Physical AI Brief

Post-training attention sparsification reduces the quadratic cumulative attention cost of pretrained Transformers by selecting a small set of context units (tokens or blocks) for each query. Existing trainable methods usually use a lightweight selector to score context units, followed by hard Top-K selection that blocks gradients from the language modeling loss. Consequently, these methods commonly distill layer-wise dense attention distributions. Although this encourages the selector to rank context units by dense attention weights in the original model, the ranking is not directly aligned with their impact on predictions under a fixed attention budget (i.e., the number of attended context units per query), potentially wasting the limited budget on less useful units. To address this misalignment, we propose Simple Attention Sparsification (SAS), a gated sparse attention mechanism that optimizes context ranking end-to-end with the language modeling loss. The key idea is to inject the selector's continuous scores into attention logits during training, allowing the loss to update the selector through standard backpropagation. We identify several choices crucial for this simple design to work well in practice: placing the gate inside the attention softmax in log form, using normalized softmax gates to calibrate historical context against the always-retained current block, and preserving continuous selector scores so the model learns relative priorities rather than only hard selections. To support long-sequence training, we implement a memory-efficient Triton kernel that integrates SAS into FlashAttention-style computation. Across reasoning, long-context understanding, and agentic tasks, SAS consistently outperforms trainable sparse attention baselines across attention budgets, with especially large gains under tight budgets, demonstrating more effective context ranking for downstream tasks.

Frontier language models are more aggressively using subquadratic attention to reduce the memory footprint and compute requirements during inference while still delivering frontier accuracy. While existing systems make dense attention-centric disaggregated serving decisions, we show that disaggregating inference around the unique arithmetic intensity and memory footprint of subquadratic attention LLMs can achieve significant throughput and energy efficiency gains on emerging DRAM-based and SRAM-only heterogeneous systems. We introduce SQD (SubQuadratic Disaggregation), a fine-grained heterogeneous disaggregation scheme that splits decode by quadratic and subquadratic attention rather than by operator type, and that applies across subquadratic attention variants. For sparse attention LLMs, we disaggregate decode into top-k selection, which must index through the full KV, and top-k attention plus FFN, which have static memory footprints. For linear and sliding-window attention LLMs, we disaggregate decode into dense attention layers and subquadratic attention layers plus FFN. In an adjusted 8xB200 heterogeneous system proxy, we observe average tokens/J improvements of 53% on GLM 5.2, 31% on Nemotron 3 Ultra, and 56% on Gemma 4 31B over the strongest GPU-only baselines. In an analytical model of a Rubin plus LPX system with fixed power budgets, we observe 1.2x to 1.5x tighter achievable latencies and up to 3.6x higher throughput over the best baseline of attention-FFN disaggregation. Our experiments also reveal architectural insights on chip and interconnect provisioning for next-generation heterogeneous systems serving subquadratic attention.

Accurate prediction of equity returns remains a major challenge in computational finance due to the non-stationary, nonlinear, and low signal-to-noise ratio nature of financial time series. This paper proposes a hybrid two-stage architecture that combines a long short-term memory (LSTM) network with an XGBoost gradient-boosted regressor for multi-horizon stock return prediction across a diversified panel of 14 U.S. equities spanning six industry sectors. The LSTM component, comprising two stacked layers with 64 hidden units, processes 60-day sliding windows of five sequential market features to produce 64-dimensional temporal embeddings that encode learned sequential market dynamics. These embeddings are concatenated with 14 hand-crafted technical indicators to form a 78-dimensional hybrid feature vector, which is subsequently passed to an XGBoost regressor tuned via 3-fold cross-validation grid search. The framework is trained on a multi-stock pooled corpus using strict chronological splits and per-stock MinMaxScaling to prevent look-ahead bias, and evaluated across four prediction horizons of 30, 90, 252, and 365 trading days. Experimental results demonstrate that the hybrid model achieves a test RMSE of 0.0949 on the 30-day horizon, roughly one-third that of the standalone LSTM baseline, while marginally matching or surpassing the XGBoost-Only baseline across the majority of stocks. Directional accuracy rises with horizon length, reaching 97.6% at 365 days; we show, however, that this largely tracks the high base rate of positive long-horizon returns in the sample, and we therefore benchmark directional accuracy against a naive always-positive predictor and treat the above-base-rate gap at short horizons as the more informative signal. A composite investment scoring framework derived from multi-horizon predictions is further proposed to support portfolio ranking and decision support.

Full-duplex evaluation often emphasizes whether an agent keeps speaking or stops. That binary cannot express a third response humans use routinely: continuing to speak while incorporating what the listener just contributed. The contribution may be a missing word, a correction or a clarification. We introduce Duplex Cue, an evaluation of this \emph{in-turn adaptation} in full-duplex voice agents. Duplex Cue separates listener intent (backchannel, collaboration, or interruption) from speaker behavior: continuing unchanged, adapting within the turn, or yielding. Adaptation includes acknowledgment as well as content revision. In a single-model case study using 300 human-confirmed cues from unscripted English conversations, we compare recorded human responses with PersonaPlex continuations generated while replaying the listener's audio. We retain 208 pairs with the ongoing speaker active at cue onset and a scorable response in each condition. On the 66 collaborative pairs, recorded speakers adapt in 68.2\% of cases, compared with 34.8\% for PersonaPlex. The model otherwise continues unchanged (42.4\%) or yields (22.7\%). These findings show why evaluating natural voice interaction requires measuring how an agent responds to a listener's contribution as well as whether it keeps speaking.

Agentic systems offer a promising way to automate embodied benchmark construction, but existing approaches typically cover isolated stages or remain specialized to predefined environments and task families. More importantly, multi-step construction produces dependent intermediate artifacts that are often passed downstream without artifact-specific verification, allowing local defects to propagate into the final benchmark. We present Embodied-BenchForge, an agentic framework that transforms user-specified evaluation intents into complete embodied benchmark artifacts. It formulates construction as Closed-Loop Benchmark Synthesis, integrating forward artifact synthesis with backward verification and repair. Skill-Orchestrated Artifact Synthesis composes typed and reusable skills into executable workflows, while an artifact dependency graph records intermediate outputs and their dependencies. Requirement-Guided Verification and Repair applies artifact-specific contracts throughout construction and uses provenance to trigger local re-execution or upstream rollback when verification fails. Embodied-BenchForge constructs six benchmarks covering diverse embodied scenarios in the Offline EQA Track, together with one interactive benchmark containing 220 executable tasks in the Interactive Embodied Track. Evaluations of representative MLLMs and embodied agents show that the benchmarks distinguish model capabilities in both observation-based understanding and closed-loop execution. Quality assessment and ablations validate benchmark quality and the effectiveness of verification and repair, while repair and skill-reuse analyses demonstrate efficient localized recovery and cross-benchmark reusability.

