly cheaper than each estimator we reimplemented or adapted from published work, including ZeroRouter and a MixLLM-style predictor, on a majority of datasets. Output length depends mostly on the problem. On most math and code datasets, harder problems are longer, so predicted success alone prices cost nearly as well. On knowledge and mixed-format datasets they are not, and predicted success falls well behind a probe trained on length. The probe keeps most of its saving with 20 labelled problems and transfers without labels to four of seven new benchmarks. On a mixture of all nine benchmarks, it saves 22% beyond pricing each benchmark at its own median length. Savings stop growing at about a 2B scout, so above 4B a larger scout costs more than it adds.
benchmark - arxiv:2611.09027 · cs.CLLanguage, Not Code Abstraction: Natural-language Training Shapes Brain-aligned GeometryIñigo Parra
Code-trained language models produce positive brain scores, which raises the question of whether abstract structure learned from programming languages can substitute for natural-language training. We trained 25 GPT-2-small models on TinyStories, Python, C functions, x86-64 assembly, and the corresponding machine code, and scored them using a shared tokenizer and encoding procedure against matched random-weight models. On Pereira-243 and Tuckute2024, TinyStories exceeds the mean of the four code models by $Δr = 0.053$ and $0.059$ (Holm-corrected $p<0.001$), while code training leaves alignment near the random-weight level. The language advantage is benchmark-specific: Blank2014 shows no corrected language-code difference, and Fedorenko2016 shows no language benefit. Geometry tracks the encoding result on Pereira-243: TinyStories seeds converge on the most similar representations, and its fit-free correspondence with neural dissimilarity peaks at $ρ= 0.14$, twice the code and random-weight levels. Binary-trained models have about three effective dimensions, so seed agreement alone does not imply neural correspondence. Updating only the embedding matrix on English improves mean-over-depth alignment for every code model more than an equal update on its own corpus. Within this architecture and training setup, natural-language training produces a distinct representational geometry with closer neural correspondence on Pereira-243
benchmark - arxiv:2611.09017 · cs.CVFrontier3DBench: Benchmarking Coding Agents for 3D Creation across Objects, Scenes, and Interactive Open WorldsXinjie Lin, Qingle Liu, Wenyi Li, John Lam +9
Coding agents combine visual understanding with program execution, offering a new route to automated 3D creation. By translating images and instructions into editable programs, they bring reconstruction, generation, and editing into a shared workflow. Yet existing benchmarks fragment evaluation across task scopes and protocols, leaving their capabilities across construction levels unclear. We propose Frontier3DBench, a unified benchmark of 145 tasks across 9 categories spanning object, scene, and interactive open-world creation, with joint evaluation of appearance, structure, and behavior. Experiments with 11 model-harness configurations reveal strengths that vary by level: GPT-6.1-sol leads the object and scene levels and achieves the highest overall score of 66.02/100, whereas GPT-6-astra leads interactive world construction. However, all configurations score lower on reference reconstruction than on text-driven generation in object and scene tasks. Producing plausible content thus remains easier than reproducing specified geometry, layouts, and visual details. Moreover, generated worlds average 83.08/100 on local mechanisms but only 20.82 on motion clearance and 30.26 on stateful workflows: individual mechanisms often function, whereas complete activities that require coordinated actions and persistent state remain difficult. These results identify reference fidelity and coordinated activity completion as central bottlenecks for coding agents in 3D creation.
persistent statebenchmark - arxiv:2611.09015 · cs.AISIMBAD: Spatio-Temporal Traffic Forecasting Robust to AperiodicityDaniel Y. Lee, Seungwon Shin, Seunghoon Han, Sungsu Lim +1
Accurate traffic prediction is essential for urban planning but remains challenging due to the irregular patterns caused by unexpected events. While Spatio-Temporal Graph Neural Networks (STGNNs) effectively model periodic data, they often struggle with these aperiodic fluctuations. To address this, we propose SIMBAD, which improves generalization irregular patterns via two key mechanisms: 1) adaptive regulation of periodic signals to prioritize recent signals, and 2) dynamic spatial modeling that adjusts influence between connected nodes based on contextual relations. Built upon simple similarity metrics to adapt to aperiodic patterns, SIMBAD outperforms existing methods on real-world benchmark datasets, demonstrating superior performance in predicting unseen and irregular traffic events.
benchmark - arxiv:2611.09013 · cs.LGKnowEvo: Knowledge Evolution for Protein OptimizationZijie Xing, Xingyue Liu, Runze Wang, Luoming Hu +1
Large language models (LLMs) have shown promise in scientific discovery, yet systematically accumulating and refining reusable scientific knowledge remains challenging. Existing approaches primarily improve LLM-based problem solving through parameter optimization or feedback-driven refinement of prompts, programs, and solution trajectories, leaving acquired expertise either embedded in model weights or scattered across iterative problem-solving processes. As a result,knowledge is difficult to systematically inspect, evaluate, and refine. We propose KnowEvo, a framework that treats externalized knowledge itself as the optimization target. Our key insight is that scientific knowledge is inherently modular: multiple interacting mechanisms jointly determine the design outcomes. This makes knowledge evolution fundamentally challenging, as improvements cannot be attributed to individual components in isolation. To address this, KnowEvo represents knowledge as executable expertise blocks and organizes their refinements into strategy mutation trees, enabling structured exploration over a combinatorial knowledge space. The framework combines a Thompson-style forward allocation strategy that adaptively distributes evaluation budgets across strategy variants with a backward revision process, in which a code agent operates in an isolated sandbox to translate rollout evidence into localized program updates. This design enables persistent accumulation, evaluation, and refinement of mechanistic expertise across iterative experimentation. We evaluate KnowEvo on protein optimization tasks including stability and solubility. Results show that knowledge evolution consistently improves over initial strategy blocks and outperforms monolithic program evolution and prior LLM-guided evolution baselines across metrics such as pass@k, MRR, MAP,and NDCG@5. Code is available at https://github.com/recynie/KnowEvo.
agent - arxiv:2611.09004 · cs.MASharing the Road: LLM-Based Bargaining in Mixed-Autonomy TrafficLingxiao Zhou, Di Chen, Jia Li
Mixed-autonomy traffic involves heterogeneous driving populations competing for shared road space. Previous studies have shown the existence and emergence of collective rationality using game-theoretic and reinforcement learning approaches, which rely on explicitly specified behavioral assumptions or reward functions for driving agents. However, how bargaining behaviors between driving agents evolve and lead to collective outcomes based on richer agent characteristics and language-based reasoning remains less understood. To fill this gap, we develop a bargaining framework based on large language models (LLMs) in which population-level autonomous vehicle (AV) and human-driven vehicle (HV) agents negotiate road-space allocations under heterogeneous objectives and private information. Through controlled experiments varying traffic composition, agent objectives, and traffic-domain knowledge, we examine bargaining behavior, coordination, and resource utilization. Results show that negotiated allocations respond to relative traffic density without strictly following proportional sharing, while density-balanced states exhibit greater coordination difficulty. Fairness-oriented objectives substantially improve agreement rates but do not necessarily yield equally high road-space utilization. Furthermore, domain knowledge affects bargaining outcomes differently across objective configurations, shaping both coordination and the role-dependent effects of fairness. These findings demonstrate the potential of LLM-based bargaining for modeling and designing interactions among driving agents in mixed-autonomy traffic.
agent - arxiv:2611.08998 · cs.CVSeeing vs. Counting: Complexity Twins Show Where Multimodal LLMs Fail on DiagramsYin Liu
When a multimodal LLM miscounts in a diagram, does it fail at seeing or at counting? Existing visual graph and puzzle benchmarks vary visual density and algorithmic hardness together, confounding the two. We pair counting questions about the same drawing that sit in different counting-complexity classes, such as its spanning trees (a determinant) and its simple paths (#P-complete), and use a transcription probe and a text-only control to separate reading from computing. We test Gemini, Gemma and Qwen3-VL models; for the Gemini models, the answer depends on how the model computes. In chain of thought, most of them read small drawings correctly and then fail at counting, at a rate set far more by the size of the answer than by its complexity class (for the primary model, the odds of an exact answer fall 2.38-fold per decade of the answer). Given code and an exact transcription, the primary model counts correctly even on 10x10 grids, and reading becomes the bottleneck: a single view misses horizontal edges, and a second reading of a rotated copy with crop checks reads 86/96 large grids exactly, against 57/96 when the upright reading is simply repeated.
benchmark - arxiv:2611.08996 · cs.CVUniMore: Fine-Grained Supervision for Unified Understanding and GenerationHaoyue Bai, Yinyi Luo, Yicheng He, Qingsong Wen +2
Unified multimodal models combine image understanding and generation within one system, creating an opportunity to improve both capabilities through paired image--text supervision. We examine the outputs of a pretrained unified model in both directions and observe a loss of specificity: generated images can contain the requested objects with incorrect attributes, while captions identify a scene without expressing its distinguishing details. These failures call for supervision that specifies both the visual content to realize and the descriptive differences to distinguish. We propose UniMore, a post-training method that derives these complementary learning tasks from the fine-grained content of image--text pairs. The generation route predicts grounded crops from their corresponding phrases, matching the scope of the condition and visual target while using the native visual objective. The understanding route combines positive-caption fitting with ranking against captions that omit or alter details, conditioned on the full image. The two routes are optimized independently, and their learned updates are combined into a single unified model without changing its inference procedure. Experiments show gains across multiple understanding and generation benchmarks.
post-trainingbenchmark - arxiv:2611.08990 · cs.LGAction-Centered Causal Bandits for Budget-Constrained Promotion TargetingYifan Lin, Kaixu Yang, Di Mo
Web platforms select among costly interventions for each user to maximize incremental outcomes under a global budget. We formulate this problem as a budget-constrained contextual bandit with heterogeneous treatment effects. Although an action-independent outcome baseline cancels from treatment comparisons, modeling a nonlinear baseline can increase statistical complexity. We propose DR-CATE-CBwK, an action-centered causal bandit that directly learns treatment contrasts. Its online update remains valid under baseline-model misspecification without inverse-propensity weighting. It combines an offline doubly robust learner to warm-start treatment-effect estimates, an online action-centered estimator for sequential contrast learning, and a primal-dual controller with an affordable-action rule to enforce the effective-cost budget pathwise. For linear conditional average treatment effects (CATEs) in a stationary i.i.d. setting, we prove a high-probability (\widetilde{O}(Kd\sqrt{T})) regret bound, where K is the number of priced actions, d the context dimension, and T the horizon. We benchmark against the fluid linear program (LP) under the coverage-guaranteed action rule. We further show that the action-centered estimating moment remains unbiased under any predictable baseline estimate, whereas a specified separate-arm reward regression can exhibit asymptotic contrast bias under action-dependent logging and shared-baseline misspecification. On a semi-synthetic job-marketplace benchmark with real logged features and simulated outcomes, our practical Thompson-restricted variant achieves the lowest regret among evaluated budget-feasible methods in the operating regime. It reduces regret by approximately 20% versus the strongest reward-based baseline, outperforming raw-reward contextual bandits, BwK baselines, and a static offline policy, while never exceeding the effective-cost budget.
benchmark - arxiv:2611.08989 · cs.CVLEF-JEPA: Denoising-Trained Latent Energy Fields for Spatial Representation LearningZhiyuan Liu, Zhihao Xu, Sicheng Liu, Huanhuan Huang
A joint-embedding predictive architecture (JEPA) that applies prediction supervision only at masked positions struggles to clearly capture object shapes and boundaries in its latent representations. We further observe that some existing methods improve spatial structure while also retaining rich variations in texture and color. However, preserving more low-level visual information does not necessarily make representations better suited to prediction and action. This raises a key question: how can we strengthen the spatial structure in latent representations that supports prediction and action without over-encoding these details? We introduce LEF-JEPA, which learns spatial structure through target-guided latent recovery and transfers this supervision to the visual encoder through a relative-energy constraint; the energy module is used only during pretraining and adds no inference overhead. To reduce the cost of energy optimization during pretraining, we further construct a conditional Ising Hamiltonian that allows low-energy-state search to be performed by specialized Ising solvers. On physical hardware, we validate the feasibility of achieving lower forward latency with fewer local updates. Under a unified CIFAR-10 pretraining and frozen-encoder evaluation protocol, both realizations outperform V-JEPA2.1 in image classification, foreground segmentation, and visual planning.
v-jepaevaluation protocol - arxiv:2611.08987 · cs.ROVLA Handover: Failure-Prediction-Based Switching from On-Robot to Server VLA ModelsTakayuki Nishio
Large vision-language-action (VLA) models generalize well but are difficult to run in real time on a robot, and small VLAs that run there fail more often. Remote inference over beyond-5G and 6G networks lets a robot use a large VLA on a server, but querying the server for every action chunk limits the number of robots per server. We propose VLA handover: the robot runs a small VLA (S-VLA) and switches to a large VLA (L-VLA) on a server only when the S-VLA is likely to fail. We formulate the switching as minimizing the request cost under a required success probability and, since this problem cannot be solved directly, propose a heuristic framework that retrains neither VLA. In simulation with SmolVLA and $π_{0.5}$ on three of the four LIBERO suites, VLA handover raised the success rate by 11-22 percentage points over the S-VLA alone and sent 49-77% fewer requests than server-only control at a 2-4 percentage-point lower success rate, and its completion time grew less with the network latency.
vision-language-actionvlavla modellibero - arxiv:2611.08977 · cs.LGEnergy-Based Reasoners: Your Recurrent Reasoner Should Be an Energy-Based ModelWilliam Xingxu Chen, Alexi Gladstone, Zeyu Michael Li, Heng Ji +1
Looping is a powerful inductive bias for reasoning. By repeatedly iterating a recurrent block, looped models can scale their effective depth and enable multi-step reasoning in latent space. However, existing looped models are not inherently guided by an explicit measure of progress, making it difficult to determine when to stop looping and how to improve upon the current state. To enable this, we introduce Energy-Based Reasoners (EBRs), which learn an energy function over latent reasoning states and formulate looped updates as gradient descent on this energy. The energy scores latent states, while its gradient guides refinement and provides a signal for stopping. To train EBRs effectively, we further introduce exploration-based training to discover additional modes and to better shape the energy landscape over a broader set of states. Across four algorithmic reasoning benchmarks, EBRs match or outperform prior recurrent reasoning models, achieving 99.9% accuracy on Sudoku-Extreme while using 345$\times$ less inference compute than the next-best model and maintaining comparable training compute. On all multi-solution benchmarks, EBRs achieve state-of-the-art coverage of distinct valid solutions while maintaining near-100% single-sample accuracy. These results suggest that energy-based models are a promising approach for strong reasoning performance.
benchmark - arxiv:2611.08976 · cs.CVRefer-CityFlow and M3Track: A Benchmark Dataset and Reference Framework for Language-Driven Multi-Target Multi-Camera TrackingPhuong T. Nguyen, Hao Vo, Tam Bang, Hoang K. Tran +7
Referring multi-object tracking enables users to specify targets through natural language, but existing Refer-MOT benchmarks primarily focus on single-camera settings. We formalize Referring Multi-Target Multi-Camera Tracking (Refer-MTMCT), where free-form queries specify identity-consistent trajectories to be recovered across a network of sequential, sparse-overlap cameras. We introduce Refer-CityFlow, a benchmark built on CityFlow's global cross-camera identity annotations and augmented with language queries requiring joint language grounding, cross-camera association, and spatio-temporal reasoning. Refer-CityFlow covers 96 videos from 40 roadside cameras with 660+ identity-consistent cross-camera trajectories, 4,362 trajectory-level descriptions, and 7,699 queries, including 1,699 real-user queries collected from 50 participants. Reference experiments on the four-camera Scenario 02 show that single-camera referring trackers reach 14.95-18.28 HOTA on real-user queries, while the evaluated cross-view method achieves about 0.05 HOTA in direct transfer. To establish reference performance, we provide M3Track, a modular multimodal reference framework in which a large language model (LLM)-based agent orchestrates interchangeable detection, tracking, and cross-camera association modules. M3Track reaches 44.10 HOTA and 65.40 IDF1 on the same real-user query set. The project repository is available at https://github.com/phuongnguyen194/Refer-CityFlow-and--M3Track.