Conversational voice agents have advanced significantly, offering increasingly natural human-machine interactions through both cascaded and end-to-end architectures. However, while recent benchmarks extensively evaluate dyadic interactions and passive audio comprehension, they largely overlook a prevalent real-world scenario: multi-party conversations. Evaluating agents in these settings is fundamentally more challenging than in dyadic interactions due to the exponentially greater conversational complexity. For voice agents to integrate seamlessly into human group dynamics, they must not only generate contextually appropriate responses but also demonstrate a nuanced understanding of open turn-taking. To address this gap, we introduce Multiparty Bench (MP-Bench), the first benchmark specifically designed to objectively evaluate conversational speech systems as active participants within multi-party contexts. MP-Bench assesses agent behavior along two primary dimensions: turn-taking awareness and response appropriateness. Additionally, we incorporate comprehension-based question-answering tasks as a complementary evaluation. By benchmarking 12 voice agents, we find that real-time voice agents stay at or below 22% on multiparty comprehension and remain near chance on multiparty turn-taking, exposing an open challenge for real-time voice agents under multiparty scenario.

Recent breakthroughs in LLM-based systems and their abilities in problem solving and coding have allowed progress in the AI for Science paradigm, potentially replacing human roles in machine learning (ML) research. However, while several frameworks of fully autonomous end-to-end ML research have been proposed, successful implementations of them are often limited to problems with narrow search spaces, like language modeling or biomedical ML benchmarks. In this paper, we explore how autonomous research can be adapted to solve open-ended, industry-grade ML problems, by considering a case study: telecom ticket retrieval, an open-ended task with degrees of freedom in representation, architecture, and training data generation. We discover that autonomous research for open-ended problems with commercial and open-source agents shows both promise and limitations: while autonomous research can excel in narrow hyperparameter optimization, it lacks human-like intuition and creativity and requires operational overhead. Even with minimal human supervision, autonomous research can reach $90\%$ of state-of-the-art performance (0.34 vs. 0.38 Recall@1) in a much shorter time period (10 weeks vs. 10 months of human work) at a modest cost (up to \$200 per Cursor campaign). Our empirical evidence recommends that human researchers and autonomous research frameworks work together for best results in ML research.

Fine-grained control of language model behaviors (e.g., steering) is among the more actionable outcomes of interpretability research. For binary concepts such as refusal, a single direction in activation space often suffices for steering. However, many concepts are not binary: Animals and Countries contain many subcategories, each with multiple instances. For these concepts, the search space over possible representation geometries is far larger than for binary concepts; it is thus not clear what geometries are most appropriate, nor what methods are most effective at recovering them. In this work, we introduce MAxBench, a geometry-agnostic evaluation framework for multinomial concept representations based on sampling from the recovered concept representation. We use MAxBench to compare 10 localization methods (covering 5 geometry types) across 6 concepts and 4 models. Using this framework, we find that (i) affine subspaces steer more reliably and have greater recall than rank-one or linear subspaces; (ii) much of this advantage is due to better non-zero offsets rather than the choice of bases; (iii) manifold steering is competitive with the best methods when applicable; and (iv) no method consistently outperforms prompting, in alignment with prior findings on binary concepts. These findings underscore the importance of expanding the scope of interpretability research and meta-evaluation to concepts with more varied structure.

Diffusion Language Models (DLMs) offer promising parallel generation capabilities but lag behind autoregressive models in complex reasoning and tool-use tasks. While Reinforcement Learning (RL) has recently been applied to enhance DLMs, standard RL approaches suffer from an exploration bottleneck. To address this, we inject reasoning priors from a stronger teacher model to guide RL exploration. In this paper, we introduce CanvasAnneal, a curriculum-guided diffusion RL framework. During the initial RL phase, we warm-start exploration by injecting teacher-generated reasoning traces into the initial diffusion canvas. As training progresses, we gradually remove this guidance and require the model to generate more of the reasoning trajectory independently. Across mathematical reasoning and tool-use benchmarks, CanvasAnneal improves over standard diffu-GRPO on MATH500, Countdown, and Tau2 and substantially accelerates reward improvement on several tasks, while gains are task-dependent. Our results suggest that structured training-time guidance can alleviate exploration bottlenecks in diffusion RL and speed up convergence on harder tasks.

Reinforcement learning (RL) has become central to post-training of large language models. Recent advances in RL for Mixture-of-Experts (MoE) models have primarily focused on improving optimization stability and training efficiency, while treating the expert selection as a fixed component. Since routing determines the sparse computation paths that induce output distributions, expert selection offers an additional source of rollout diversity. Through empirical analysis, we find that perturbing expert routing effectively alters model output and increases rollout diversity, which is similar to increasing the decoding temperature. However, direct perturbation can activate unsuitable experts and substantially degrade rollout quality. Motivated by these observations, we introduce Expert-Space Exploration Reinforcement Learning (ESRL), an architecture-aware framework that explicitly explores the expert-routing space of MoE models. ESRL preserves high-confidence experts as anchors, and restricts stochastic routing to a plausible candidate pool, thereby retaining reliable computation paths. The perturbation strength is further adapted according to router entropy to avoid over-perturbation. To mitigate the routing mismatch introduced by perturbation, ESRL records the expert paths used during rollout and replays them during policy optimization. Experiments demonstrate that ESRL achieves the best performance across MoE backbones with top-K, top-1, and shared-expert routing, as well as across mathematics, science, and code tasks without additional sampling or computational cost. Specifically, ESRL on Qwen3-30B-A3B achieves the best among all compared methods, improving average Pass@1 and Pass@8 over GRPO by 3.2 and 4.5 percentage points, respectively. Further analyses of expert utilization and training dynamics provide insights into how exploiting MoE-specific routing structure benefits RL training.