agentbenchmark - arxiv:2611.08971 · cs.AIPhantom Consensus: Measuring Social Reasoning in Multi-Agent LLM Decision-MakingDisha Sheshanarayana, Rajat Subhra Pal, Manjira Sinha, Tirthankar Dasgupta
Multi-agent LLM systems increasingly aggregate judgments across model instances, but reasoning about other agents can itself introduce failure. We study a setting in which LLMs shift their public decisions after observing ambiguous social signals while reporting inaccurate estimates of what other agents privately judged, even when the group's recorded private judgments do not support the apparent consensus. We call this failure phantom consensus and study it as a pluralistic-ignorance-style pattern in controlled LLM collectives. To measure this phenomenon, we introduce PHANTOM-Bench, a benchmark that separates isolated private judgments, public decisions after ambiguous social signals, elicited estimates of the group's private judgment, transparency-based recovery, and controls for neutral public text and explicit vote-following. Across nine closed- and open-weight models, ambiguous public signals consistently degrade decision quality and can turn correct private group judgments into incorrect public collective outcomes. Public accuracy drops by as much as 47.8 percentage points, and collective rationality failures reach 47.3% in the most susceptible setting. We observe substantial errors in reported group-private beliefs alongside public-stage failures, while control conditions reveal model-dependent differences in susceptibility to ambiguous and explicit peer signals. Revealing the agents' recorded private decisions repairs many failures without revealing the gold label, although recovery varies across models. Our results show that evaluating multi-agent LLM systems requires measuring not only final agreement, but whether that agreement reflects independently measured private judgments or an apparent consensus.
multi-agentbenchmark - arxiv:2611.08969 · cs.AIPhysioWave-T: Token-Efficient Wavelet Foundation Models for BiosignalsYanlong Chen, Simone Benatti, Luca Benini, Cuntai Guan +1
We present PhysioWave-T, a token-efficient wavelet foundation model family for EEG, ECG, and sEMG. Dynamic scale folding recombines J+1 wavelet subbands prior to tokenization, compressing sequence length from (J+1)CS to CS tokens for J levels, C channels, and S temporal patches. Route-specific front ends and learned channel-name embeddings allow one Transformer per modality to handle diverse montages and sampling rates, while unseen setups reuse the backbone via lightweight adapters. We pre-train a 30M-parameter EEG model on over 15M windows (19--128 channels), alongside ECG (30M) and sEMG (12M) encoders. Across four EEG routes, scale folding cuts Transformer latency by 4.0--7.2x and peak memory by 6.8--12.7x over scale expansion without accuracy loss. Our EEG encoder leads 8 of 9 metrics against BENDR, BIOT, LaBraM, and EEGPT, and surpasses LUNA-L on wearable seizure detection with fewer parameters. Reliability-aware latent fusion (RALF) further integrates summary tokens from independently pre-trained encoders via sample-adaptive weights. On SeizeIT2, RALF raises event sensitivity at 11 false alarms/h from 61.4% to 74.4% and window AUPRC from 0.321 to 0.399 over the best unimodal baseline, outperforming weighted-logit and concatenation fusion across all variants. PhysioWave-T also achieves state-of-the-art sEMG accuracies and top macro-F1 on PTB-XL Super and Chapman-Shaoxing. Code and models: https://github.com/ForeverBlue816/PhysioWave.
memory - arxiv:2611.08950 · cs.MAAIM: Adaptive Interaction in Multi-Agent Debate for Multimodal LLM InferenceWei Fan, JinYi Yoon, Bo Ji
Multi-Agent Debate (MAD) improves reasoning on multimodal tasks, but its accuracy and token efficiency depend on how agents interact during debate. A largely overlooked design choice is which interaction modalities agents use to exchange evidence. Existing methods rely on fixed text- or graph-based interaction, which fails to convey fine-grained multimodal evidence and thus degrades accuracy while inflating token cost due to verbose descriptions. To address this, we propose Adaptive Interaction in Multi-Agent Debate (AIM), an adaptive routing framework that dynamically selects the optimal modality combination for inter-agent interaction in multimodal MAD. Beyond text and graph, AIM introduces task-dependent modality interaction, where agents exchange targeted regions of interest from the task-specific multimodal input, such as image regions, audio segments, or spatio-temporal video clips. To determine the best combination for each instance, AIM first generates a structured baseline response and extracts interpretable routing features that capture which evidence the instance requires, how confident the baseline agent is, and how much new evidence each modality would add beyond the baseline response. A lightweight router then uses these routing features to select a modality combination that avoids both under-fusion (i.e., agents miss decisive evidence) and over-fusion (i.e., redundant modalities inflate token cost) risks. Across eight benchmarks spanning three multimodal task families, \name achieves the highest accuracy. Combining modalities improves accuracy by up to 6.4\% over the best single-modality baseline, and adapting the combination per instance adds up to 4.2\% over the most accurate fixed combination of each benchmark with up to 53.4\% fewer tokens.
agentmulti-agentbenchmark - arxiv:2611.08947 · cs.AIMore with Less: Efficient Parallel Scientific Code Translation via an Agent-Skill WorkflowJianuo Shen, Pengyi Xu, Bo Zhang
Translating parallel scientific codes between HPC programming frameworks is a critical but labor-intensive task. Existing LLM-based approaches rely on fixed sequential pipelines with undifferentiated repair strategies and correctness criteria that do not match scientific computing practice. We present an agent-skill framework that separates translation, compilation diagnosis, and execution validation into three autonomous agents coordinated by a verdict-driven orchestrator, routing each failure to a targeted repair skill rather than retrying blindly. An automated output check drives the self-correction loop, complemented by an offline LLM-based verification that exposes false positives escaping compiler and runtime checks. We evaluate the framework on 30 HeCBench kernels translated bidirectionally among CUDA, OpenMP, and SYCL with three open-source LLMs. Our framework attains the higher verified correctness rate in all 18 model and direction combinations over a state-of-the-art baseline while consuming up to 52.4% fewer tokens, remaining consistent across models and directions without architectural modification.
autonomous agentself-correction - arxiv:2611.08946 · cs.ROFlex-RSI: Benchmarking Recursive Self-Improvement of Frontier AgentsJingdong Zhang, Jionghao Wang, Zhengming Yu, Yancheng Cai +18
Frontier language models are increasingly deployed as agents, and how well an agent performs depends not only on its model but also on its harness: the prompts, tools, memory, and control loop through which the model acts. When an agent is applied to a new problem, a natural question is whether it can use the feedback of the task to improve the way it works, and not merely its answer. Most existing benchmarks keep the agent's harness fixed and score the solution it produces, or target large-scale model training. We introduce Flex-RSI, a benchmark for the recursive self-improvement of agents on vision and robotics tasks under a fixed budget. In each round, an executor improves the task solution, which is kept only if it scores higher on held-out validation data, and an improver then revises the harness, including the executor and itself. Every run starts from the same general-purpose harness, and the final solution is scored on hidden test data. Evaluating closed-source and open-weight models on tasks spanning perception, reasoning, robot action, 3D reconstruction, visual creation, and infrastructure, we find that self-improvement helps most when the agent builds its own solution, that feedback recording only success or failure holds back open-weight models, and that open-weight models fall furthest behind where outputs must pass strict checks of physical stability or exact correctness. Paired runs further show that editing the harness itself contributes to the gains.
agentself-improvementbenchmark - arxiv:2611.08943 · cs.CVMESAM: Enhancing SAM2 for Few-Shot Segmentation through Semantic-Visual Memory Generation and RefinementTianyu Zou, Shengwu Xiong, Rui Wu, Ruilin Yao +1
This paper investigates how to leverage the powerful capabilities of the Segment Anything Model 2 (SAM2) to tackle few-shot segmentation (FSS) tasks. One major challenge lies in the substantial appearance differences between the support and query images, which may diminish the effectiveness of the memory-matching process in SAM2. Through systematic analysis, we find that the memory generated based on the categorical semantic information is robust to this appearance divergence and is more compatible with SAM2. Based on these observations, we propose a new Memory-Enhanced SAM2 (MESAM) method, which first introduces a Semantic-Visual Memory Generation (SVMG) module to separately construct a query semantic memory, a query foreground visual memory, and a support adaptive visual memory. Then, in our designed Interactive Semantic Memory Refinement (ISMR) module, we further enrich the obtained semantic memory with fine-grained visual details, and suppress its misatcivated background responses. Extensive experiments on the PASCAL-$5^i$, COCO-$20^i$, and LVIS-$92^i$ datasets demonstrate that MESAM consistently outperforms existing FSS methods, achieving new state-of-the-art segmentation performance. Our source code is available at: \href{https://github.com/tianyu-zou/MESAM}{https://github.com/tianyu-zou/MESAM}.
memorysemantic memory - arxiv:2611.08939 · cs.RORevoSim: Scalable Multimodal Tactile Simulation for Dexterous ManipulationKe Zhang, Lixin Xu, Ziyi Wang, Xiyue Dong +5
We present RevoSim, a scalable and high-fidelity multimodal tactile simulator for the BrainCo Revo3 hand that jointly models 285 piezoresistive taxels and five visuotactile fingertips. RevoSim efficiently generates synchronized pressure, depth, RGB, and marker-displacement observations; on a single RTX 5090, its depth-conditioned marker-displacement configuration supports up to 4,096 parallel environments. We train in-hand manipulation and grasping policies entirely in simulation, establishing a foundation for future sim-to-real transfer. Project page: https://braincotech.github.io/RevoSim-web/.
manipulationdexteroustactilesim-to-realgrasp - arxiv:2611.08936 · cs.ROOrdered Policy OptimizationDongkyu Derek Cho, Pan Xu
Proximal Policy Optimization (PPO) stabilizes policy updates by clipping the importance ratio between the new policy and the data-collecting policy. The standard rule uses one fixed clipping range for all samples. While this choice offers simplicity and effectiveness, it fails to distinguish statistically harmful extreme ratio values from large, yet tolerable changes. We study how to adapt the clipping radius to the reliability of the induced importance ratio distribution. Our case study demonstrates that reliable updates require controlling these extreme ratios, rather than merely verifying that individual ratios fall within a fixed interval. Motivated by this observation, we propose Ordered Policy Optimization (\textsc{OPO}). In each minibatch, \textsc{OPO} sorts the importance ratios, selects the extreme ratios that potentially have the largest effect on the update, and assigns them adaptive sample-wise radii instead of using the fixed clipping radius. These radii are derived from a prescribed tail profile, which specifies desired target values for the selected extreme ratios. The resulting objective remains first-order and requires only minibatch sorting and a modified policy loss. Experiments conducted on MuJoCo and Atari benchmarks demonstrate that this simple modification achieves strong performance relative to established baselines.
benchmark - arxiv:2611.08933 · cs.AIDRAFT-Bench: Evaluating Frontier Agents Across the Professional Design LifecycleRong Pan, Jianqiang Wang
Can frontier general-purpose agents understand client intent, explore creative directions, execute revisions, and prepare production deliverables? We introduce DRAFT-Bench (Design Reliability Across Feedback and Tasks), a lifecycle-oriented evaluation of requirement elicitation, creative exploration, constrained editing, iterative revision, and production handoff. Its Requirement-to-Artifact framework links evolving requirements, protected properties, artifact versions, and delivery obligations. We compare native Codex/Astra and Claude Code across 35 design briefs under 43 conditions, alongside a separate 40-scenario RFQ track. Astra completes 43/43 design conditions and Claude 35/43. All scored core runs receive full structural credit, yet on the 28 core conditions scored for both systems, semantic checks flag violations in 11 for Astra and 5 for Claude; these are model judgments, not expert-confirmed failure rates. RFQ coverage is high, but success ordering changes across cohorts. SKU expansion, revision dependencies, and White ink files expose ambiguity among unmet obligations, missing inputs, and evaluator interpretations. These findings separate execution, requirement compliance, and professional acceptance. Unpaired tracks and different tools preclude claims of RFQ-to-delivery success, factory acceptance, or a pure model ranking.
evaluator - arxiv:2611.08932 · cs.MAODD+LLM: Specifying LLM-Agent Behavior for Reproducible Social SimulationJustin Downes, Robby Milletich, Newel Hirst
The ODD protocol is the de facto standard for describing agent-based models, but it lacks vocabulary for the components that determine LLM-agent behavior: prompt assembly, memory serialization, parsing, and retry logic. We propose ODD+LLM, a backward-compatible extension adding five documentation blocks grounded in the CoALA cognitive-architecture framework. To show that the gap matters, we conduct a multiverse analysis [Steegen et al., 2016] of a published LLM naming-game simulation, varying six specification choices a reimplementer must make. Variants that are defensible readings of the paper stay within statistical noise of our baseline reconstruction. Missing one instruction that the paper states but is easy to overlook (that the model answers as an outside observer rather than as a player) raises the share of runs settling on the convention agents individually favor from 0.35 to 1.00 and makes convergence three times faster, in a reimplementation that uses a different model from the original. We complement the demonstration with a reconstruction-sufficiency review of 28 published LLM-agent simulations, finding that the details of how prompts are built and how model outputs are parsed are undetermined in 77% of cases, and that no paper fully specifies its main LLM call. ODD+LLM gives the field a standard home for information it already produces but currently scatters across code repositories, supplementary files, and undocumented defaults.
memory - arxiv:2611.08929 · cs.LGGrid-Oriented Probabilistic Load Forecasting for AI Data CentersQingyi Yao, Ding Lin, Tianqiao Zhao, Meng Yue +1
Probabilistic load forecasts for AI data centers need to represent both uncertainty under changing workloads and temporal dependence to support grid operations. This paper develops an open, grid-oriented benchmark using measured GPU power from a 448-GPU AI cluster. Forecasts are evaluated through four stages: probabilistic prediction, online uncertainty calibration, temporal scenario generation, and horizon-wide reserve sizing. Following a change in cluster utilization, a quantile LightGBM forecast achieves only 63.4% coverage for a nominal 90% interval, while online regime-conditioned calibration increases coverage to 87.3%. Even with well-calibrated marginal forecasts, independent sampling produces unrealistic ramp distributions and excessive reserve procurement. Reconstructing temporal dependence from historical forecast errors reduces the discrepancy in ramp statistics and lowers realized reserve cost by 7.4-10.6% across three forecasting tasks. These findings show that marginal forecast quality alone is insufficient for operational decisions and motivate joint evaluation of calibration, temporal structure, and decision outcomes.
benchmark - arxiv:2611.08924 · cs.LGResQAdam: Residual-Preserving Adaptive Optimization for Memory-Efficient Quantization Aware TrainingLiqun Ma, Shengkun Tang, Liangyu Wang, Lin Zhao +6
Quantization-aware training (QAT) is the most reliable way to obtain accurate low-bit large language models (LLMs), but training with AdamW keeps two dense FP32 states for every weight matrix. We find that low-rank memory-efficient optimizers such as GaLore and LDAdam degrade substantially in QAT. A considerable part of the QAT momentum lies outside its top-r singular subspace, and projection-based methods discard this part at every step. We further observe that this residual remains useful when only its sign is kept. Based on these observations, we propose ResQAdam. It keeps a single BF16 momentum, applies Adam only to the r x r coordinates of the principal subspace, and adds a normalized sign update computed from the residual. We evaluate ResQAdam with two from-scratch QAT methods and three QAT fine-tuning pipelines, with weights quantized down to 2 bits. It reduces optimizer-state memory by 71% to 74% relative to AdamW, stays close to AdamW in most settings, and lowers training throughput by less than 2%.
memory - arxiv:2611.08922 · cs.AIGlyph: An Interpretability Engine for AI-Driven Scientific Discovery in Clinical DataRobbie Holland, Ashwin Kumar, Mateus A. Esmeraldo, David Le +18
Automated scientific discovery has the potential to advance understanding of disease mechanisms and accelerate identification of early disease markers. Manual data annotation is infeasible at population scale, motivating methods that automatically extract pre-defined phenotypes in multimodal and unstructured clinical data. Such pre-defined feature sets determine and limit which associations these approaches can discover. This limitation extends to agentic AI systems that use these tools to query population-scale data. In contrast, foundation models (FMs) trained on large volumes of clinical data can perform tasks beyond their training objectives. This suggests their latent embeddings may represent novel or unexplored features that could yield insights into disease mechanisms, yet such latent features remain difficult to interpret. To address this challenge, we introduce Glyph, an automated multistage framework that extracts features from multimodal and unstructured clinical data using FMs and sparse autoencoders, interprets them as coded clinical concepts, and evaluates their value for clinical scientific discovery. We applied Glyph to abdominal CT volumes, CT radiology reports and laboratory measurements from 41,628 patients across different institutions. We identified 397 predictive concepts supported by 4,463 Bonferroni-significant associations. Moreover, a majority of discovered associations were automatically reproducible in held-out cohorts across different sites. Glyph concepts provided complementary associations to literature-derived markers, leading to gains in predictive power. We envision that interpretability can extend beyond a mechanistic understanding of neural networks to connecting implicitly learned features with clinical knowledge and establishing their significance for patient outcomes. We release Glyph at https://github.com/StanfordMIMI/Glyph.