Visual goal and dynamics prediction can provide language-conditioned robot policies with both a target outcome and a representation of action-dependent scene changes. We bring these predictions into action generation and selection through a shared trajectory model. Dynin-Robotics implements this formulation on Dynin-Omni, an omnimodal masked-diffusion backbone, representing language, visual observations, goals, and actions as discrete tokens. By varying conditioning and target spans, the same model learns action prediction, action-conditioned next-observation prediction, terminal goal-state prediction, and trajectory-to-instruction reconstruction. These interfaces support test-time scaling through goal prediction, action-candidate evaluation, and joint refinement of action and future-state predictions. We continually pretrain the model on approximately 1.33 million trajectories from 48 Open X-Embodiment datasets and adapt it separately to downstream domains. On two VLABench tasks, robot pretraining improves adaptation within a fixed Stage-2 step budget, and the full objective mixture improves shifted-instruction success over Policy-only post-training under the same coupled decoder. Combining goal guidance with joint action-next-state denoising further improves shifted-instruction success over action-only decoding; the benefit depends on how the predictions are composed. Dynin-Robotics achieves competitive performance on LIBERO and zero-shot LIBERO-Plus, together with a 78.4% average success rate across four manipulation conditions on a Franka Research 3 robot. An optimized block-parallel implementation accelerates model-side action decoding by up to 29.2x relative to the base implementation under the reported profiling setup. These results support shared trajectory modeling as a common interface for learning complementary robot objectives and composing their predictions during control.

Symmetries play a central role in reducing the complexity of reinforcement learning problems, yet most existing approaches rely on fixed group actions or predefined state abstractions. Classical reinforcement learning algorithms typically assume a globally structured Markov decision process with uniformly applicable actions and transitions, an assumption that limits their ability to exploit modularity and local, context-dependent regularities present in many realistic environments. We propose a reinforcement learning framework using groupoids to capture local, state-dependent symmetries and support the dy- namic discovery of equivalence structures during interaction. The agent maintains orbit representatives together with transporters that map raw states to canonical forms, enabling learning and decision-making to be performed in a symmetry-reduced space while preserving local distinctions. Empirical results demonstrate that the proposed groupoid-based approach improves sample efficiency and convergence in dense and large-scale environments exhibiting strong partial symmetries, yielding substantial performance gains over standard Q-learning. These findings show that dynamically exploiting local symmetry provides a practical and mathematically principled route to scalable and generalisable reinforcement learning.

As a key model compression technique, knowledge distillation aims to transfer knowledge from a high-capacity teacher model to a lightweight student model for enhancing the latter's performance. In this work, we reviewed the feature knowledge distillation for 3D-CNNs and observed that most feature distillation methods in video analysis are simple adaptations of those used in image analysis, often neglecting the differences of video features in the temporal dimension. To address this issue, we proposed Label-Guided Knowledge Distillation (LGKD) to guide the distillation of student model features using ground truth labels. Our method entails two components: sample-wise distillation and class-wise distillation, enabling the student model to learn feature representation of the teacher model at two levels. Sample-wise distillation utilizes label information and the teacher's probability distribution to guide the learning of features that significantly impact temporal accuracy while mitigating noise. Meanwhile, class-wise feature distillation employs a prototype network to further capture the relational knowledge among samples within the same category, enhancing the student's ability to learn higher-dimensional semantic information and improving model generalization. To demonstrate the effectiveness and superiority of our method, we conducted comprehensive experiments on two benchmark action recognition datasets, UCF101 and HMDB51, achieving competitive results.

This work enhances global path planning via a pure-pursuit controller with multi-model kinematic switching that sustains plan fidelity across diverse terrains. The system includes a traversability graph for terrain analysis, a Heading-Aware A* algorithm for generating feasible paths, and a multi-model Pure Pursuit controller for dynamic tracking. A core innovation is adaptive kinematic modeling, enabling real-time switching between kinematic models based on terrain features and robot states. This adaptability optimizes path efficiency and energy use in challenging scenarios. We validate the approach in simulation on different platforms, namely the Artaban quadruped and the X3 quadrotor drone, showcasing improved performance, robustness, and adaptability over standard baselines.

Optimal transport (OT) compares distributions and aligns datasets in machine learning, yet unregularized discrete OT requires a linear program with quadratically many transport variables. We propose HELLO, a hierarchical solver that casts large-scale discrete OT as edge localization and uses dual potentials to guide both coarse-to-fine initialization and within-level refinement. Initialization propagates coarse dual potentials across a recursive subsampling hierarchy to assign candidate edges. Refinement then iteratively inserts the largest dual violators in each row and column until the relative KKT residual meets a prescribed tolerance, while budgeted pruning ensures linear memory complexity. For exact-arithmetic refinement, we prove finite termination at a global optimum under a symbolic lexicographic rule. At the million-point scale, HELLO attains lower transport objectives with order-of-magnitude runtime improvements over strong baselines across feature dimensions from single digits to thousands. It further scales to 1.28 million samples per marginal in 8192 dimensions on a single H100, using 41.6 GiB peak GPU memory while satisfying a full relative KKT residual below $10^{-6}$. Beyond standard discrete OT, the framework supports general pairwise costs and serves as a scalable balanced-OT oracle for semi-discrete OT, Gromov--Wasserstein, unbalanced OT, and OT-based Flow Matching.

Low reported scores on leading physics benchmarks, including those featured in the Artificial Analysis Intelligence Index (2026), suggest that frontier language models still struggle with advanced physics, a demanding test of their scientific reasoning and quantitative problem-solving abilities. Yet this impression does not always align with domain experts' experiences using these models in their work. We revisit these reported findings by evaluating frontier models on six widely used physics benchmarks and auditing them with experts, focusing on text-only problems with verifiable final answers. For each subfield of physics, faculty and graduate researchers with relevant expertise carefully review problem statements, reference solutions, and model responses to distinguish genuine model errors from grader errors, incorrect reference solutions, and ambiguous or underspecified questions. Most audited cases initially evaluated as incorrect reflect these benchmarking issues rather than errors in the models' physics reasoning. We then ask experts to address these benchmarking issues by correcting erroneous reference solutions and repairing or excluding flawed questions. We find that GPT-5.6-Sol's measured mean@4 rises from 47.3% to 78.7% on HLE-Physics and from 61.0% to 87.2% on CMT-Benchmark, while its corrected pass@4 reaches 94.4% on the 54 retained CritPt challenges. Corrected scores are computed on the retained evaluation subsets following expert review. Scores on the audited subsets of UGPhysics, PRISM-Physics, and PHYBench also rise substantially after correction. These findings suggest that current benchmarks substantially understate frontier models' ability to solve well-posed physics problems. Near-saturation on these closed-ended tasks highlights the need for more demanding, expert-validated evaluations.