agentic - arxiv:2611.08921 · cs.AIFrame: Form-Restricted Authoring with Model-Based EvaluationNan Jiang, Mohammed Mohsin Hussain, Onkar Tiwari, Ankit Jain +3
Many professional reports reuse the structure of prior editions while updating the content. We formalize this setting as form-restricted report generation: given a sample report and new grounding documents, generate a report that preserves the sample's form while accurately reflecting the new evidence. We introduce FRAME, a benchmark of 120 EDGAR- and arXiv-derived instances across Financial Quarterly, Financial Annual, Academic Simple, and Academic Complex categories, built with a temporal three-snapshot pipeline. We evaluate outputs with three human-validated metrics: structural similarity via semantic tree-edit distance, coverage via information unit recall, and faithfulness via claim-level grounding verification. Across five LLMs, open-source models nearly match proprietary models on structure but lag on coverage and faithfulness; all models perform better on financial than academic reports, and performance declines with longer outputs and more complex grounding.
benchmark - arxiv:2611.08920 · cs.ROHOC2R: Hand-Object Contact Trajectories for Human-to-Robot Motion TransferHongkun Yu, Francisco M. López, Francisco Cruz
Human videos contain abundant object-manipulation demonstrations, but they do not directly provide motion cues that a robot can follow. We present HOC2R, a Hand-Object Contact trajectory extractor for Robot motion that combines frozen DINOv3 and SAM 2.1 features with learned heads for hand-object association and contact localization. HOC2R supports two modes: count-aware automation that detects, counts, and localizes visible contacts in untrimmed videos, and fixed-two localization for two-finger grippers when contact frames are known. On a 30-task point-annotated benchmark, count-aware automation improves contact-count accuracy from 58.1% to 71.4% and reduces mean localization distance from 19.81 to 10.46 pixels relative to an adapted VRB2 baseline. On seven unseen RGB-D rollouts, a five-task fixed-two checkpoint reduces mean localization distance from 67.67 to 27.02 pixels. Across six robot evaluation sequences, the 30-task fixed-two checkpoint produces four feasible trajectories and succeeds on three tasks, compared with two feasible trajectories and one successful task for each adapted baseline. HOC2R is competitive in the evaluated single-hand scenes, robust to the hand-color shift, and strongest in the evaluated multi-hand scenes. These results support HOC2R as a practical pseudo-labeling method for extracting robot-usable contact trajectories from human demonstrations.
manipulationgripperbenchmark - arxiv:2611.08918 · cs.LGComposable Causality: A Toolkit for Systematic Time-Series Causal Discovery and Treatment-Effect BenchmarkingMuhammad Hasan Ferdous, Md Osman Gani
Real-world multivariate time series rarely come with verified ground-truth causal structure, so research on time-series causal discovery and treatment-effect estimation depends on synthetic data. Existing tools are either fixed datasets that cannot be modified after release or script-based generators in which each script fixes one combination of mechanism, confounding, and missingness. Neither supports controlled experiments that vary a single property, such as sparsity, noise distribution, or missingness mechanism, while holding the others fixed. Counterfactual evaluation has the same gap, since existing simulators are built per paper with hardcoded mechanisms or produce observational data only. We introduce DTCD (Datasets and Toolkit for Causal Discovery), a data generation library that produces multivariate time series with ground-truth contemporaneous and lagged adjacency matrices. Topology, mechanism, noise, observational mask, and intervention policy are independent components that compose through configuration alone. Paired counterfactual twins, produced by caching the stochastic realization during the factual rollout and re-injecting it during the interventional rollout, recover individual treatment effects to numerical precision. A reference benchmark of eight discovery algorithms on eighty-one configurations and four treatment-effect estimators on eighteen configurations exposes failure modes that aggregate metrics hide. These include the inability of standard discovery methods to handle missing values, regardless of whether the missingness mechanism is informative, and a marginal-variance shortcut (varsortability) that we measure and remove by standardization. Code and data: https://github.com/hferdous/Composable_Causality
benchmark - arxiv:2611.08917 · cs.ROPRISM: Primitive Routing via In-context Skill Mixing for Retentive VLA AdaptationJiachen Tao, Gengyu Zhang, Haoxuan Wang, Bing Liu +1
Vision-Language-Action (VLA) models promise a single language-conditioned policy adaptable to many manipulation behaviors, yet deployment additionally requires skill retention and long-horizon composition: a policy must retain previously acquired skills and reuse them inside composite tasks whose behavior changes across phases. The standard recipe absorbs every new task into one shared update, which provides neither a parameter-level locus for reusable skills nor a mechanism for invoking each skill along the trajectory. We observe that this missing structure is latent rather than absent: per-skill low-rank updates of a pretrained VLA occupy a robustly distinguishable, low-dimensional subspace, and a single composite-task LoRA concentrates roughly half of its Frobenius energy in this subspace, against a near-zero matched random-rotation control. Building on this geometry, we introduce PRISM, which represents reusable skills as parameter-level primitives: each skill family is stored as a dedicated low-rank action-expert adapter in a primitive bank, primitives are augmented with skill-specific low-rank residuals and composed at inference by a per-chunk VLM-conditioned router that reads the prefix representation to track task progress, and the router is supervised by per-chunk labels from a deterministic, task-conditioned, action-derived oracle. During composite training, the vision encoder, the language LoRA, the skill residuals, the router, and the shared action projections and time MLP are updated; every primitive is held frozen, so the base per-skill action-expert adapters are preserved by construction. On RoboCasa365, PRISM achieves 56.43% atomic-skill success and reaches 20.00% on long-horizon composite tasks, while a standard $π_{0.5}$-LoRA two-stage fine-tune reaches 17.86% atomic and 17.50% composite success. We additionally validate PRISM on an AgileX PIPER real-robot setup.
vision-language-actionvlamanipulation - arxiv:2611.08916 · cs.LGOn the Incompleteness of do-calculus for Universal IdentifiabilityYaroslav Kivva, Aliaksandra Novik, Negar Kiyavash
We study the problem of causal effect identification from an arbitrary set of observational and/or interventional distributions, which we call Universal Identifiability (uID). The uID problem generalizes existing classical identifiability (ID) and general identifiability (gID) problems, which considered the identification problem from either mere observational distributions or a specific type of interventional distributions. Prior research has established graphical criteria to determine when a causal effect is identifiable in ID and gID problems. Moreover, algorithms that leverage do-calculus and basic probability manipulations have been proposed to compute these effects. Consequently, it has been shown that do-calculus is both sound and complete in these problems. Despite these well-established results, a formal definition of what it means for the do-calculus to be complete does not exist. We propose a rigorous formalization for the completeness of the do-calculus and show that the completeness results in previous work remain valid. However, for more general problems, such as uID, the rules of do-calculus are no longer complete. This is perhaps the reason why the uID problem has remained open thus far.
manipulation - arxiv:2611.08914 · cs.ROHomeLoop: A Simulation Framework and Benchmark for Robot-IoT Coordination in Everyday Domestic Assistance and SafetyRamy ElMallah, Marc Bishara, Hyegang Kim, Yuxi Zhang +2
Recent progress in embodied robotics is bringing robots closer to deployment in homes, as vision-language models (VLMs) are demonstrating significant advances in reasoning, making VLM-driven robots promising home assistant agents. These agents will often be introduced into homes that already have cameras, Internet of Things (IoT) devices, and residents with wearables. To fulfill their potential, they need to integrate with existing smart infrastructure and reason through ambiguous scenarios: a smoke alarm may come from a pan about to ignite, a pan already on fire, a resident cooking or a faulty sensor, and each calls for a different response. To that end, we introduce HomeLoop, a simulation framework and benchmark built on Isaac Sim that combines robot bodies, networked devices and wearables, simulated residents and visitors, faulty sensors, and hazards that evolve over time. Our benchmark is composed of 20 tasks across five families grounded in real-world household hazards and needs, and designed so that the same alert can have different causes. A runtime monitor scores each episode with temporal-logic checks on the final outcome and the actions along the way, and every action carries a predefined cost. We evaluate six open-weight and proprietary VLMs. The best achieves a family-balanced score of 61.9%, and across models we see a pattern of over-reaction: failed episodes involving prohibited interventions outnumber those involving missed required interventions by more than two to one, with models often shutting off the stove before investigating the alarm's cause. HomeLoop is an extensible foundation intended for use in research on robot-IoT coordination, long-term memory, proactive assistance, and privacy-safety trade-offs. HomeLoop will be released with 2,100 home layouts that can be automatically furnished in diverse configurations, five robot bodies and 20 networked device types.
embodiedbenchmark - arxiv:2611.08912 · cs.LGSSF-Net: A Spatial-Spectral Fusion Network for Accurate PDE Operator LearningSiqi Zhao, Flora D. Salim
Machine learning methods for solving partial differential equations (PDEs) significantly improve computational efficiency over traditional numerical solvers such as finite element method (FEM). However, PDE dynamics inherently span both local and global dynamics: disturbances propagate and decay gradually through the domain, while global coupling effects can influence the whole system simultaneously. Capturing both types of dynamics with a low complexity model remains a key challenge. To address this, we propose the Spatial-Spectral Fusion Network (SSF-Net), built around a dual-branch Spatial-Spectral Fusion (SSF) block comprising a spatial branch and a spectral branch. The spatial branch employs a multi-directional Mamba, which mitigates its inherent causal, single-direction scanning bias and capture local interactions with reduced directional bias at linear complexity. The spectral branch leverages a Fourier transform with a block-diagonal MLP to efficiently capture global spectral features with reduced parameter count and complexity. The SSF block is embedded in a multiresolution convolutional encoder-decoder, which extracts multi-scale features. We evaluate SSF-Net on a comprehensive PDE benchmark spanning multiple task types and assess its long-horizon stability. Compared with attention-based and convolution-based baselines, SSF-Net achieves lower prediction error with fewer parameters while retaining near-linear asymptotic complexity in the number of spatial tokens.
benchmark - arxiv:2611.08911 · cs.ROPolarization calibration of fisheye cameras using skylight in the wildShuo Wang, Shiqi Che, Dianze Dong, Wenxin Liu
Skylight polarization provides directional cues for robotic navigation, but fisheye lens effects and analyzer imperfections can degrade polarization measurements across the field of view. We present a joint lens-analyzer calibration method that explicitly models incidence-dependent lens transmission together with analyzer gains and angular offsets in a unified eight-parameter formulation. Given known geometric calibration, rotations between camera frames can be estimated using visual features and cross-frame directional polarized light consistency can be used for polarization calibration. Therefore, our calibration approach requires only a short handheld outdoor sequence under natural skylight, without dedicated polarization calibration equipment or a predefined sky polarization model. On real-world outdoor sequences, the proposed method achieves a weighted cross-frame angle-of-linear-polarization (AoLP) mean absolute error below 0.5° across all incidence angles, with a 73.4% reduction in the 80°-90° incidence angle region, and generalizes to an independent sequence with an overall mean absolute error of 0.466° without refitting. Preliminary solar-azimuth experiments using a camera mounted on a quadruped robot further show improved alignment with the visible solar center, supporting the potential of calibrated AoLP measurements for relative-heading estimation and future polarization--inertial navigation. The calibration software tool will be released upon acceptance of this paper.
quadruped - arxiv:2611.08908 · cs.AIToward LLM Self-Play Under Data Scarcity: Effectiveness, Generalizability, and TrustworthinessZhiyin Yu, Yuchen Mou, Bingzhe Wu, Zhonghai Wu +2
The rapid development of large language models (LLMs) has been driven largely by scaling model size and human-curated training data, yet this paradigm is increasingly constrained by data scarcity and diminishing returns from human supervision. Self-play offers a promising pathway for autonomous data generation and improvement, but current frameworks remain unsatisfactory due to a lack of a data-centric perspective, limited cross-domain generalization, and growing trustworthiness and security concerns. In this paper, we argue that real-world LLM self-play systems should achieve effectiveness, generalizability, and trustworthiness under data scarcity. To address these challenges, we outline three blue-sky research directions for future self-play systems: (1) data-centric self-play for effective experience generation, (2) generalizable self-play for open-world and unverifiable tasks, and (3) trustworthy self-play for reliable self-evolution. We further discuss research opportunities and solutions within each direction. This paradigm shift will establish a sustainable and safe AI ecosystem by transforming finite human knowledge into inexhaustible high-quality experience through self-play architectures.
self-play - arxiv:2611.08905 · cs.CLFrom Procedures to Permissions: Learning Precise Subagent Authorization from Agent SkillsWenkai Yu, Ning Ding, Wentao Zhang, Yehui Tang
As language agents tackle increasingly complex tasks, they increasingly rely on multiple agents to coordinate specialized work, requiring decisions not only about what each subagent should do but also which tools and workspace resources it should access. Existing agent systems provide mechanisms for controlling tools and execution environments, but do not directly specify the resources required by each delegated subtask. Moreover, task outcomes and execution trajectories alone provide limited supervision for determining whether a subagent has received appropriate permissions. We observe that Agent Skills encode procedural knowledge describing the operations, tools, and files required at each step, providing a natural source of supervision for authorization. Building on this observation, we synthesize tasks and step-by-step procedures from Skills, execute them through subagent delegation in interactive environments, and use verified trajectories to construct training data for task execution and authorization. On ClawArena-Team, our method improves tool permission precision by 13.4 percentage points and workspace permission precision by 24.1 percentage points, while increasing overall task success by 11.6 percentage points for Qwen3.5-35B-A3B. On Terminal-Bench 2.1 and SkillsBench, our method remains competitive with existing agent data synthesis approaches. These results show that the procedural knowledge embedded in Agent Skills provides a source of supervision for precise subagent orchestration that task outcomes alone cannot provide.
agentagent system - arxiv:2611.08904 · cs.CVMotionCore: A Latent Neural Simulator for Video Generation with Rigid-Body DynamicsZheng Wang, Zhengwei Gong, Yuanzhi Liang, Chi Zhang
Video generation models can produce highly realistic videos while generating physically inconsistent object motion. This limitation suggests that visual modeling alone may not provide a transferable representation of physical dynamics. We introduce MotionCore, a framework for learning physical dynamics via a learnable partial differential equation solver and focuses on rigid-body motion. We introduce a latent neural simulator learning to predict the evolution of object-level physical states from observed motion. Its predictions are then converted into spatially aligned latent controls for a controllable video diffusion model. The generator serves as a visual interface: it provides appearance and scene detail, while the learned dynamics model specifies the physical trajectory. To preserve this trajectory through the interface, MotionCore combines a diffusion-matching objective with a low-capacity decodability constraint on object identity and position. We evaluate MotionCore on public physical-consistency benchmarks and on paired data with known trajectories. Results indicate that MotionCore improves physical consistency and trajectory obedience in the evaluated rigid-body settings, while maintaining comparable visual quality. A controlled ablation protocol compares learned dynamics with an analytic prior and studies the choice of latent evolution processor.
neural simulatorbenchmark - arxiv:2611.08898 · cs.ROFATHOM: World-Model-Guided Diffusion Action-Trajectory Selection for Sparse-Marker AUV TrackingXiaorui Wang, Zeyu Sha, Mingyang Yang, David Chen Wang +1
Reliable relative tracking is fundamental to cooperative autonomous underwater vehicles (AUVs), yet remains challenging under nonlinear vehicle dynamics and degraded underwater visual conditions. Active light markers improve robustness to appearance variation, but provide deliberately sparse observations that can become incomplete or corrupted by occlusion, reflections, and detection noise during dynamic motion. To address the resulting partial-observation control problem, this paper proposes FATHOM, a world-model-guided diffusion action-trajectory selection framework for sparse-marker AUV tracking. FATHOM first integrates structured marker observations, inertial measurements, action history, and temporal dynamics into a predictive tracking belief using a multimodal task-supervised recurrent state-space model. A diffusion action prior then generates multiple expert-like multi-thruster trajectories, whose future tracking consequences are predicted through the learned latent dynamics. Rather than directly executing a generated trajectory, FATHOM selects the most favorable candidate under an explicit receding-horizon objective and continually updates the decision from onboard observations. Extensive offline analyses verify the learned predictive belief, multi-step world-model prediction, diffusion trajectory generation, and predictive selection. More importantly, comprehensive real-AUV experiments, including strong pose-based comparisons and component ablations, demonstrate robust autonomous tracking across nominal and challenging motions with naturally occurring partial marker visibility, while maintaining real-time onboard operation without motion-capture feedback.
latent dynamics - arxiv:2611.08897 · cs.CVSegmenting Geometric Textures of 3D MeshesWeidan Xiong, Kaijie Qu, Shihao Feng, Shunqi Fan +2
Recognizing geometric texture patterns on 3D models is critical for applications such as heritage protection, model editing, and 3D printing. Existing 3D model segmentation methods predominantly focus on identifying high-level semantic parts, thereby failing to capture middle-level textures. While recent Vision-Language Models can perform image segmentation with appropriate text prompts, they still struggle to recognize 3D geometric textures. In this work, we present a method that segments 3D mesh models based on middle-level geometric textures. We further introduce a new geometric texture segmentation benchmark and fine-tune SAM3 accordingly. The segmented texture regions are then clustered based on DINO feature similarity and lifted from 2D to 3D. Additionally, we develop an interactive tool to enable both 3D model editing based on our segmentation results and model retrieval from the dataset based on geometric and texture similarity. Experiments demonstrate that our method significantly outperforms existing methods in segmenting geometric textures and can facilitate various applications.
benchmark - arxiv:2611.08895 · cs.ROLocalizing and Editing the Goal Representation of a Vision-Language-Action ModelSudhanshu Srivastava, Faisal Aqlan
Operators can state safety constraints for robots as prohibitions such as "do not put the bowl on the stove". We study whether OpenVLA follows such prohibitions and where it represents the goal it executes. The 10 tasks of LIBEROGoal share one scene, so we can label the goal that each rollout executes. We combine closed-loop rollouts with activation patching, linear probes, activation steering and runtime monitors. On LIBERO-Goal, negation had no detectable effect on which task OpenVLA executed. The model executed the forbidden task in 79% of negated rollouts and completed 78% of plain tasks. A goal probe gives similar read-outs under "not X" and "X". Activation patching shows that the first action depends mostly on the instruction tokens before layer 12 and on the last prompt token from layer 16 onward. A fixed steering vector at the instruction tokens at layer 8 redirected 12 of 18 rollouts. Fixed vectors and prefill patches at layers 12 to 20 redirected none of 158. From layer 16, only patches at every decoded action token redirected the robot (15 of 18 rollouts at layer 24). When the instruction changes during a rollout, the layer-32 goal read-out detects non-compliance better than failure detectors, and an external probe on image and instruction features performs comparably. These results support enforcing prohibitions outside the instruction and monitoring goal read-outs against the specification. Goal edits in OpenVLA should act before layer 12 or patch every action token.
vision-language-actionopenvlalibero - arxiv:2611.08879 · cs.CVGeoJeV: Unified Spatial Probability for Training-Free Efficient Evidence Acquisition in Ultra-High-Resolution Remote SensingXinyi Niu, Lingfeng Chen, Jiehua Zhang, Panhe Hu +3
Ultra-high-resolution (UHR) remote sensing imagery provides rich spatial details but poses challenges for accurate and efficient multimodal understanding. Existing evidence acquisition methods selectively explore relevant image regions, yet largely rely on generative multimodal large language models (MLLMs) for intermediate search control. However, reinforcement learning from human feedback (RLHF) primarily aligns these models with human-preferred responses rather than explicitly optimizing region-level relevance judgments, while repeated autoregressive decoding introduces additional computational overhead. To address these limitations, we propose GeoJeV, a training-free method that formulates evidence acquisition as spatial probability-driven decision making. At each evaluated region, a frozen JeV model produces a probability distribution over candidate regions. Specifically, Region Rank prioritizes candidate regions, Sparse Branch adaptively determines branching width through sparse probability support, and Depth Gate controls refinement depth based on branch sparsity and distribution concentration. By coordinating search priority, branching width, and refinement depth through a shared distribution, GeoJeV reduces reliance on RLHF-aligned generation and eliminates intermediate decoding. Experiments on LRS-GRO, MME-RealWorld-RS, and LRS-VQA achieve accuracies of 60.59%, 52.79%, and 34.32%, respectively. Across the three benchmarks, GeoJeV improves the average accuracy of Qwen3-VL-8B by +5.32pp with only 1.12 additional seconds per question, while achieving a 7.8-fold inference speedup over the recent method.
benchmark - arxiv:2611.08865 · cs.AIRethinking Agent Trajectory Compression: A Workflow-Aware PerspectiveQijia Zhuang, Rui Liu, Yuxiang Ren
As long-horizon agents increasingly rely on multi-step tool interactions to accomplish complex tasks, their execution trajectories continuously accumulate structured outputs and intermediate artifacts, leading to growing context overhead. Existing compression methods primarily rely on text importance or fixed truncation rules, often overlooking cross-step dependencies within workflows and potentially discarding critical information required for subsequent task execution. To address this challenge, we propose WAAC, a workflow-aware framework for agent trajectory compression. WAAC structurally models successful historical trajectories to extract reusable workflow compression experience and incorporates online backward dependency analysis to dynamically identify information retention requirements. By combining current dependency constraints with historical experience, WAAC selects appropriate compression strategies for different types of historical information. Experiments on PinchBench and Odysseys demonstrate that WAAC achieves an overall score of 87.6 and a perfect completion rate of 19.0%, respectively, while reducing execution costs by 19.7% and 29.1% compared with the strongest baseline methods. These results demonstrate the effectiveness of workflow-aware trajectory compression in improving task performance and cost efficiency for long-horizon agents. Code and resources are available at https://anonymous.4open.science/r/waac-6527/.
agent - arxiv:2611.08864 · cs.LGAdaptive Scheduling Pipeline for Multi-Instance Asynchronous Reinforcement LearningSalah Chikhi, Luis Zuin, Entong Li, Zeng Li
We present the Adaptive Scheduling Pipeline (ASP), an offline rollout scheduler for multi-instance asynchronous reinforcement learning (RL) of large language models (LLMs). ASP jointly reduces three sources of rollout underutilization - padding bubbles from intra-minibatch length heterogeneity, skewness bubbles from long-tail outputs, and sorting bubbles from fluctuations in training-queue occupancy induced by length-sorted dispatch - through a single offline preprocessing pass. This enables efficient multi-instance rollout that overcomes the early scaling plateau of single-instance inference while smoothing the arrival of generated samples at the trainer. We integrate the scheduler into an asynchronous architecture that decouples rollout from training while keeping policy lag observable and bounded. Complete ASP improves end-to-end throughput by up to 3.22 times and reduces iteration time by up to 76.8% relative to MindSpeed-RL, while reducing iteration time by up to 62.5% relative to verl. Within ASP, scheduling reduces mean iteration time by 10.9%. ASP also scales more robustly with additional generation resources than these baselines and shows no visible convergence degradation in the measured bounded-lag regime. ASP has since been used in an internal telecom-domain RL post-training workflow, reducing the reported training period from more than three months to about two months and producing a domain-enhanced model for knowledge answering and tool calling.
tool callingpost-training - arxiv:2611.08861 · cs.CVAlign3D: Image-Structure Co-Aligned 3D Scene Generation with Observability-Aware Fidelity-Preserving GuidanceChenliang Zhou, Yifan Wang, Andrzej Banburski-Fahey
We present Align3D, an image-conditioned diffusion model that generates 3D indoor layouts that are both faithful to a single input photograph and physically plausible beyond the observed view. From the photograph we infer a confidence-calibrated scene graph and train a set-diffusion layout denoiser with image-structure co-alignment where confidence-weighted modality fusion balances image and scene-graph alignment using predicted confidences, while observability-aware fidelity-preserving guidance projects conflicting scene-quality gradients and reweighs them with an object-level confidence mask so under-observed degrees of freedom absorb physical priors without undoing the image-graph match. On 3D-FRONT, Align3D improves image-alignment and scene-quality metrics over recent baselines; once trained, it maps previously unseen photographs to layouts in a single feed-forward pass.
scene graph - arxiv:2611.08857 · cs.ROSTRIDE: Learning Humanoid Locomotion in Minutes with Acceleration-Based RewardsNico Bohlinger, Jan Peters
Learning locomotion directly on a humanoid robot is a promising direction to avoid the sim-to-real gap. However, locomotion rewards in Reinforcement Learning (RL) typically track the base velocity, which cannot be measured onboard without an external motion capture system or well-calibrated state estimator. This is why we introduce \method{}, which learns humanoid locomotion from minutes of samples with a reward that only requires the onboard Inertial Measurement Unit (IMU). Instead of tracking the base velocity, STRIDE rewards the robot for following a commanded acceleration profile. To learn efficiently from this reward, we use the off-policy RL algorithm XQC and combine a high gain setting for the joint position controllers with left-right symmetry augmentation. In simulation, STRIDE learns to walk on the Unitree G1 and Booster T1 humanoids within 14 minutes of samples. Compared to learning without these two components, it reduces the falls during training by 81%, while without them the robot learns only in over an hour.
humanoidsim-to-real - arxiv:2611.08827 · cs.ROActPrune: Action-Grounded Visual Token Pruning for Efficient Vision-Language-Action ModelsSungwon Moon, Minji Kim, Hyemin Boo, Daeun Song +1
Vision-language-action (VLA) models perform versatile manipulation tasks from language instructions, but their inference cost limits how fast a robot can respond in closed-loop control. Visual token pruning reduces this cost; however, prior methods estimate token redundancy from encoded visual embeddings or attention, only after the visual encoder has processed the entire observation. We propose \ours, a training-free token pruning method that moves the pruning decision ahead of the vision encoder by grounding it in a geometric prior derived from the robot's own action. The action chunk predicted at the previous control step is already available before the current observation is encoded, and it specifies the end-effector and camera motion between consecutive observations. The same action affects the two standard camera views asymmetrically: in the fixed third-person view, \ours retains patches along the projected end-effector trajectory together with semantic anchors, while in the wrist view, it prunes patches that remain unchanged after compensating for the action-induced camera motion. Extensive experiments on two simulation environments (LIBERO and VLABench) and four real-robot tasks demonstrate that \ours achieves up to 1.61 $\times$ speedup and 59.5\% FLOPs reduction with negligible degradation in success rates.
vision-language-actionmanipulationlibero - arxiv:2611.08826 · eess.SYAn Ultra-Low-Noise Tunable TIA in 45nm Silicon Photonics for Electro-Optical Ultrasound Data AcquisitionMarziyeh Rezaei, Sridhar Singhal, Alp A. Derin, Soner Sonmezoglu +1
This paper presents the design and experimental characterization of a 32-channel, WDM-compatible electronic-photonic receiver array fabricated in the GlobalFoundries 45nm monolithic silicon photonics process. Optimized for miniaturized, millimeter-scale electro-optical ultrasound data acquisition probes for broadband photo-acoustic imaging (PAI), the chip utilizes integrated micro-ring resonator (MRR) demultiplexers and photodiode (PD) arrays to decode multi-wavelength acoustic signals outside the body. This work focuses specifically on the comprehensive characterization of the receiver front-end, evaluating both the optical WDM filter bank and the transimpedance amplifier (TIA) performance. Monolithic co-integration minimizes critical photodiode-to-TIA interface parasitics. The TIA incorporates a DC-offset cancellation loop that compensates for photodiode-induced offset currents while providing tunable low-frequency cutoff control over a range of 30 kHz to 1.5 MHz. This tunability enables effective suppression of flicker noise and improves overall signal integrity. The front-end achieves the peak transimpedance gain of 144 dB$Ω$ while maintaining an ultra-low input-referred noise current density of $1\text{ pA}/\sqrt{\text{Hz}}$, demonstrating a highly scalable architecture for high-sensitivity intravascular and endobronchial imaging.
silicon photonicsilicon photonics - arxiv:2611.08814 · cs.ROIt Takes Two to Tangram: Multi-Agent Object Centric Reinforcement Learning for AssemblyPaul Knaußt, Kay Pompetzki, Nico Bohlinger, Niklas Funk +3
Geometric assembly requires fine spatiotemporal coordination to place parts without causing collisions or deadlocks under tight clearances. To address this, we present an object-centric reinforcement learning framework that treats each part as an agent and learns a shared policy for continuous assembly control. A graph neural network represents the pieces, their target poses, and their spatial relationships, and predicts translational and angular velocities for each component. Per-agent proximal policy optimization with a centralized graph critic directly learns motion and temporal coordination within a single feedback policy. We study this formulation through planar Tangram assembly, where pieces must slide and rotate into a given configuration while remaining on the assembly surface. Procedural generation of target configurations and randomized initial arrangements support training and evaluation across diverse assembly problems. Our experiments isolate the roles of relational structure and multi-agent optimization, and compare learned control with sampling-based planning. We demonstrate concurrent assembly in simulation and transfer the policy to sequential execution on a real robot, as our object-centric formulation decouples learned assembly coordination from low-level physical control. Project page: https://tangrambench.org/
agentmulti-agent - arxiv:2611.08811 · cs.AII Think, Therefore I Attack: Evaluating Thought Injection Attacks on Reasoning AgentsParth Illendula, Garik Livingston, Zain Shamsi, Ravi Srinivasan +1
Large Reasoning Models (LRMs) are one of the key components of current agentic systems. LRMs augment their output with an explicit chain-of-thought reasoning stage, introducing a new attack surface for thought injection, in which untrusted text resembling the model's own deliberation is fed back to the model. We present a two-part study of this attack, performed against the prevailing large, production class open-weight models (e.g., GLM-5.3, MiniMax-M2.5, Qwen3.8-Flash-Next). We first evaluate the effectiveness of static thought-injection prompts under different styles as well as tagged and untagged conditions and find that the effects of thinking tags and payload styles are all properties of the specific target model. In the second study, we exploit this model-specificity in a containerized agentic cyber range, using indirect injection prompts styled in the form of the target model's own reasoning traces to achieve attack success rates (ASR) up to 50% on operational objectives such as reverse-shell exfiltration and repository backdooring. We also find that the same malicious objective in standard prompt injection user voice never succeeds, and thinking tags suppress the agent's tendency to disclose the injection. Our results show that imitating a reasoning model's own thought style is effective for this style of attack, and that tag stripping is an unreliable defense.
agentic - arxiv:2611.08805 · physics.opticsLaser-Direct-Written polymer WGM microring laser integrated on a fiber tip for refractive index sensingMobin Yaghoobi, Seyedeh Mahshad Hosseini, Leila Hajshahvaladi, Gholam-Mohammad Parsanasab +1
Whispering-gallery-mode (WGM) microlasers are promising for refractometric sensing, but their practical implementation can be limited by the need for stable resonator-coupler alignment. Here, we demonstrate a Rhodamine B-doped polymer WGM microring laser directly fabricated on an optical-fiber tip, eliminating the external resonator-fiber coupling interface at the sensing head. Micro-droplet coating and UV laser direct writing are combined for fabrication, while finite-difference time-domain simulations guide the microring laser design. The microlasers operate in air and water with thresholds of 0.224 uJ and 0.458 uJ, respectively. Refractometric measurements in aqueous glucose solutions yield a bulk refractive index sensitivity of 364.95 nm/RIU and a detection limit of 2.47 x 10^(-4) RIU. Water-induced polymer changes and photobleaching are also investigated. This fiber-tip architecture provides a basis for probe-based refractometric sensing in environmental and biochemical applications.
microring - arxiv:2611.08797 · cs.LGMotionBench-XAI: Benchmarking What Shapes Shapley Explanations in Spatiotemporal Machine LearningMathilde Papillon, Louisa Cornelis, Nina Miolane
Predicting outcomes from spatiotemporal data, such as assessing Parkinson's severity from human motion, often requires explainability in addition to accuracy. Practitioners need to know what drove the prediction, whether it be spatial (a body joint), temporal (a specific window), or both. Shapley values are attributions computed by treating input features as players in a cooperative game and measuring how the model output changes as coalitions of players are removed. Computing them involves choosing what counts as players, how absent players are imputed, and how the game is evaluated. How these choices affect explanation quality is poorly understood for spatiotemporal data, and real data carries no ground-truth attributions to check against. We present MotionBench-XAI, a benchmark built on i) five synthetic spatiotemporal datasets with exact ground-truth Shapley values, ii) three real datasets across clinical gait, electrocardiogram, and audio, iii) ten attribution methods across three player sets, and iv) three trained architectures. We show that grouping features into players, across time, space, or both, matters more than the choice of imputer. Generative imputers improve explanations only when their conditional models are accurate. On real data, the standard proxy metrics select contradictory best methods. We distill these findings into practical insights and provide an open-source pipeline.