Large language models (LLMs) have achieved strong performance on a wide range of natural language tasks, and recent benchmarks suggest that they are increasingly adept at multi-hop reasoning. However, these benchmarks are typically short-horizon, requiring only a small number of retrieval or inference steps, and provide limited evidence of reliability on real-world tasks that involve following manuals spanning hundreds of pages with complex, interdependent guidelines. In this paper, we introduce Tasks over Application Manuals (TAM), a benchmark for evaluating long-horizon procedural reasoning. We construct TAM by curating real-world tasks from two domains: ICD-10-CM clinical coding (mapping medical conditions to diagnostic codes) and U.S. federal sentencing (computing crime sentencing guideline outcomes, specifically offense levels), with human-validated labels. Each task requires following an authoritative manual with tens of thousands of rules and executing a sequence of interdependent steps across different sections to produce an exact answer. We evaluate general-purpose prompting approaches, including retrieval-augmented generation, ReAct-style prompting, and an agent-harness baseline on GPT-5, and find that the best exact-match performance remains extremely low: 1% on ICD-10-CM coding and 15.5% on sentencing tasks. These results show that current benchmarks may overestimate LLM reasoning ability and miss a key challenge: reliably following long, rule-based procedures. The complete TAM data and code are publicly available.

Binary-choice truth benchmarks ask models to choose between a correct and an incorrect answer, but if the two answers differ systematically in surface-level features, models can exceed chance without performing the intended reasoning. We show that this failure mode is detectable and can be exploited by downstream classifiers. In TruthfulQA, a simple six-feature logistic classifier achieves substantial accuracy in separating correct from incorrect answers. We further show that similar surface-level artifacts are present in additional benchmarks. To counteract this, we developed a general mechanism to clean them by removing the most leakage-reinforcing pairs. We release a version of TruthfulQA with surface-feature leakage reduced close to chance and provide a mechanism, Audit-Prune, so that the datasets can be cleaned before release.

We investigate the effect of pretrained T5 model scale and explicit motion features on pose-to-text Indian Sign Language Translation (SLT) for the WSLP 2026 Shared Task. Pose sequences are projected into the embedding space of T5 through a lightweight pose encoder, with the complete model fine-tuned to generate English text. The shared task data used for this work consists of a test set with 5,334 examples and a validation set with 5,257 examples. We compare T5-small, T5-base, and T5-large, and additionally introduce a motion-augmented variant, T5-small + Motion, that adds explicit frame-to-frame pose differences to the input representation. T5-small achieves the best BLEU and ROUGE scores among the spatial-only models, while T5-large obtains the highest chrF score. Augmenting T5-small with motion features yields the largest single improvement observed in our study, substantially improving BLEU over the spatial-only baseline and making it the strongest model overall on this metric. Our submitted system ranked 5th on the official WSLP 2026 SLT testing leaderboard. The source code and trained models are publicly available on GitHub and HuggingFace.

We introduce an adaptive multi-parameter variant of ADMM and prove that it exhibits local superlinear convergence once the set of active constraints has been identified. In simulations, the proposed algorithm consistently outperforms OSQP, a standard ADMM solver, in terms of iteration count. We then implement an MPC solver based on our method and compare its runtime against a broader selection of state-of-the-art algorithms. Evaluations on challenging benchmark problems reveal that our approach delivers competitive performance both in terms of average and worst-case solve times, without being limited to coarse tolerances, as is typically the case for standard ADMM implementations and first-order methods.

Mixture-of-Experts (MoE) creates a structural advantage for offloading: only a small fraction of activated experts need to reside in device memory, and if they can be loaded in time for computation, offloading can in principle approach full-load performance, where all model weights reside in device memory. Yet translating MoE's structural advantage into practical offloading gains remains challenging. We propose SeqMoE to bridge this gap. To maximize expert hits, we build predictive memory management: (i) Sequence-to-sequence prediction. We are the first to recast expert activation prediction as sequence modeling, enabling accurate multi-step, multi-layer forecasts that provide a long and reliable window for downstream decisions. (ii) Joint prefetch scheduling. We formulate prefetch scheduling as Job Sequencing with Deadlines to maximize expected expert hits and improve bandwidth efficiency. (iii) Forecast-driven caching. Leveraging the recursive nature of sequence modeling, we introduce a probabilistic Belady policy for future-aware eviction. To eliminate execution bottleneck, we develop (iv) Graph-compatible offloading runtime. We derive general runtime principles encompassing compute-transparent expert placement and synchronization-free orchestration disciplines for end-to-end graph capture. With 45% expert residency, SeqMoE averages a 96.97% hit rate and 80.22% of full-load performance, advancing the state of the art in MoE offloading.

Robots are increasingly used in diverse application areas, where autonomous navigation plays a central role. As these systems become more widespread, improving their energy efficiency is critical to extending operational time and reducing environmental impact. The Robot Operating System (ROS) is a widely adopted middleware for robotics, offering a rich set of configurable packages. However, this flexibility can result in suboptimal software configurations in dynamic environments, negatively affecting both performance and energy consumption. This paper investigates the impact of ROS 2 package reconfigurations on the energy efficiency of mobile robot navigation. We conduct a controlled experiment in two warehouse-like scenarios (small and large) with varying obstacle layouts and Costmap 2D configurations (essential to the Nav2 stack). Through repeated trials, we measure energy usage, power profile, CPU load, memory consumption, and navigation performance. Results show that configurations must be carefully chosen for the specific robotic environment, and we were able to identify critical settings that lead to good and poor performance and energy consumption.