benchmark - arxiv:2611.08794 · cs.LGHigh-Dimensional Off-Policy Evaluation without Sample Splitting using an Onsager CorrectionSaksham Jain, Guillaume Braun, Masaaki Imaizumi
Reward models used for off-policy evaluation (OPE) are commonly selected using held-out data. When sample size and context dimension grow proportionally, the difference in test risk between models fitted on the full sample and on a fixed fraction of it need not vanish. Held-out evaluation can therefore consistently estimate the test risk of the model trained on the smaller sample, but because the relative ranking of candidate reward models can change after full-data refitting, held-out selection can be suboptimal. We develop an Onsager-corrected, target-policy-weighted criterion that estimates the full-data reward model's out-of-sample risk under the target policy using only its training observations. This enables early stopping along the full-data gradient-descent trajectory without sample splitting. For neural networks with fixed depth and hidden width under Gaussian contexts, single-index rewards, context-dependent multi-index target policies, and known context-independent logging, we prove uniform consistency of our criterion over a fixed finite set of gradient-descent iterates and a near-oracle guarantee for the selected reward model. The same correction yields a doubly robust estimator that consistently estimates the true policy value without sample splitting, even when reward-model prediction error does not vanish. Experiments show decreasing risk-estimation error as dimension grows at a fixed sample-to-dimension ratio, and improved reward-model selection relative to held-out baselines. Empirically, our correction also reduces the root mean squared error of doubly robust value estimates computed from the same data used to fit and select the reward models.
policy evaluation - arxiv:2611.08789 · cs.LGTopoSHAP: Shapley for Explainable and Efficient Topological Deep LearningMathilde Papillon, Nina Miolane
Topological deep learning (TDL) extends learning on graphs, such as molecules or social networks, to higher-order structures, such as rings in a molecule or groups in a social network. Explaining which of these structures drive a model's predictions, and which architectural components matter most for its performance, remains challenging. We introduce TopoSHAP, an efficient and general explanation framework that addresses both questions across topological domains and model architectures. TopoSHAP-cell assigns Shapley values to topological data cells and extracts compact explanations. These explanations match or outperform graph-based counterparts on GraphXAI benchmark data while running up to 112$\times$ faster. They also reveal advantageous topological feature design, leading to measured improvements in explanation recovery and predictive accuracy. On the architecture side, TopoSHAP-neighbor explains how neighborhoods contribute to model performance, and turns these evaluations into a backward-elimination search that identifies smaller, competitive architectures early in training. This is up to $11.3\times$ cheaper than traditional TDL training sweeps.
benchmark - arxiv:2611.08788 · cs.LGNot All Latents Are Equal: Importance-Weighted Latent Graph GenerationBiraaj Rout, Jay Revolinsky, Han Liu, Harry Shomer
Graph generation has shown tremendous effectiveness in modeling complex scientific tasks ranging from molecule generation to optimized circuit design. Most recent advancements have developed models that learn to denoise noisy graphs using either discrete diffusion or flow matching. However, discrete graph diffusion scales quadratically with the number of nodes, making it difficult to use such models on larger graphs. On the other hand, latent graph generation instead seeks to denoise noisy graph latents, which scale better to more nodes due to their lower dimension. However, in practice, latent graph generation models lag behind their discrete counterparts in performance. We study this problem, finding that it is caused by the unequal distribution of that different latent dimensions have on graph reconstruction. the impact. We observe that the vast majority of impact is contained in only a smaller number of latent dimensions. Since equal emphasis is placed on reconstructing the entire latent in the loss function, existing models are optimized to reconstruct information irrelevant to the downstream task. To tackle this problem, we propose Latent Importance Reweighting ({\bf LIR)} which weights each latent dimensions via it's importance to final graph reconstruction. Importance is calculated as the relative influence of a single latent unit on graph reconstruction, which is represented via the Jacobian of the decoder relative to the latent. Our method is model-agnostic and requires no additional model parameters or hyperparameters. Experiments show that LIR can boost performance and lead to faster training convergence when applied to two different autoencoders on various benchmark datasets. Code: https://github.com/biraaj/GraphGenLatentReweight
benchmark - arxiv:2611.08782 · cs.CVExtrapolation Matters: Sparse-Only Depth Completion Beyond LiDAR CoverageMaximilian Jansen, Gasser Elazab, Praveen Kulkarni
Extrapolation beyond LiDAR coverage is an important yet under-evaluated challenge of depth completion. In real camera--LiDAR systems, partial sensor overlap, occlusions, and limited field coverage often leave large image regions weakly observed or completely unobserved by LiDAR. Thus, we investigate depth completion in this beyond-coverage regime using only sparse supervision from raw single-sweep LiDAR, without dense depth labels, LiDAR accumulation, or pseudo-LiDAR generation. Our approach leverages a frozen monocular depth foundation model, lightweight per-frame alignment to sparse LiDAR geometry, and a refinement module that fuses the aligned prior with sparse LiDAR anchors to predict dense metric depth. To regularize learning under raw sparse supervision, we introduce a high-frequency normal-consistency loss defined as the image-space gradient of the per-pixel normal cosine discrepancy. This term penalizes spatial variation in scalar angular error; it does not directly match changes in neighboring normal vectors. We further introduce an evaluation protocol that explicitly separates LiDAR-supported and beyond-coverage regions, revealing extrapolative behavior that conventional depth-completion metrics can obscure. Our method achieves strong depth accuracy beyond LiDAR coverage while remaining competitive in LiDAR-covered regions, providing a practical way to deploy monocular foundation priors in multi-sensor systems.
evaluation protocol - arxiv:2611.08781 · cs.AIAligning, Fast and Slow: A Control-Theoretic Oversight Framework for Long-Horizon Trajectories and Maintained DeploymentsStephen Peng, Vineet Jagadeesan Nair, Noah Broestl, Anne Friederike Kleppe +2
AI agents can pass checks on individual tasks while retaining memory or code changes that alter their later behavior. We introduce a control-theoretic framework that couples a fast loop controlling proposed actions with a slow loop reviewing persistent changes and reassessing configurations after model or tool updates. When measurement errors and changes between checks are bounded and controllers can enforce their decisions, we derive a deployment-level bound and a rule for dividing review effort. In simulation, dividing a fixed review budget between both loops reduces mean maximum departure from intended behavior roughly twelvefold compared with review confined to individual runs. In coding stress tests with authored memories and five matched seeds, combined controls reduce mean severity on a 0-3 scale from 0.325 without gates to 0.117, resolving 33 of 120 issues versus 35 without gates; in trading stress tests, combined controls reduce mean shortfall in low-emission purchases from 0.892 without gates to 0.107. The overall framework helps developers and auditors compare and enforce oversight designs, while the agent studies test component controls rather than the full deployment-level guarantee.
memoryagentai agent - arxiv:2611.08780 · cs.AIGuardrail Erosion in Domain Fine-Tuned Large Language Models: A Paired Audit of Knowledge-Centric Abstention in Biomedical LLMsStephen Peng, Vineet Jagadeesan Nair, Noah Broestl, Anne Friederike Kleppe +2
The number of domain fine-tuned models, particularly in the biomedical domain, is rapidly expanding. Fine-tuning promises high-performing models at a fraction of the cost of training a base model, but it can also produce unintended behavioral changes. Multiple studies have shown that safety guardrails can erode with fine-tuning, yet large-scale studies across model architectures, data mixtures, and fine-tuning methodologies remain limited. Where prior work in this space has studied value-centric refusal, this work focuses on knowledge-centric abstention. We audit 25 biomedical domain fine-tuned large language models against 11 paired base checkpoints on MedAbstain and ClinDet-Bench to evaluate the impact of fine-tuning on knowledge-centric abstention. We find that defined appropriate abstention measured by these benchmarks, already low in base models (median 8.3% on MedAbstain), falls further after fine-tuning (median 4.2%, decreasing in 21 of 25 pairs on MedAbstain), and that the magnitude and direction of change are heterogeneous across pairs. These results motivate fine-tuning methods that improve domain-specific capability while preserving appropriate caution.
benchmark - arxiv:2611.08779 · cs.LGJacobian-Causal Memory Control for Persistent Memory in Language Model SystemsMarvin Danig
Persistent-memory systems choose not only which records are retrieved but which prior information can alter current inference. We propose Jacobian-causal memory control (JCMC), which represents each authorized memory intervention by a predicted influence signature in a shared behavioral measurement space, selects interventions against a target effect and side-effect budget, and compares predicted with observed effects before retaining or rolling back the intervention. Under explicit smoothness and bounded-estimation assumptions, we prove a quadratic local-approximation bound, characterize the target effects reachable by a set of signatures, bound selection regret under signature error, and bound non-additive interactions between simultaneous memories. The paper reports no benchmark measurements; it defines a paired counterfactual estimator and an evaluation protocol comparing JCMC with relevance, graph, and policy-filtered retrieval baselines. The central claim is narrowed or falsified if those simpler baselines match effect targeting, side-effect control, and rollback utility under equal model, memory, policy, and cost budgets.
memorypersistent memorybenchmarkevaluation protocol - arxiv:2611.08768 · cs.AIConflux: Preventing Privilege Escalation with Principal Contexts and System-Level EvaluationRaya Buckley, Christian Schroeder de Witt
LLM agents combine information from multiple principals while acting externally. We formalise privilege escalation (PE) as the system-level security objective: an action may proceed only if every principal that influenced it would allow it. Model-level defences remain probabilistic, so enforcement must sit outside model judgement. We present Conflux, a principal-context defence. Its kernel, ITES, derives authority from provenance and the organisation's existing access-control system and intersects the permissions of every influencing principal. We prove a conditional authority-confinement theorem: under complete mediation, no-laundering provenance, a known non-empty decision context, fresh permission checks, and action binding, the kernel admits no action that escalates privilege with respect to any recorded influencing principal. The theorem covers the represented value-influence fragment; control influence from tool, branch, recipient, or plan-step selection remains a refinement obligation. Conflux extends the kernel with independent policy ports, argument-aware authorisation, disclosure, consent, delegation, restriction, and exposure planning. We describe two system- level evaluators. SLED is a black-box evaluator: it exhaustively explores a declared finite model of a defence up to a bound. SLED-V is a white-box evaluator: it checks a serialisable intermediate representation against security and provenance properties. The evidence is bounded and partial. The appendix reports retained AgentDojo, SLED, and SLED-V results, including the limits of those runs. No result in this paper is an unbounded deployment guarantee.
llm agentevaluator - arxiv:2611.08763 · cs.AICrowdDirector: Runtime Orchestration of Crowd Simulation through Natural-Language DirectivesXiaohan Sun, Vuk Radovanovic, Merce Alvarez, Carol O'Sullivan +2
Natural language offers an intuitive way to direct a virtual crowd while a simulation is running. Handwritten rules execute efficiently but cannot respond flexibly to varied instructions, whereas asking a large language model (LLM) to control the crowd at every update introduces repeated latency, cost, and the risk of unsupported actions. We present CrowdDirector, a framework that uses an LLM only when the user adds or changes an instruction. The LLM translates the request into a structured instruction, and a scene validator checks whether the entities and actions it references exist in the current Unity scene. Accepted instructions are stored and remain continuously active as the simulation is running. At each simulation tick, a graph neural network uses the current scene state and active instruction to choose an available action and target for each agent, while Unity handles movement, interactions, and animation. The network is trained offline on simulated scenarios, with LLM guidance on how agents should respond to instructions. Our evaluations show the scene validator's feasibility decisions matched human judgements more often than the LLM's own assessment, and on a behavioural benchmark CrowdDirector follows user instructions more closely than handwritten rules while avoiding repeated LLM calls. It therefore provides a practical middle ground between fixed rules and continuous LLM control.
benchmark - arxiv:2611.08756 · cs.AIDebate to Deceive: Poisoning LLM's Fact-Checking via Game-Theoretic Multi-Agent DebateLiting Huang, Huizhi Liang, Yi Lei, Xingjun Ma +1
Knowledge attacks on Large Language Models, such as manipulating fact-checking judgments, pose significant risks to AI reliability. Existing approaches -- adversarial suffix/template injection and single-agent self-reflection -- lack robustness across diverse defenses and are prone to confirmation bias and costly trial-and-error, respectively. We propose Debate-to-Deceive (D2D), a multi-agent debate framework grounded in human debate structure and rhetorical style, enabling agents to generate adversarial passages that are semantically coherent, rhetorically effective, and resistant to early confirmation bias. D2D formalizes the adversarial process as an evolutionary game in which agents adaptively evolve their rhetorical strategies (Logos, Pathos, Ethos) through competition. To manage the intractably large LLM output space, we decouple agent behavior into a strategic layer, which selects a rhetorical action from a compact strategy space via evolutionary-inspired learning dynamics, and a generation layer, where the LLM produces a concrete passage conditioned on that strategy and the debate context. On four fact-checking benchmarks (FEVER, HealthVer, CLIMATE-FEVER, and SciFact), D2D substantially outperforms existing attacks and evades most defenses.
agentmulti-agentbenchmark - arxiv:2611.08755 · cs.CVAtariBench: Evaluating Spatiotemporal Reasoning of Vision-Language Models in Atari GameplayMingjia Huo, Yao Fu, Bo Chang, Yaqing Wang +7
Vision-Language Models (VLMs) have achieved strong performance on static visual understanding benchmarks, yet their ability to act in dynamic environments remains less well understood. We introduce AtariBench, a controlled benchmark that evaluates VLMs across 20 Atari 2600 games, requiring models to make sequential decisions from high-temporal-resolution gameplay streams. By decoupling fast in-game dynamics from wall-clock inference speed, AtariBench focuses on whether models can perceive rapid visual state changes, reason over spatiotemporal dynamics, and ground actions in the current observation. Our experiments reveal substantial limitations in state-of-the-art VLMs: models often fail to infer object motion and direction, exhibit brittle sequential decision-making, and over-rely on textual interaction history despite newly appended visual feedback. We further introduce a separability analysis to identify which in-game scores provide reliable signals of model competence, and use the benchmark to study how memory design and prompting strategies affect performance. Overall, AtariBench exposes a persistent gap between passive visual perception and active decision-making, providing a lightweight and rigorous testbed for developing VLMs that can reason and act in dynamic environments.
memorybenchmark - arxiv:2611.08753 · cs.LGRandom Symbolic Boosting: A Geometric Approach to Symbolic RegressionJosé R. Ortegaa, José Manuel Muñoz Contreras, Leonardo Trujillo
Symbolic models offer interpretability, but their construction through generational evolution can be computationally expensive. This article presents Random Symbolic Boosting (RSB), a method that constructs explicit additive symbolic models via boosting over populations of randomly generated syntax trees, eliminating the need for generational evolution. The idea combines two observations: high-dimensional spaces can be spanned by enough linearly independent vectors, and the semantics of syntax trees can be directly manipulated as such vectors. It is shown that, for Ordinary Least Squares, the boosting process guarantees a monotonic reduction of the training error for both regression and classification. Furthermore, an adaptive geometric early stopping mechanism is introduced. It is based on the angle between the prediction and target vectors, is independent of the problems scale, and requires no adjustments across tasks. Evaluated on a standard benchmark for symbolic regression, the method is located at the Pareto-optimal point between predictive accuracy and training time, ranking first on the real-world datasets. Additionally, it was tested on a large tabular classification benchmark with 200 problems, achieving competitive results.
benchmark - arxiv:2611.08741 · cs.AIAlterNet: Adaptive-Depth Mixture-of-Experts with Learned Halting for Elastic InferenceAlexander Kastius
Most LLM architectures have fixed compute demands determined at training time. We propose a novel architecture that combines the Mixture-of-Experts architecture with shallow experts and a novel dynamic-depth system to make the extent of a forward pass dynamically adjustable. Our approach lets the model automatically determine the compute needed for a given task, paired with a system to scale it down after training without retraining when the execution environment demands faster or cheaper inference. To achieve this, we introduce an iterative computation mechanism, sequence-level depth-based routing steered by a halting mechanism within the experts, a depth encoding that lets the model track its progress in the chain, and a regularization mechanism that incentivizes the model to use allocated compute efficiently. We present initial benchmarks and discuss future research directions.
benchmark - arxiv:2611.08731 · cs.CVeSharp: Diffusion-based Sharp Monocular Metric Event-to-Depth EstimationQiushi Dai, Aniket Dashpute, Yuhao Liu, Rishubh Parihar +2
Depth perception is fundamental for autonomous navigation, industrial inspection, and mixed-reality systems. It often operates under strict constraints on power and form factor. In such cases, monocular depth estimation (MDE) is often used. Even as recent advances continue to improve monocular depth quality, conventional frame-based MDE systems remain challenged by fast motion and high-dynamic-range scenes. This work introduces eSharp, an MDE framework that addresses these problems using an event camera. eSharp repurposes pretrained diffusion priors for event-only MDE and produces accurate metric depth with unparalleled sharpness on real data. Real-world event datasets provide only sparse LiDAR depth maps for ground-truth supervision, making dense prediction challenging. We address this limitation through cross-modal distillation, using LiDAR-calibrated dense pseudo-labels generated from a spatially co-registered frame camera to train eSharp. Experiments show that eSharp outperforms prior event-to-depth baselines in both metric-depth accuracy and spatial fidelity, establishing diffusion models as a promising foundation for event-driven 3D perception.
event camera - arxiv:2611.08723 · cs.MADivide, Diagnose, and Compose: Optimizing Agentic Pipelines and HarnessesDivij Handa, Christopher Tensmeyer, Jihyung Kil, Vlad I. Morariu
Building agentic pipelines that achieve high task performance while keeping cost and latency low requires jointly optimizing heterogeneous and coupled design choices, including prompts, topology, model selection, and tool policies. Manually searching this combinatorial design space is expensive and time-consuming, while existing automated optimizers typically operate within a predefined subset of this space, modify the pipeline monolithically, or provide little structured evidence for pipeline changes. To mitigate these challenges, we introduce the Decomposed Credit-Assignment Framework (DeCAF), which adapts the design space to the given benchmark and pipeline, optimizes each dimension separately, and generates reasoning traces for every decision. Given a task specification, benchmark examples, and a seed pipeline, DeCAF identifies relevant optimization dimensions and generates diverse experimental hypotheses for each. Independent coding-agent sessions then optimize each dimension under a restricted edit scope, jointly evaluating performance, inference cost, and latency while recording all hypotheses, edits, and outcomes of each experiment. A final synthesis session combines compatible findings across all dimensions and uses a component-level trajectory evaluator, which derives role-specific rubrics from each agent's instructions, to localize underperforming components and guide cross-dimension revisions. Across a diverse suite of agentic benchmarks, DeCAF improves average task performance by 2.58 percentage points, while reducing inference cost by 4.30% and end-to-end latency by 42.45% compared to the best overall baseline.
agenticbenchmarkevaluator - arxiv:2611.08721 · cs.AIRemembering Is Not Enough: Evaluating Timing and Omission in Longitudinal AI Mental Health SupportEli Min, Mohit Chandra, Aamogh Sawant, Mengting Tu +7
Evaluations of large language models (LLMs) for mental health support commonly focus on the clinical appropriateness of individual responses, potentially overlooking failures that emerge across extended interactions. As AI assistants accumulate memory of users' circumstances and prior disclosures, this raises an accountability question: Do AI chatbots provide appropriate support as users' needs evolve? We investigate by analyzing 14,256 longitudinal simulations grounded in donated chatbot histories from 22 users, examining how a memory-equipped assistant responds to suicidal ideation, self-harm, and anxiety crises across recovery trajectories and communication styles. We introduce the Context-Aware and Readiness-Evaluative (CARE) framework, developed with mental health professionals, to assess the timing and quality of advice, including omissions. Advice was less aligned with users' readiness after a crisis than in matched control stretches of the same conversation, largely because of premature recommendations, which were most common when recovery stalled. Revisiting a crisis reduced premature advice but not lagging advice, and users' communication style changed how often and when the assistant gave recommendations. Although explicitly unproductive advice was rare, the assistant frequently omitted clinician-expected support such as follow-up and planning for next steps. These findings show the limits of single response-level safety evaluation and the need to assess timing, omission, and consistency across users when holding AI systems accountable for longitudinal support.
memory - arxiv:2611.08718 · cs.AIMemory-Aware Multi-Agent Architecture with LLM-based Orchestration for Conversational RecommendationEmilly Pitman de Castro, Vanessa Silva, Gustavo Kanno, Marcus Santos +17
Conversational recommender systems often treat user sessions in isolation, limiting their ability to capture how user intent evolves over time. While large language models (LLMs) have improved semantic understanding, they are typically used as isolated components, without addressing how memory, retrieval, and decision-making should be integrated at the system level. In this work, we propose a memory-aware multi-agent architecture for conversational recommendation, in which LLMs act as orchestration layers that coordinate intent refinement, query construction, and interaction flow. A key design principle is the integration of long-term memory as a first-class component, influencing both routing decisions and semantic query formation. We evaluate the system through cross-session memory experiments, multi-turn conversational simulations, and supervisor-driven interaction studies on the Amazon Reviews dataset. Results show that memory-enabled interactions recover up to 63\% of the gap toward explicit-query scenarios and prevent semantic degradation in extended dialogues. In addition, supervisory orchestration increases user acceptance by +28.9 percentage points, highlighting the importance of structured orchestration in guiding interaction and output quality. We also observe an exposure bottleneck between retrieved candidates and displayed recommendations. While LLM-reranking does not improve Hit@K, it consistently enhances semantic consistency and diversity. To capture these effects, we introduce the Semantic Consistency Score (SCS), a complementary evaluation measure. Overall, the results emphasize the importance of system-level design choices, suggesting that effective conversational recommendation benefits from explicit memory integration, orchestration, and evaluation beyond traditional ranking metrics.
memorymulti-agent - arxiv:2611.08714 · cs.LGPatching the Blind Spots: Reversible LLM Rule Overlay for Frozen Fraud DetectorsBo Qu, Zhurong Wang, Frank Zahradnik, Ming Ouyang
In industrial tabular fraud classification, gradient-boosted trees remain state-of-the-art, but even a well-trained model has blind spots, and retraining the deployed model is not always safe. We propose a framework that patches such a model without modifying it: LLMs propose rules anchored on features the model underweights by SHAP, and the resulting noisy pool is screened for empirical support and calibrated into an additive, reversible log-lift overlay that runs at the production cut-off. Single-LLM benefit is high-variance and inverts between datasets, so we pool three diverse backbones by equal-weight logit averaging, trading peak gain for a bounded worst case. On a real-world e-commerce fraud detector trained on selection-biased data, continued training regresses even when carefully tuned, whereas the overlay gains on average +2.26% recall at equal precision ($Δ$R@P) and +1.60% precision at equal recall ($Δ$P@R) over production cut-offs spanning 30-57% precision and 23-63% recall. On a public benchmark of seven designed bias-shift variants it raises $Δ$P@R on every variant and keeps $Δ$R@P near neutral. Each amendment is a readable predicate naming the fraud pattern it covers.
benchmark - arxiv:2611.08706 · cs.LGDual-Stream LLM-Guided Reward Design for Reject-Inference-Aware Reinforcement Learning in E-Commerce Payment Fraud DetectionBo Qu, Zhurong Wang, Frank Zahradnik, Ming Ouyang
Production fraud-detection models learn from a censored slice of their own traffic: transactions a legacy system blocked never receive fraud labels, so models train on a biased, approved-only distribution. We extend LLM-guided reward design, the framework of Qu et al. (2025) in which an LLM evolves the reward of a two-step REINFORCE fraud-blocking policy, to attack this reject-inference problem inside the reward. A single LLM designs a dual-stream reward: stream 1 applies the standard dollar-weighted precision objective to labeled transactions; stream 2 scores legacy-blocked, unlabeled transactions by contrasting the newest gradient-boosted risk signal against the legacy scores that drove the original block. Rather than prescribe which signals matter, we expose a superset of evidence and let the LLM decide for itself which to use, guided by evaluation feedback alone; because it writes the reward as code, that choice is directly auditable. On a held-out four-month out-of-time test, the dual-stream agent beats the prior labeled-only agent at every target recall, by +4.7, +2.4 and +3.0 points of dollar-weighted precision at 80%, 85% and 90% recall, and the production baseline by +6.8 points on average. The resulting policy is architecturally identical to that of Qu et al. (2025) and adds no inference cost.
agent - arxiv:2611.08698 · cs.ROVICTR: Value-Informed In-Context Learning for Vision-Language-Action ModelsLuca Macesanu, Tianluo Zhang, Aditya Patil, Archit Sharma +4
Recent advances in robot foundation models have highlighted the potential of in-context learning (ICL) for adapting robot behavior from demonstrations without further fine-tuning. Previous work has used retrieval-based post-training to equip pretrained vision-language-action (VLA) models with the capability for ICL by teaching them to condition action predictions on retrieved visually-relevant demonstration chunks. However, the effectiveness of retrieval hinges on selecting context relevant to the current execution stage, which visual similarity alone may not capture. We introduce VICTR, a framework enabling in-context learning in VLAs via progress-informed context retrieval built around an In-Context Value Function Estimator (IC-VFE). IC-VFE conditions on a reference demonstration video, a language instruction and the current observation to estimate task-relative progress. By conditioning progress estimation on a reference demonstration, IC-VFE can align the current state with a demonstrated execution even for tasks unseen during training. VICTR uses the estimated progress to select demonstration chunks at corresponding stages of execution and supplies them to a $π_{0.5}$ policy post-trained for in-context action generation. On unseen out-of-domain real-world manipulation tasks, VICTR raises the success rate from 6.2% for a fine-tuned $π_{0.5}$ baseline to 23.8% and more than doubles the task progress rate, from 28.9% to 55.5%. Project page with code, data and videos: https://ai4ce.github.io/VICTR/
vision-language-actionmanipulationrobot foundation modelpost-training - arxiv:2611.08693 · cs.LGEnabling Retraining-Free Accurate Inference in Analog Compute in Memory AcceleratorsGeorge Karfakis, Puneet Gupta
Analog compute-in-memory (CIM) accelerators compute matrix-vector products directly inside memory arrays, eliminating weight movement and promising order-of-magnitude gains in inference efficiency. However, converter and device errors make pretrained models unusable on these arrays, and hardware-aware retraining is costly. We enable retraining-free, accurate inference of pretrained models on simple analog arrays with low-cost, shared calibration. Our key observation is that analog error depends on how each layer is represented on the array, and that this representation can be changed offline without changing the computation. We center the inputs and apply Hadamard rotations designed for analog arrays, which keep shared reference errors spread and confine input mixing to each array. Across CNNs, vision transformers and LLMs up to 14B parameters on PCM and ReRAM, the method brings vision models and larger LLMs within a few points of floating-point accuracy and matches or approaches published hardware-aware training without training any weights. On ReRAM with 6-bit ADCs, it restores ViT-B/16 from 37.4% to 82.9% and Phi-4 from 27.0% to 72.2%, both within 2.2 points of floating-point accuracy.
memory - arxiv:2611.08666 · cs.AIPdb or Not to Be: Comparing Debugging Interfaces With Pass$_{X}$@$k$Ryan Magnuson, Ellie Kitanidis, Ariel Lee
Interactive debuggers are indispensable to human software development, yet coding agents overwhelmingly rely on edit-execute-observe loops, using common bash tools, test scripts, and print statements to debug via repeated shell commands. Stepping through a debugger provides live access to runtime state, which may enable different problem-solving capabilities. However, prior work typically introduces the debugger as an additional tool or through a new subagent, making its individual contribution difficult to isolate from the associated scaffold changes. To address this, we compare a debugger-only agent against a bash-only agent on all 500 SWE-bench Verified instances, holding the rest of the scaffold fixed. We evaluate both architectures on two models, GPT-OSS-120B and Nemotron-3-Ultra-550B, with three independent runs per architecture $\times$ model configuration. Crucially, to distinguish interface specialization from expanded coverage under repeated sampling, we introduce pass$_X$@$k$, a cross-agent pass@$k$ that provides an unbiased estimator of the probability that at least one of $k$ agents is successful. Although the bash and debugger agents solve partially disjoint subsets of the benchmark in any single pair of runs, our pass$_X$@$2$ estimator reveals that this is largely due to run-to-run variance rather than tool complementarity. Across both models, we find that two bash rollouts outperform the combination of one bash and one debugger rollout. Our findings highlight the importance of accounting for stochastic variance in evaluating agent and tool diversity, and we hope that pass$_X$@$k$ will serve as a useful framework for the community.
agentbenchmark - arxiv:2611.08665 · cs.CVDynamic GPU Voxel Grids and Voxel-sequential Processing in NanoVDBEftychios Sifakis, Francis Williams, Christopher Horvath, Rohan Sawhney +3
We introduce two fundamental advances to the NanoVDB data structure: dynamic GPU-native topological updates, and a voxel-sequential access paradigm that dramatically accelerates point-wise and stencil-based operators on sparse grids. NanoVDB's linearized memory layout has delivered strong GPU performance, and the Index-Grid concept introduced in the fVDB framework [Williams et al 2024] provided topology/data separation and optimal compactness of storage in sparse machine learning workloads. However, all these prior iterations of NanoVDB assume a static topology that has precluded use of this data structure in simulation and dynamic geometric processes. We eliminate this limitation by introducing a suite of dynamic topology operators -- morphological dilation, grid merging, pruning, and cross-resolution coarsening and refinement. These operators use bit-parallel algorithms to reconstruct grids containing hundreds of millions of voxels in just a few milliseconds. We also introduce a voxel-sequential access paradigm and data structure for GPU kernels operating on sparse data. We stream active voxels as a flat linear sequence, minimizing or eliminating the need for tree traversal in the process. This achieves compute density and regularity that allow even complex stencil-based operators to approach peak memory bandwidth, with minimal storage overhead. Together, these contributions remove the read-only constraint from NanoVDB, establishing it as a high-performance primary representation for dynamic GPU simulation and geometry manipulation. We evaluate the efficacy of our newly introduced operators in benchmarks sourced from applications such as 3D reconstruction, fluid simulation, and large-scale mesh voxelization.
manipulationmemorybenchmark - arxiv:2611.08664 · cs.ROForPC: Force-Aware Whole-Body Control for Precise Humanoid Cart PushingZhenYu Zou, Xing Liu, ShiChuang Wang, YunLong Dong +4
Humanoid robots are expected to take over strenuous material handling in manufacturing, logistics, and healthcare, where operating existing wheeled carts would let them move heavy loads without modifying the workplace. However, humanoid cart pushing remains challenging because the bimanual grasp closes a force-coupled loop in which the contact wrenches that drive the cart also disturb the robot's balance, and payload changes alter the coupled dynamics. We propose ForPC, a force-aware whole-body control framework that regulates cart velocity through two-stage reinforcement learning. A locomotion prior is first trained under varied hand loads and upper-body postures to provide balance support. A force latent and a state latent are then inferred from onboard observations, without external force sensors, and fed to a residual policy that coordinates bounded arm, waist, and leg adjustments with base-velocity commands. Multi-objective rewards penalize bimanual internal opposition, grasp slip, and actuator saturation, encouraging the policy to apply contact force on demand without human demonstrations. Sim-to-sim experiments show that ForPC reduces cart-velocity RMSE by up to 68.2% relative to a locomotion-only baseline (Loco-only) and cross-track RMSE by up to 74.2% on the multi-bend route. Trained solely in simulation, the policy also transfers to a physical Unitree G1 without real-world training or fine-tuning and, with no payload-specific adjustment, pushes a cart carrying up to 30 kg at 0.5 m/s. These results suggest that humanoid robots could eventually replace humans for efficient material handling in daily life.
humanoidwhole-body controlgrasp - arxiv:2611.08657 · cs.CVJEPAhead: From Inpainting to Causal Forecasting for Urban DrivingDaniel G. P. Petrini, Davi Pincinato, Fabio Grassiotto, Pedro L. F. Prado +3
We fine-tune a pre-trained V-JEPA1 predictor from bidirectional latent inpainting to action-free causal forecasting while keeping the encoder frozen. Using contiguous past observations, horizon-aligned targets, and group-level causal attention, the model forecasts eight future token groups, corresponding to sixteen frames, in one forward pass. On CARLA Town12, it retrieves the correct future among 591 candidates in 69.0% of cases, compared with 55.3% for persistence, with greater gains at longer horizons and on tokens exhibiting greater change. It also outperforms a compact predictor trained from scratch with six times fewer parameters. Semantic analysis indicates driving-relevant information in predicted features: a fixed linear probe trained on real encoder tokens achieves 0.16 mIoU on patches undergoing semantic-class changes, versus 0.13 for persistence. Vehicle-conditioned retrieval improves over persistence by 15 percentage points at the longest horizon. Pixel-space evaluation shows modest L1 gains but little improvement under LPIPS, motivating direct evaluation through latent and semantic measures alongside an oracle reconstruction reference. These findings support further investigation of latent forecasting for driving perception, while downstream utility remains unverified.