Text-based person search (TBPS) aims to retrieve images of a target person from a large image gallery based on a given natural language description. Most existing methods rely on supervised learning with manually annotated image-text pairs. In this paper, we explore unsupervised TBPS, with only unlabeled images. We propose GTR+, a two-stage generation-then-retrieval framework. In the generation stage, we introduce a tiered description generation framework designed to produce fine-grained and stylistically diverse textual descriptions through a three-tier sequential process. The base tier leverages an automated question-and-answer mechanism to generate basic visual attribute descriptions; the intermediate tier enhances fine-grained detail using an inter-sample contrastive mechanism; the advanced tier further enriches textual diversity via a stylized expansion mechanism. In the retrieval stage, to mitigate the impact of noisy pseudo texts, we develop an adaptive confidence-weighted retrieval learning framework. We model image-text pairs as clean or noisy using a Gaussian Mixture Model, calibrated by real-time image-text similarity and static text generation probability from the prior stage, yielding adaptive sample weights during training. Beyond that, we also contribute LargeFine-Person, a large-scale TBPS dataset with high-quality, fine-grained, and diverse textual annotations, enabling a practical and generalizable TBPS pre-training benchmark under unsupervised setting. Experiments on multiple TBPS benchmarks demonstrate the effectiveness and generalization of both GTR+ and LargeFine-Person. Code is available at: https://github.com/Flame-Chasers/GTR.

Large language models (LLMs) are increasingly deployed as tutors, but it is unclear whether they support all students equally well. We introduce \textbf{EduFair-Bench}, a benchmark for auditing the pedagogical fairness of LLM tutors---whether tutoring quality varies systematically with student demographics. EduFair-Bench pairs a multi-domain question bank (mathematics, physics, chemistry) with a controlled simulation in which a fixed LLM student interacts with each tutor across nine demographic levels spanning four dimensions: gender, immigration background, first language, and socioeconomic status (SES). Tutoring quality is scored on five turn-level pedagogical metrics and four conversation-level dimensions, using an LLM judge validated against three-annotator consensus on 180 tutor turns. Bias is measured via paired Wilcoxon signed-rank tests and bootstrap effect-size confidence intervals. Two ablations (demographic cues conveyed through names; conflicting demographic information between tutor and student) disentangle tutor-driven from student-driven bias. Across five tutors, we find that model capability and demographic fairness are largely orthogonal: the smallest model is the most consistent while the four more capable tutors all exhibit wide demographic gaps with no clear capability-to-fairness ordering, pedagogy-specific RL training redistributes rather than removes bias, and language- and immigration-related cues produce larger gaps than gender- and SES-related cues.

Proposed governance framework for autonomous robotic systems, introducing a three-layer compliance architecture (ARC) instantiated through model safety validation, cognitive certification benchmarks, and operational authorization standards.

Robust grasping of everyday objects remains challenging for parallel-jaw grippers, particularly when handling sloped or asymmetric items that induce torque-driven rolling and shear slip. These challenges become even more severe in domestic environments such as kitchens, where objects are often wet or submerged, drastically reducing friction between the gripper and the object. To address these issues, we present a waterproof passive adaptive gripper that combines local and global adaptability for high-performance grasping in the submerged environments. The proposed gripper features a fully waterproof design that integrates a passive rotational joint for global self-alignment on sloped surfaces and passive variable-stiffness pads for local surface adaptation. Experiments in both dry and underwater conditions on various cylindrical and conical objects demonstrate superior holding capability and grasp robustness compared to rigid-pad and fixed-joint baselines. The proposed design offers a practical and robust solution that successfully enables stable underwater manipulation of diverse everyday objects, effectively addressing a critical gap in current underwater robotic grasping for daily-life applications.

Handwritten mathematical expression recognition (HMER) is conventionally scored by exact-match rates and string-similarity metrics that are blind to where an error occurs: two predictions with identical token-error counts receive identical scores whether they misplace a subscript or swap the operands of a fraction. Render-based character detection matching (CDM) aligns glyphs robustly but remains position-blind---on controlled fraction-operand swaps it scores 0.8595 where position-aware scoring yields 0.6253. Tree-edit metrics exhibit a complementary blind spot: rewrites outside the parser's normalization coverage are penalized as structural errors (0.8552 where render-based metrics score 1.0). We propose PA-CDM, a position-aware metric that couples character detection matching with position-forest encoding and divergence-level weighting; StructPerturb v2.0, a frozen benchmark of 1,340 controlled perturbation pairs across 15 type--intensity cells; and a cross-metric consistency protocol combining a sensitivity matrix, a human study, and LLM-judge calibration. In a six-annotator study, PA-CDM attains the highest correlation with human judgments among seven automatic metrics (Spearman rho=0.9535, n=990). A frontier LLM judge correlates slightly higher (rho=0.9613) but is costly, nondeterministic, and API-dependent; PA-CDM approaches it at zero marginal cost with deterministic, diagnosable behavior.

Large-scale multi-label text classification assigns a small subset of relevant labels to each document from a vocabulary containing thousands or tens of thousands of candidate labels. Although pretrained language models have improved semantic text representations, most representation-based approaches center their prediction pipelines on a primary encoder or combine auxiliary features within a single ranker. The complementarity between heterogeneous language models therefore remains insufficiently explored. We propose DualMLC, a dual-branch framework that processes the same document through an autoregressive decoder-only language model and a bidirectional encoder. Each branch maintains its own representation pathway and independently estimates relevance scores over the shared label space. DualMLC combines the two score vectors through late logit fusion, allowing shared evidence to reinforce relevant labels and branch-specific evidence to compensate for limitations in the other branch's representation. DualMLC achieves state-of-the-art results on three widely used large-scale multi-label text classification benchmarks. Ablation results further confirm that integrating the heterogeneous predictors produces stronger rankings than either branch alone. The source code is publicly available at https://github.com/huiyegit/DualMLC.

Dense retrieval systems increasingly rely on multi-billion-parameter language models, whose memory and computational requirements make large-scale indexing, frequent corpus updates, and low-latency serving costly. We present a three-stage training pipeline for developing compact and efficient retrievers that remain competitive with substantially larger models. The pipeline combines cross-lingual alignment, relational knowledge distillation, and contrastive fine-tuning. It requires no original ground-truth relevance labels, relying exclusively on supervision generated by strong embedding models and rerankers utilised as teachers. Using this pipeline, we develop PolDense and EuroDense, both supporting contexts of up to 8,192 tokens. PolDense is a family of six Polish retrievers ranging from 17M to 1B parameters. EuroDense is a 435M-parameter retriever supporting nine European languages. We conduct an extensive evaluation covering 41 Polish and 150 multilingual retrieval tasks. The results demonstrate strong quality-efficiency trade-offs. PolDense-1B outperforms the evaluated retrievers with up to 9B parameters, while the PolDense family forms the Pareto frontier across model sizes. Among the evaluated models below 1B parameters, EuroDense ranks first in both task-averaged and language-averaged performance and leads in seven of nine languages. We release all models publicly.

This paper addresses the wind farm power maximization problem in the presence of wind direction changes. Specifically, a model-free Modified Twin Delayed Deep Deterministic Policy Gradient with Behavior Cloning (MTD3-BC) algorithm is proposed to tackle this task through yaw control under varying wind direction conditions. MTD3-BC is an offline reinforcement learning (RL) algorithm that aims to infer good behavior from only a precollected offline dataset. Additionally, to ensure smooth and moderate yaw adjustments, a new action consistency term is introduced into the policy optimization objective. Unlike online RL methods, MTD3-BC does not require extensive interactions with a wind farm simulator during training, significantly reducing computational costs and training time. A wind tunnel experiment is conducted to validate the effectiveness of the algorithm under varying wind directions. The results demonstrate that MTD3-BC successfully mitigates wake effects, delivering farm-level power gains of approximately 10\% over the baseline greedy strategy and performance on par with a data-calibrated model-based wake-steering benchmark, while requiring no wake model and only a small fraction of the training cost of online RL. To our knowledge, this is the first time an offline RL wind farm control policy has been validated and demonstrated experimentally.

Fine-grained robotic manipulation depends on understanding parts, not only whole objects. Existing 3D foundation models tend to be either generalized but object-aware, or part-aware but limited to closed-set taxonomies, which weakens zero-shot transfer. We study text-conditioned 3D part segmentation, where a free-form phrase selects a functional part on point cloud. We introduce UniPart, a feed-forward cross-modal 3D Transformer that conditions CLIP text embedding. To scale supervision, we build LangPart-1M with 160K+ Objaverse assets and 8M text to part pairs using multi-view consistent part generation. We further manually label a high-quality subset, LangPart-4K, for fine-tuning and evaluation. UniPart achieves strong zero-shot results on open-vocabulary part benchmarks and transfers to language-conditioned part grasping in real world.

LLM-based agents rely on heterogeneous interaction capabilities to accomplish complex tasks. Existing approaches often distribute these capabilities across multiple LoRA adapters, which increases adapter storage requirements and introduces routing overhead during inference. A single LoRA avoids this overhead, but learning from diverse agent trajectories under a fixed rank budget presents two challenges. First, trajectories with different interaction traces and parameter gradients can induce equivalent changes in decision distributions, causing repeated updates to overemphasize redundant behavioral changes. Second, an aggregated update may exceed the rank budget of the adapter, and approximating it in weight space can distort the decision changes that it is intended to produce. We propose BQ-LoRA, a low-rank adaptation framework that organizes trajectory updates through a local behavior quotient manifold. It contains two modules, i.e., behavior quotient balancing (BQB) and decision preserving compression (DPC). BQB constructs the quotient manifold from decision distributions and reweights trajectory update directions according to their local density in the quotient tangent space. DPC projects the balanced gradient onto the intrinsic fixed rank tangent space and refactorizes the resulting target by jointly controlling effective weight error and distortion of decision distributions. Experiments on AppWorld and BrowseComp-Plus compare BQ-LoRA with standard LoRA and recent low-rank adaptation methods, while separate ablations evaluate the complementary contributions of both components.

Estimating the 3D center of mass of unknown objects is challenging when grasping is infeasible, geometry is irregular, or mass distribution is uneven. We present a force-based method that estimates CoM height and mass from a single sub-critical tipping experiment by a robot manipulator. The robot applies a quasistatic elevated push and retract motion, using force-angle measurements recorded during tipping to identify parameters from the object trajectory. Our proposed push-retract cycle mitigates frictional bias, enabling generalized fitting. We experimentally validate our method using a robot manipulator with a six-axis force torque sensor on varying types of objects without prior shape information and without specific models. We also propose a method to prevent toppling, keeping the object in a sub-critical tipping regime by leveraging a safety margin. In experimental studies, our method recovers mass, CoM height, and toppling angle with relative errors below 5.0 percent across all unknown objects. This work demonstrates reliable 3D inertial parameter estimation under proper safety thresholds in tipping. Our proposed method informs and enables reliable non-prehensile manipulation and robotic grasping of challenging objects that were previously infeasible.

In autoregressive forecasting, long prediction rollouts provide distant supervision, but backpropagation through time (BPTT) carries gradients from those losses through many autoregressive steps. Repeated Jacobian products can make distant gradients dominate the update while amplifying predictable signal and unpredictable noise together; a large distant gradient therefore need not carry reliable learning signal. Motivated by this observation, we introduce Internal Dual-Wiener routing (Internal-DW), a principled backward-only intervention that preserves the full forward rollout and all horizon losses while reliability-weighting internal gradient routes. At each residual block, we derive bounded Wiener gains for the identity and nonlinear routes that balance preserving predictable learning signal against suppressing unpredictable variation, and estimate them from route-level gradient statistics and an explicit noise model. In a controlled system with known gradient signal-to-noise ratio (SNR), we show that distant gradients can grow even as their SNR falls, and that Internal-DW reduces held-out error in recovering predictable gradient signals and improves forecasting. On four history-dominated, weak-drive testbeds, Internal-DW reduces forecast error by 5.2%-13.8% relative to full BPTT, outperforms gradient clipping and Jacobian regularization on all four, and outperforms validation-selected truncated BPTT (TBPTT) on three. It also extends or preserves the fitted optimal training-horizon range across these four testbeds. Across the full benchmark suite, the current Internal-DW estimator has a clear applicability boundary: its benefit diminishes or reverses when usable history is limited or when the selected sampler fails to represent dominant drive-dependent variation. The results show that retaining long-horizon supervision does not require trusting every backward contribution equally.