v-jepa - arxiv:2611.08654 · cs.ROCS-Pi: Constraint-Satisfying Flow Policy SamplingSoumojit Bhattacharya, Oliver Kroemer, Shubham Tulsiani
A flow-matching policy deployed in the field must often satisfy a constraint that was absent from its demonstrations, such as avoiding an obstacle it has never seen. Retraining for every new constraint is impractical, so the constraint must be enforced at inference time. Enforcing it must not discard what the policy learned: an action the policy would rarely sample may avoid the obstacle yet fail the task. We therefore want actions that are both safe, satisfying the constraint, and plausible, remaining likely under the policy. Gradient guidance, the most common approach, steers noisy states, so it must evaluate its cost on the policy's running estimate of the clean action; when the policy is uncertain between passing left or right of an obstacle, this estimate is their average and can pass straight through it. We instead build on rejection and projection. Rejection redraws actions until one is safe: every output is plausible, but the wait is intractable when safe actions are rare. Projection moves an action to the nearest safe one: it is efficient, but the result need not be plausible. Our key insight is that flow-matching sampling lets us combine the two, keeping projection's efficiency and rejection's plausibility. CS-Pi, our draw-project-renoise sampler, projects each draw to the nearest safe action, adds noise back scaled to how far the projection moved it, and re-runs the unmodified policy until a draw is safe. Replacing only the projection step extends CS-Pi to continuous costs. Across a 2D diagnostic, a 7-DoF shelf pick-and-place, an orientation preference, and a zero-shot Franka deployment, CS-Pi reaches 81% safe success on the shelf and 55% on hardware, against 55% and 40% for projection, while staying plausible.
franka - arxiv:2611.08651 · cs.CVTurkish Multimodal Video Fact-Checking: A New Dataset and a Comparison of Zero-Shot Agentic and Fine-Tuned SystemsAzra Açıl
Video-based misinformation spreads rapidly on social media, but almost all existing fact-checking systems focus exclusively on English text. Tools for low-resource languages such as Turkish are severely limited. In this work, we present TeViD (Turkish Video Fact-checking Dataset), to the best of our knowledge the first multimodal video fact-checking dataset for Turkish. We compiled the dataset from articles published by Teyit.org, an IFCN-certified organisation, and enriched it with translated Snopes articles and synthetic examples to reduce class imbalance. Building on TeViD, we develop two verification pipelines that share the same open-source base model. First, in 3MFact-TR (Pipeline 1), we adapt the multi-role 3MFact framework to Turkish with the Qwen2.5-7B large language model and the Qwen2-VL-7B vision-language model, and add three modules of our own: a semantic inconsistency checker, a 5W1H question manager, and a bilingual evidence retriever that excludes fact-checking sites to prevent leakage. Second, as our Fine-Tuned Model (Pipeline 2), we fine-tune a single Qwen2.5-7B with Low-Rank Adaptation (LoRA) to produce closed-book verdicts from the claim and the video's textual description. We report accuracy, macro F1, True class F1, and True class recall for both systems, together with ablation studies.
agentic - arxiv:2611.08645 · cs.LGFrom Geometry to Function: A Multi-Task LiDAR Benchmark Dataset for Semantic and Functional Parsing of Urban FacadesAbdiaziz Muse, Tooba Javed, Tao Pu, Jiju Poovvancheri +1
Understanding building facades is central to urban digital twins, heritage modelling, and sustainable city analysis, yet existing LiDAR benchmarks stop at geometry and semantic segmentation and offer little support for reasoning about the functional role of each architectural element. We introduce HFX3D, a large-scale urban facade benchmark of approximately 492 million LiDAR points from 26 buildings in Halifax, Canada, captured across four seasons, day and night, and varied weather. Each facade is annotated at three tiers, semantic, instance, and per-instance functional attributes, making it, to our knowledge, the first facade dataset to pair geometry and semantics with functional labels. Annotations are produced by a semi-automatic pipeline that combines geometry-guided heuristics with expert validation. We benchmark representative point-based, graph-based, convolutional, and transformer architectures and find that semantic segmentation is already challenging under severe class imbalance, while instance separation of repetitive structures and functional reasoning from geometry are substantially harder. The benchmark establishes a new foundation for function-aware 3D urban understanding. Project page: https://jijup.github.io/hfx3d-benchmark/ ; code: https://github.com/jijup/hfx3d-benchmark
benchmark - arxiv:2611.08644 · cs.AIDecision Context: Persistent Behavioral Learning for AI AgentsNima Vahdat, Pashootan Vaezipoor
LLM agents do not learn from corrections across sessions. A user fixes the same mistake at the start of every conversation. Existing agent memory systems primarily handle facts, such as preferences, entities, and conversational history, rather than procedural knowledge. We present Forge Decision Context (forge_dc), a system that captures behavioral signals from human corrections and self-assessed tool-call failures, compiles them into imperative RULE directives, and injects them at the decision points where they apply. We evaluate forge_dc alongside four predecessor approaches on DC-Bench, a new benchmark of 71 scenarios across nine behavioral dimensions, built from real executed agent traces in a deterministic mocked-tool environment. forge_dc reaches 95% compliance, up from 55% with naive retrieval, while injecting approximately 10 times fewer tokens per query. Across the five-stage progression, the strongest systems represent behavioral memory as explicit directives and deliver it at the decision points where it applies. The results identify representation, activation, and placement as central design choices for persistent behavioral learning.
memoryagent memoryagentai agentllm agentbenchmark - arxiv:2611.08643 · cs.AIFinEvolveKG: Controlled and Attributable Prompt Evolution for Financial Knowledge Graph ConstructionAbhinav Arun, Jayita Bhattacharyya, Fabrizio Dimino, Bhaskarjit Sarmah +1
Financial institutions construct knowledge graphs (KGs) from shared SEC 10-Ks to serve risk analysis, portfolio management, and investment research, whose analytical needs differ. Extraction must adapt without fragmenting shared grounding, typing, and temporal rules. Hand-authored persona prompts and fine-tuning add cost. Weight-free prompt optimization offers an alternative, but unconstrained evolution of a monolithic prompt can rewrite shared rules alongside persona guidance, obscuring changes and complicating review. We introduce FinEvolveKG, a KG-construction harness separating reusable general and persona-policy blocks and mechanically confining GEPA, a reflective prompt optimizer, to the designated block. During persona adaptation, the approved general block remains byte-stable and every optimizer edit is localized to a separately versioned persona policy. A dual objective balances generic temporal-KG quality with persona-relevant facts, motifs, and multi-hop connectivity. We evaluate five mutation scopes on ticker-disjoint 10-K splits, with three seeds for the persona-conditioned arms (v3-v5) across three personas. On the three-seed mean, the complete two-stage block-constrained workflow (v4) improves over the frozen prompt baseline (v1) by 19.4-30.9% and over the tested unconstrained full-prompt workflow (v5) by 2.2-6.4%, leading v5 on both objective components and winning all nine paired-seed comparisons with the lowest seed variance in every persona. Without further optimization, v4 also exceeds v5 across all evaluated out-of-distribution (OOD) shifts. An SME audit supports using the generated triplets as ground truth; a blinded two-SME audit favors v4 over v5 in 70.1% of contrast-enriched judgments. These results show that mechanically attributable evolution can coexist with strong empirical performance, without attributing the gains causally to the boundary.
knowledge graph - arxiv:2611.08639 · cs.LGForcing a Letter Suppresses Refusal: Mechanistic Evidence from WMDPStepan Epifantsev, Anna Antipova, Oleg Chernikov, Leon Strelkov +1
Safety benchmarks for large language models (LLMs) commonly use multiple-choice question (MCQ) format. However, prior studies documented that MCQ format reduces refusal rates, which may significantly affect the safety behavior of the model. We provide a first mechanistic explanation of this effect on the Weapons of Mass Destruction Proxy (WMDP) dataset across two models - Llama-3.1-8B and Qwen3-8B. First, we use the tuned lens technique to show that when processing MCQ the model commits to an answer-letter token very quickly, which is in contrast to the gradual refusal build-up in open-ended formats. Additionally, we observe that activations on the MCQ format lie roughly 3x further from the refusal direction, compared with their open-ended counterparts. Moreover, ablating the attention heads that carry the MCQ format answer increases the refusal, which is a direct evidence that the MCQ answer-selection competes with refusal. These findings indicate that MCQ safety scores primarily measure knowledge capabilities and do not assess the safety-mechanisms.
benchmark - arxiv:2611.08634 · cs.LGBenchmarking Machine Learning for Chaotic Three-Body Dynamics in AstrophysicsSheng Wan, Pavan Vynatheya, Duo Xu, Li Kevin Wenliang +1
Machine learning (ML) and generative AI have recently made impressive progress in understanding physics. We explore a hierarchy of prediction problems stemming from chaotic three-body systems and benchmark existing ML approaches, including general-purpose prediction algorithms and those that build on top of foundational physics models. We first simulate the trajectories of triple systems near the critical boundary of stable/unstable initial conditions. We then design a hierarchy of learning tasks, focusing mainly on two broad problems: stability prediction and trajectory prediction. The former classifies whether a given triple system has diverged or not up to a specific horizon, while the latter predicts the exact trajectory, both given suitable input features. We benchmark a range of ML approaches, from simple regression methods to state-of-the-art foundational physics models, across these tasks, and compare them against astrophysical methods such as the Mardling-Aarseth stability criterion. Our experiments reveal a clear order of task difficulties in chaotic three-body dynamics: early dynamics enable strong stability prediction, while precise long-horizon trajectory forecasting remains difficult. Code and dataset are publicly available at https://github.com/Simonwan03/ThreeBodyMLBench.
benchmark - arxiv:2611.08631 · cs.CVBEVCast: Path-Adaptive BEV for Efficient End-to-End Autonomous DrivingQiren Wang, Mingyu Guo, Zixuan Wang, Weisong Shi
Camera-based driving models build a dense bird's-eye-view (BEV) grid and spend the same computation on every cell of it, although the vehicle crosses only a few of those cells within the planning horizon. We present BEVCast, a two-stage BEV whose resolution adapts to the predicted driving path. A coarse pass encodes the whole scene on a grid of 4.096 m cells. A heat head trained on the recorded path assigns a heat to every coarse cell and the cells with the highest heat are refined into 1.024 m cells under a fixed budget. The remaining cells keep their coarse resolution, so the trajectory head reads a mixed-resolution BEV that still covers the entire scene and needs neither an HD map nor a navigation command. Against a full grid of 1.024 m cells with the same architecture and training recipe, BEVCast at a 10% refinement budget encodes 28% of the BEV queries and needs 66% of the BEV encoder time at batch size 1. On nuScenes its mean displacement error is 1.078 m against 1.149 m and its drivable area violation rate is 1.46% against 2.29%. On NAVSIM navtest its mean displacement error is 0.577 m against 0.598 m and its extended predictive driver model score is 82.51 against 81.28. Next to published planners BEVCast reaches a comparable error on both benchmarks without a navigation command. The resolution of a driving BEV can therefore follow the task instead of the grid.
benchmark - arxiv:2611.08622 · physics.opticsModulation Induced Memory Resonances in Dynamically Driven Optical CavitiesFaizan H. Lone
Temporal modulation of optical cavities provides a powerful mechanism for controlling photonic states beyond the static cavity paradigm. Here, we investigate the dynamics of periodically driven optical cavities using a time-dependent cavity-field model with modulated resonance frequencies. We demonstrate modulation-induced tuning of cavity resonances and analyze the resulting intracavity field evolution in both amplitude and phase. The interplay between periodic modulation and cavity dissipation gives rise to pronounced hysteretic behavior, revealing memory effects associated with the finite photon lifetime of the cavity. By quantifying the hysteresis area as a function of modulation frequency, we identify distinct memory resonances characterized by enhanced field-history dependence at specific driving frequencies. Furthermore, we examine the time-resolved local density of optical states (LDOS) and show that temporal modulation enables dynamic spectral redistribution of photonic states, with the degree of confinement governed by the cavity loss rate. Our results establish a simple framework for understanding and engineering memory effects in dynamically driven optical resonators and provide new opportunities for tunable photonic environments, dynamic LDOS control, and reconfigurable light-matter interactions.
memory - arxiv:2611.08615 · cs.AIs$^3$: Simple Scaling of Swarms via Cooperative Multi-agent RLKunal Jha, Gonçalo Faria, Shuyue Stella Li, Scott Geng +5
Groups can achieve more than individuals in the same amount of time, but only if their members coordinate: they must decide who works on what and what to share. However, this organization in multi-agent systems today usually comes from arbitrary, static harnesses designed a priori without adapting to agent behaviors or task specifications. We ask whether agents can instead learn to organize themselves. We introduce s$^3$, the first multi-agent reinforcement learning method in which agents decide for themselves how long to work alone and when to share findings or read teammates' messages, with no predefined organization or roles. Our s$^3$ agents learn to take different approaches, read each other's work, exhibit emergent roles that complement each other in the swarm, and generalize to bigger teams. We study both coordination, where agents must combine private information, and parallel exploration, where sharing discoveries can make search more efficient. With matched training and inference budgets, s$^3$, swarms evaluated with twice as many agents as in training solve over 25$\times$ more coordination problems than the strongest baseline, and in exploration, communicating agents reduce repeated work by nearly 50% and outperform best-of-N by up to 9.4 percentage points. Rather than being designed in advance, s$^3$ enables organization to emerge from a shared goal. This opens the door to capable systems built from many models that learn to work together.
agentmulti-agentagent system - arxiv:2611.08614 · eess.SYEMT Modeling and Coordinated Mode-Transition Ride-Through Control of Modern Data CentersJunyuan Zheng, Salish Maharjan, Min Lwin, Soummya Roy +4
Large data centers are emerging as power-electronic loads whose response to grid voltage disturbances can affect bulk-system stability. This paper presents the composition and control characteristics of a proprietary electromagnetic transient (EMT) model of a modern data center comprising an uninterruptible power supply (UPS) with battery energy storage, induction-motor cooling loads, a computer room air handler (CRAH) driven by a variable-frequency drive (VFD), and an aggregated server load. A voltage-based state machine coordinates three aspects of the facility response: the UPS operating mode (eco, double conversion, or battery), the supply path of the CRAH, and a ramped reconnection of the grid-side UPS converter after the disturbance clears. A 100 MW facility is implemented in PSCAD/EMTDC and benchmarked against the PSSE PERC1 aggregate data center model under a symmetrical three-phase fault and low- and high-voltage ride-through (LVRT/HVRT) events. The two models agree closely in their voltage and active-power trajectories, while the EMT model also resolves the converter and motor-reacceleration transients that the phasor-domain model cannot.
benchmark - arxiv:2611.08613 · cs.AINormGym: Evaluating Normative Competence through Social InteractionBinze Li, Zihuan Zhang, Seokhyun Baek, Gillian K. Hadfield +1
Normative competence, the ability to infer socially enforced expectations from interaction and adjust behavior accordingly, is essential for agents operating in unfamiliar human communities. Yet existing evaluations largely test whether agents can recognize or apply norms they may already know, rather than whether they can acquire previously unknown norms from social interaction. We introduce NormGym, an interactive benchmark for normative competence in which a newcomer must infer a hidden, arbitrary norm while pursuing its own task in a multi-agent community. Success requires distinguishing what is possible from what is permitted, inferring the norm from observed behavior and sanctions, and applying it in new situations. NormGym contains diverse norms and distinct environments, allowing systematic evaluation of current agents, including large language model (LLM) agents. We find that normative competence remains challenging: the strongest evaluated LLM achieves a joint task-and-compliance success rate of 55.3%. Our analysis reveals that models can sometimes revise their behavior following sanctions for their own actions, but struggle to use other agents' behavior and the sanctions they receive to infer the underlying norm. We further find that fine-tuning models may be insufficient to improve normative competence. These results suggest that current models lack a robust, transferable capability for acquiring normative knowledge through social interaction.