A medical claim's correctness often depends not on the claim alone, but on the clinical structure around it. A claim may require a lab reference range, a causal or conditional link, or patient-specific details to be judged correctly, and atom-level decomposition can fragment these dependencies, leaving the verifier with clinically incomplete claims. We reformulate medical fact-checking around snippet-level verification, where clause-grouped units preserve local clinical structure. We introduce MedSNIP-Bench, a human-annotated benchmark for snippet-level medical fact verification, and MedSNIP, an automatic snippet-generation pipeline. MedSNIP-Bench covers 276 consumer-health and clinical-vignette responses, segmented into 2,524 snippets with dual in-general and in-patient-context labels and six structural pattern codes. MedSNIP is evaluated against human snippet boundaries on MedSNIP-Bench and then used to generate snippet-level units for external corpora. Across MedSNIP-Bench, HealthFC, and MedHallu, snippet-level verification preserves or improves false-class F1, with gains concentrated where answers are long enough to fragment and where the verifier is strong enough to exploit the recovered structure. The largest merge-pattern gain is on causal-conditional clinical chains. It also reduces verifier calls by 24-73%, though the saving survives end-to-end only when decomposition is cheap, which an open-weight decomposer makes possible at no loss of chunking fidelity.

Magnetic levitation (MagLev) systems have great potential for application in high-mix, low-volume manufacturing due to their scalability and flexibility, enabling highly reconfigurable in-machine material flow. However, their manipulation capabilities remain largely unexploited, as current applications almost exclusively focus on transportation. To enable grasping and manipulation directly on MagLev systems without requiring additional costly handling equipment, such as industrial robot arms, we present the Gripper MagBot, a low-cost parallel 6-DoF manipulator with an integrated 1-DoF gripper that mechanically couples three MagLev movers. The Gripper MagBot supports two operating configurations: a default mode and a single-track mode, selectable depending on the required stability and workspace footprint. To reconfigure a machine, the MagBot can be autonomously dropped off and picked up using a docking station. We showcase pick-and-place examples in simulation, as well as with the real Gripper MagBot using our inverse kinematics controller. CAD files, assembly instructions, a component list, and videos are available at https://sites.google.com/view/gripper-magbot.

We introduce GenOR-Twin, a neuro-symbolic framework that bridges the translation gap between unstructured operational logs and rigorous mathematical optimization. Our architecture uniquely positions Large Language Models as semantic translators rather than direct solvers, ensuring that the system retains the feasibility guarantees of exact combinatorial methods. { \color{red}We design a dynamic constraint injection mechanism (the runtime translation of qualitative disruption events into formal mathematical constraints) that allows the system to structurally modify the optimization problem's feasibility region in real-time based on qualitative human inputs. The resulting bidirectional coupling---where operational observations update the virtual model state and optimized decisions are reflected back into the Knowledge Graph---satisfies the synchronization requirement of a proper Digital Twin. The framework features an adaptive decision policy} that automatically selects between low-complexity schedule repair and full re-optimization by analyzing the available system slack. Finally, we demonstrate the generalization of this approach across six distinct optimization domains, {\color{red}turning static models into resilient systems that adapt to the operational uncertainty and variability of real-world environments.}

The widespread adoption of data-driven, energy-efficient model predictive control (MPC) in buildings remains hindered by substantial effort to collect data and train models for individual buildings. Transfer learning (TL) has consequently gained increasing attention for target building modeling, as it reduces data requirements and modeling effort by reusing pretrained source models. However, these TL models are typically evaluated only on prediction accuracy in the target, without testing downstream control performance. To address this gap, we apply a state-of-the-art TL approach - pretraining a generalized model on multiple source buildings using standard operational data - within an MPC setup in a target building. We show that this approach is insufficient to achieve satisfactory control performance. As a solution, we introduce generalized models pretrained on excitation-based operational source data - purposefully probed inputs that explore the building's state-action space. For evaluation, we apply the generalized models via zero-shot (i.e., without fine-tuning) to 32 simulated target buildings and assess MPC performance. Our results show that excitation-based generalized models achieve the strongest control performance among all benchmarks, outperforming an online linear model-based MPC and a PI controller by 6.4% and 36.9%, respectively. By combining strong control performance with the ability to generalize across multiple buildings, without requiring any target-specific data, our approach reduces MPC setup cost and simplifies its widespread deployment in the building sector.

Medical benchmarks are dominated by single-turn, multiple-choice clinical cases that poorly reflect real consultations. Practically, clinicians elicit evidence interactively and patient communication varies widely. We introduce MedRoundsQA, a multi-turn diagnostic benchmark derived from 1,387 board-exam cases across 17 specialties. Each case is converted into a structured 24-slot clinical record, and then instantiated as controlled doctor-patient dual-agent dialogues under varying patient personas, with the underlying clinical content held fixed. We further classify cases by difficulty using model-based uncertainty to enable easy-to-hard analysis. Evaluations of fifteen LLM doctor agents show that (i) moving from a single-turn diagnosis on the standardized records to multi-turn consultations causes large degradations of roughly 13-39 points; (ii) more turns reliably improves question relevance, but diagnostic accuracy exhibits diminishing returns and typically plateaus after 6-12 turns; and (iii) patient persona differences can shift diagnosis accuracy by about 7-8 points (lowest to highest education), highlighting equity risks that single-turn benchmarks miss.

Mapping observed system behavior to standardized frameworks like MITRE ATT&CK is essential for threat-informed defense, but remains largely manual. Existing automated methods depend on Cyber Threat Intelligence reports, which offer only retrospective accounts of attacks. Low-level telemetry, i.e. kernel-level system calls, instead provides evidence of adversary behavior, yet its volume and complexity have limited its use for automated mapping. We present a methodology that collects kernel-level events via eBPF, correlates attacker commands into a provenance graph, and derives compact graph representations suitable for LLM-based reasoning. These representations are mapped to the MITRE ATT&CK framework using both pure LLM prompting and retrieval-augmented generation (RAG) grounded in the ATT&CK knowledge base, producing ranked technique candidates along with supporting rationales. We implement this methodology as an end-to-end pipeline, named Trace2ATT&CK and evaluate it on 347 Linux Atomic Red Team tests using locally deployed open-weights LLMs. RAG consistently improves ATT&CK mapping performance over pure prompting, while provenance graph substantially outperforms raw telemetry. These results show that local inference over graph-based behavioral descriptions can make automated ATT&CK mapping from kernel-level telemetry operationally viable, without compromising data confidentiality.