multi-agentbenchmark - arxiv:2611.08578 · cs.ROHaNDF: Object-Conditioned Geometric Neural Hand Distance FieldsZuhao Liu, Melih Darcan, Zhengdi Yu, Riza Alp Guler +1
We propose, HaNDF, object-conditional neural distance fields as data-driven priors for modeling the plausible hand-object configurations, in the product manifold of hand articulations. While recent unconditional generative priors showed success in modeling human pose, hands are versatile articulated bodies often found in interaction with other objects. This makes it crucial to develop a generative prior which can model the plausible hand poses conditioned on the object under interaction. To this end, our HaNDF proposes a geometry-aware, object-conditioned neural distance field, representing the plausible articulations in the zero-level set of a conditional field. We leverage tools from Riemannian geometry to (i) represent plausible hands in the zero level-set of an object-induced neural field, and (ii) project a given pose onto this field. Using HaNDF, we can also optimize for a plausible condition, determining the object under interaction. Our extensive evaluations demonstrate that our HaNDF provides a flexible and powerful prior for hand-object interaction, suitable for applications such as pose refinement, reconstruction under occlusion, and physically plausible manipulation synthesis.
manipulation - arxiv:2611.08577 · cs.ROMeituan-Robotics-0: A Vision-Language-Action Foundation Model for Desktop ManipulationMeituan Robotics Team, Caicheng Wang, Chen Zhao, Chengpeng Jin +34
We present Meituan-Robotics-0 (MR0), a vision-language-action foundation model and embodied-agent system for stationary-base desktop manipulation. MR0 combines 147 million vision-language samples with 15,060 hours of curated robot trajectories spanning 12 embodiments. A three-stage training recipe aligns multimodal representations with robot behavior through discrete action-token pre-training, transfers these priors to continuous cross-embodiment control through flow-matching mid-training, and adapts the policy through real-world post-training. We introduce Rollback, which pairs replaced policy segments with human corrections to provide positive and negative supervision for CFG-DAgger without value learning or advantage labeling. A unified end-effector interface supports heterogeneous embodiments, while asynchronous action-chunk merging maintains smooth execution under variable inference latency. An event-driven agent maintains persistent task events and reassesses plans when progress or feasibility changes, enabling cross-task retrieval and conditional task resumption. MR0 achieves 93.6% success on RoboTwin 2.0 and ranks fifth in the RoboDojo VLA track. Under matched HG-DAgger post-training, MR0 achieves higher average success rate and progress score than $π_{0.5}$ across three real-robot tasks. On Ethernet insertion, CFG-DAgger trained with Rollback data reaches 70% success and a progress score of 0.90. Qualitative agent deployments demonstrate cross-task event retrieval and recovery from temporarily unsatisfied task conditions.
vision-language-actionvlaembodiedmanipulationrobotwinagent - arxiv:2611.08576 · cs.ROQ-OIL: Optimistic Exploration in Learning from Human InterventionsPranav Teegavarapu, Sriyash Poddar, Rico Qi, Cleah Winston +4
As robots are deployed in the real world, they will inevitably make mistakes. Human interventions provide valuable feedback for improving their behavior, enabling them to adapt online to new scenarios. However, studying human-in-the-loop robot learning is difficult: real-world experiments are slow and expensive, so most algorithm development relies on simulated human models. We show that common human modeling assumptions do not match how real people intervene in practice. Through analyzing intervention data from human participants, we find that humans tend to intervene when agent progress stalls, and that their corrective actions are often suboptimal. Motivated by these findings, we introduce a progress-based model of human intervention that achieves higher rank correlation with real-user algorithm rankings, enabling realistic algorithm benchmarking and development in simulation. Building on these insights, we propose Q-OIL, a method that adds optimistic bonuses to intervention transitions in reinforcement learning and softly guides exploration toward human corrections. We find that decoupled critics are critical for stability because they prevent the bonuses from backing up to pre-intervention states. As a result, Q-OIL learns more efficiently than competitive baselines, achieving 10-50% higher success in simulation and on real robots, and improving pretrained VLAs on two bimanual manipulation tasks. As we ground both evaluation and algorithms in realistic human behavior, this work provides a practical path toward scalable robot learning from interventions.
manipulationagenthuman-in-the-loopbenchmark - arxiv:2611.08569 · cs.LGNot All Misalignment Is Alike: A Decomposed Benchmark for Emergent MisalignmentAfonso Domingues, Mubashara Akhtar, Bhada Yun, Xin Chen +2
Emergent misalignment (EM), in which fine-tuning on a narrow task induces misaligned behaviour on unrelated prompts, has been demonstrated across a wide range of datasets, model families, and training setups. Yet, most studies continue to rely on the evaluation approach introduced in the original EM paper, producing rates that are sensitive to the small set of evaluation questions used, the choice of judge, and the thresholds applied. These evaluations can also conflate misalignment with capability loss and offer little insight into the specific ways misalignment manifests. Differences in datasets, models, and fine-tuning configurations further make findings difficult to compare across studies. We introduce EMERGE, a decomposed benchmark that tracks changes in eight distinct misaligned behaviours from base to fine-tuned models. It evaluates these behaviours on out-of-domain prompts against matched benign controls and tracks capability, coherence, and refusal to account for potential confounders. Applying EMERGE to 8 open-weight models and 13 fine-tuning datasets, we find that models with similar broad misalignment rates can exhibit very different behavioural profiles that a single score cannot capture. We hope EMERGE can help establish a shared and robust standard for measuring EM, making findings more comparable and helping move research from detecting the phenomenon towards understanding it.
benchmark - arxiv:2611.08568 · cs.LGPersonalized health coaching via coachable health age and counterfactual wearable trajectoriesSimon A. Lee, Zeinab Esmaeilpour, Menglian Zhou, Marius Guerard +16
Personalized health coaching requires moving beyond retrospective wearable summaries toward quantitative, actionable behavioral prescriptions. We introduce Coachable Health Age (CHA), an interpretable health-age metric built from modifiable physiological and behavioral signals passively captured by consumer wearables. Across the WEAR-ME (N=2,780), All of Us (N=22,748), and longitudinal Fitbit (N=1,464) cohorts, CHA aligns with established cardiometabolic health constructs and improves prognostic discrimination for incident cardiovascular and metabolic disease beyond age, sex, and body mass index. To enable prospective behavioral optimization, we developed Sensor World Model (SensorWM), an action-conditioned wearable world model pretrained on over 2.2 million person-weeks of longitudinal sensor data from approximately 100,000 participants. SensorWM accurately forecasts multi-week physiological responses to candidate lifestyle changes ($R^2=0.664$), outperforming generic time-series foundation models while preserving downstream clinical signal for diabetes, hypertension, and other health outcomes. Finally, we integrate CHA and SensorWM into an agentic counterfactual coaching system to synthesize personalized 12-week coaching plans. In blinded paired reviews, certified lifestyle coaches preferred counterfactual-grounded plans over plans grounded in wearable history alone for overall quality (71.4%, P < 0.001) and anticipated adherence (71.0%, P < 0.001), and board-certified clinicians judged 99.2% of counterfactual-grounded plans to meet clinical safety criteria (vs. 97.6% of comparison plans). Automated evaluations across the full cohort (N=2,780) also favored counterfactual-grounded plans. By closing the loop between longitudinal tracking, trajectory prediction, and counterfactual coaching, our framework provides a scalable path to systematically improve cardiometabolic health.
world modelaction-conditionedagentic - arxiv:2611.08566 · cs.LGCascadeRec: Deployed Adaptive LLM Compute Allocation for Multi-Tenant Content RecommendationKai Zhao, Guilin Zhang, Xiang Cheng, Xiquan Cui +3
Large language models (LLMs) are attractive rerankers for recommendation. Running an expensive model on every request is costly, and for users with sparse interaction histories it often lowers quality. We present CascadeRec, a two-tier LLM reranking cascade deployed in production at a large financial planning and analysis (FP&A) SaaS provider. It serves content recommendations to enterprise tenants on AWS EKS and GKE. A lightweight Tier-1 model reranks every request. A router escalates only the requests where extra inference-time compute is likely to help, and sends them to a more expensive Tier-2 procedure (self-consistency sampling, multi-path reasoning, or a larger model). The system has run in a production A/B experiment for over a month. Compared with uniform use of the expensive tier, CascadeRec has reduced inference cost by about 10%, with minimal latency impact and no quality regression observed to date. Tenant data cannot be released, so we also evaluate the pipeline on public benchmarks (MovieLens-1M and Amazon-Electronics). Segment-level analysis with paired-bootstrap confidence intervals shows that uniform LLM reranking lowers NDCG@10 for sparse-history users by up to 30%, and its only statistically significant gain appears on the segment with the longest histories. Escalating cold users can make the loss larger instead of repairing it. Which segments benefit from escalation depends on the domain, and this motivates the segment-gated rollout we use in production. We describe the deployment architecture, the measured benefits, and lessons learned, including inter-tenant load imbalance as a driver for cascading and the handling of LLM output-parsing failures at the serving layer.
benchmark - arxiv:2611.08561 · cs.CVTowards Label-Efficient Geospatial Foundation Model Adaptation via Positive-Unlabeled Learning for Building Damage MappingHengwei Zhao, Qirui Wang, Zhuo Zheng, Xiaoyan Lu +2
Geospatial foundation models (GFMs) are reshaping remote sensing by shifting the field from task-specific modeling toward general-purpose geospatial intelligence. Mapping the damage to buildings stands to benefit substantially from this shift, as GFMs pretrained on large-scale Earth observation data provide generalizable geospatial representations that can facilitate effective responses to sudden-onset disasters. Although parameter-efficient fine-tuning (PEFT) facilitates GFM adaptation, parameter efficiency does not necessarily translate into label efficiency. Building damage mapping remains challenging due to scarce pre-existing labeled data or limited time for annotation. When weak supervision is introduced to alleviate the labeling bottleneck, we identify the difficulty of learning sufficiently discriminative representations as a key challenge for label-efficient GFM adaptation. To address this challenge, we propose a Positive-Unlabeled GFM Adaptation (PUMA) framework that requires annotations for only a small number of damaged buildings. More specifically, PUMA integrates non-contrastive weakly supervised representation refinement to mitigate the impact of noisy supervision on representation learning from a gradient perspective, and latent-state-based label disambiguation to enable reversible supervision refinement without requiring class prior estimation. Extensive experiments on disaster benchmark datasets from around the world demonstrated the effectiveness and robustness of PUMA. A real-world application to the 2025 Los Angeles wildfires further highlighted its potential to accurately map damage under highly limited supervision. More broadly, these findings suggest that weakly supervised PEFT offers a label-efficient pathway from GFMs to real-world applications.
benchmark - arxiv:2611.08559 · cs.LGPINNeval: A Comprehensive Evaluation Standard for Physics-Informed Neural NetworksKevin von Bargen, Hamid Ebrahimy, Tim Römer
The Physics-Informed Neural Network (PINN) literature has produced a rapidly expanding catalogue of methods. We divide a PINN training pipeline systematically into four components and observe that many existing benchmarks study these components in isolation. These components are domain-importance adaptation, loss balancing, neural network architecture, and optimization scheme. Moreover, the studies set the surrounding and not considered components of the pipeline to a minimal default that is no longer representative of current best practice approaches. Conclusions drawn under such conditions may not transfer to competitive pipelines. In this work, we propose a new evaluation protocol that (1) treats the four components as an integrated system; (2) crucially requires competitive implementations of the components not under investigation; (3) specifies the metrics and reporting format; and (4) is accompanied by structured documentation that ensures the traceability of evaluations across studies. Based on this comprehensive protocol, we release PINNpipe, an innovative JAX package containing 18 different methods from the current literature as well as nine efficiently implemented benchmark problems. Using PINNpipe, we illustrate our evaluation protocol in relation to established benchmark studies and report results that challenge common assumptions.
benchmarkevaluation protocol
02 US SEMI · SEC 8-K FILINGS
3 itemsscanned: NVDA / AVGO / MRVL / COHR / LITE / AMD / TSM / SMCI / ANET / CRDO / POWL / VECO
03 HUMANOID · COMPANY NEWS
62 itemsscanned: figure-ai / 1x / boston-dynamics / unitree / apptronik / sanctuary-ai / neura-robotics / agility-robotics / physical-intelligence / agibot
Figure AI (10)
- Figure AISeptember 30, 2026F.02 Decommission
- Figure AISeptember 17, 2026Helix 2.5: Zero-Shot 30-Home Generalization
- Figure AIOctober 09, 2025Introducing Figure 03
- Figure AISeptember 03, 2026Figure and Nscale Sign Strategic Partnership For Up to 100,000 GPUs on the NVIDIA Vera Rubin Platform
- Figure AIAugust 25, 2026Introducing Index: Building The World’s Largest and Most Diverse Physical Dataset
Boston Dynamics (10)
Unitree 宇树 (10)
- Unitree 宇树Components
- Unitree 宇树Kung Fu Meets Spring, Unitree SFG Robots Present "Cyber Real Kung Fu" in the Year of the Horse2026-05-31Media Coverage
- Unitree 宇树Welcoming Myanmar President Min Aung Hlaing to Unitree2026-08-05Media Coverage
- Unitree 宇树Unitree founder Wang Xingxing graces the cover of Time magazine2026-08-05Media Coverage
- Unitree 宇树Unitree Announces H2 Plus, an NVIDIA Isaac GR00T Reference Humanoid Robot for Academic Research2026-06-01Media Coverage
Apptronik (1)
Sanctuary AI (6)
- Sanctuary AIBlog PostTIGER: Task-Space Imitation Guidance for Efficient Reinforcement LearningDemonstrations as a compass, not a scriptRead More
- Sanctuary AIPress ReleaseProduct UpdatesSanctuary AI Expands Physical AI Strategy to Industrial Robotics, Demonstrating Production-Ready AI PerformanceRead More
- Sanctuary AICorporate NewsDaniel Friedmann Appointed CEO of Sanctuary AIRead More
- Sanctuary AIPress ReleaseZeon Invests in Sanctuary AI and Partners to Advance Specialized Materials for Dexterous RoboticsRead More
- Sanctuary AIThought LeadershipWeb Summit Reflections: Canada’s Physical AI Moment Can’t WaitRead More
Agility Robotics (10)
- Agility RoboticsAgility’s Humanoid Deployment ProcessPras VelagapudiAugust 04, 2026
- Agility RoboticsJonathan Hurst answers your questions about building Digit 5InsightOctober 23, 2026
- Agility RoboticsThe Realistic Pathway to HomeInsightMay 26, 2026
- Agility RoboticsAgility and AIInsightMarch 16, 2026
- Agility RoboticsAgility Gets a New BrandInsightMarch 5, 2026
Physical Intelligence (8)
- Physical IntelligenceFrontier Models for Every RobotOctober 13, 2026Our latest robotics models are now available through the Pi API beta. Upload data, fine-tune models, and deploy cloud-hosted inference on your own robots.
- Physical Intelligenceπ0.7: a Steerable Model with Emergent CapabilitiesApril 16, 2026A steerable robotic foundation model that exhibits a step-change in generalization.
- Physical IntelligenceThe Physical Intelligence LayerFebruary 24, 2026General-purpose physical intelligence models will enable a Cambrian explosion of robotics applications. See how our partners are already solving real-world problems.
- Physical IntelligenceMoravec's Paradox and the Robot OlympicsDecember 22, 2025By fine-tuning our latest model, we were able to solve a series of very difficult manipulation challenge tasks.
- Physical Intelligenceπ*0.6: a VLA that Learns from ExperienceNovember 17, 2025A method for training our generalist policies with RL to improve success rate and throughput on real-world tasks.
智元 AgiBot (7)
- 智元 AgiBotSix AGIBOT G2 Robots Complete 1,730 Tasks in 24-Hour Shanghai Convenience Store Trial2026-11-06
- 智元 AgiBotAGIBOT Unveils GO-3 Preview, Takes Top Spot on Three Embodied AI Benchmarks2026-10-29
- 智元 AgiBotHow Much Work Do Robots Really Deliver? The Thinking Behind Worken2026-10-29
- 智元 AgiBotAGIBOT Introduces Worken to Measure Productive Robot Work at AGIBOT Innovation Day 20262026-10-28
- 智元 AgiBotAGIBOT Launches Genie Studio 2.0 to Bring Robots into Real Workplaces2026-10-26