The proliferation of highly capable open-weight Small Language Models (SLMs) democratizes access to advanced cybersecurity capabilities, posing a escalating risk as these models can bypass proprietary API guardrails when deployed locally. However, SLMs deployed as autonomous agents often struggle with long-horizon, exploratory tasks like cybersecurity Capture The Flag (CTF) challenges due to context bloat and cognitive degradation from accumulated tool-call outputs. To understand and mitigate this cybersecurity threat, we introduce \textit{context segmentation}, a two-level agentic framework that divides complex exploitation tasks into manageable, contextually isolated sub-problems. Evaluating on the \texttt{picoCTF} dataset using memory-constrained \texttt{gemma-4} models, we demonstrate that for the E4B model, our strategy acts as an intelligent search, achieving competitive rewards with superior token efficiency compared to brute-force retries, and successfully solving 18.52\% of tasks that standard agentic execution fails to complete. Code is available at https://github.com/9xeb/context-segmentation.

Blinded physician evaluation has been considered by many to be the gold standard for assessing clinical reasoning in large language models (LLMs). This is difficult to scale; thus, prior studies typically rely on small physician panels, often from a single institution or specialty, which both limits the scientific questions investigated and makes it unclear whether findings would be reproduced with a different set of evaluators. To more rigorously and scalably study clinical reasoning in AI models, here we introduce PrecepTron, an LLM fine-tuned for physician-level evaluation of open-ended responses. PrecepTron was trained using low-rank adaptation (LoRA) of a 32-billion-parameter model on a small number of physician examples. We also release GRAND-ROUNDS, a new large-scale physician-annotated benchmark of 9,217 scores by 11 physicians across seven studies. We show that frontier LLMs in typical "LLM-as-a-judge" approaches often disagree with physicians and with each other, but fine-tuning PrecepTron on a small number of cases enables physician-level consistent scoring across tasks. We use PrecepTron to reproduce headline findings from five influential studies assessing LLMs for clinical care in JAMA, Science, and Nature Medicine without new human grading. Using PrecepTron, we then pose new questions about how LLMs reason in medicine that would have been infeasible with human grading alone, including measuring the diagnostic accuracy of frontier LLMs when clinical cases are provided piecemeal, even token by token. Together, PrecepTron and GRAND-ROUNDS provide a foundation for reproducible, large-scale study of how LLMs reason in medicine. All code, data, and labels are made freely available for researchers.

Long video understanding often behaves like a visual needle-in-a-haystack problem: query-relevant evidence is sparsely distributed across long temporal spans, while packing dense frames into a single VLM context incurs \textit{context rot} and high cost. Existing video agents often rely on query-agnostic offline preprocessing or ad hoc tool sets, which can miss query-specific details and waste computation. In this work, we present VideoXAgent, a purely online video-agent harness for long video understanding that starts from the given video file and user query, plans and decomposes the task, invokes specialized expert tools on demand, and aggregates multimodal evidence to produce a final answer while resolving conflicts among observations. To support this on-demand invocation, we design a suite of heterogeneous expert tools guided by a data-driven taxonomy of atomic capabilities, spanning scripts, VLMs, and domain models (e.g., detection, OCR, ASR, face recognition). The harness further enforces objective evidence prompting and budget-aware control to curb hallucination and non-termination. Across Video-MME-Long, LongVideoBench-Long, LVBench, and MINERVA, VideoXAgent is competitive with frontier LMMs and video agents under a smaller context footprint---about 50k tokens of agent context per sample, even on hour-long videos. In particular, on complex video-reasoning benchmarks such as MINERVA, it matches this level while using only about 15\% of the context of a 1,024-frame dense-packing baseline. Notably, the harness remains effective with a visually weak or even text-only orchestrator, suggesting that strong long-video understanding can emerge from progressive agentic evidence seeking rather than from packing the full video into a single context. Project page: https://go-agent-x.github.io/video_agent_harness/

Linear attention and state-space models provide linear-time sequence modeling, but their recurrent memory remains a second-order tensor (a matrix), limiting the order of interactions that can be represented in the state. We introduce the RunningTensor, which generalizes this memory to an order-$o$ tensor, updated by a rank-1 outer product and read by contracting against $o-1$ vector queries. Order $2$ recovers linear attention; we study order $3$ as a proof of concept, retaining both recurrent and parallel forms while remaining linear in sequence length $T$ and improving working memory capacity from $\mathcal{O}(W^2)$ to $\mathcal{O}(W^o)$. On synthetic multi-query associative recall, RunningTensor outperforms linear-attention and SSM baselines. After pretraining, it also improves performance on language-understanding and non-synthetic retrieval tasks, suggesting that higher-order recurrent state can provide useful additional memory capacity beyond matrix-valued state.

Unlearning benchmarks such as TOFU and MUSE certify forgetting by reading the model's final answer, where a model that refuses to answer already counts as having forgotten. We show that this model-level certificate does not transfer once the model is deployed as an agent. We introduce K-Bench, a benchmark that scores LLM unlearning under agentic deployment. K-Bench inspects all six channels a ReAct agent exposes, including its chain-of-thought (CoT), tool calls and tool observations, and elicited summary. A query counts as leaked if the secret appears in any of them. Each experiment places the secret in exactly one of the agent's three sources (the weights, the prompt, or the retrieval store). The K-Score is computed separately for each source and credits forgetting only when the agent remains usable. Clearing the answer channel does not make the secret unrecoverable. On structured retrieval, the secret stays verbatim in the tool-observation channel and the aggregate leak rate is unchanged. When the secret lives in the prompt or the retrieval store, TOFU and MUSE report no leakage, while the deployed agent still leaks it on 22--86\% of queries. When the secret is in the weights, none of the twenty evaluated published methods demonstrably removes it, and only an input-corruption intervention reaches selective forgetting under the evaluated observer. The top-ranked method changes across base models. A refusal-tuning method resists the evaluated extraction without verified knowledge removal.

Stock price prediction remains challenging due to the non-stationary and noisy nature of financial time series. Existing deep learning models often rely on rigid decision-level fusion, ad hoc hyperparameter tuning, and compressed final-layer outputs, causing information loss, overfitting, and limited cross-market generalization. We propose VertiFuseX, a hybrid LSTM architecture using penultimate-layer vertical fusion of multi-scale temporal representations. VertiFuseX stacks and reweights penultimate features from LSTM, Bi-LSTM, and St-LSTM branches, integrates a parallel DNN stream, and jointly optimizes all components via backpropagation under a fixed hyperparameter configuration. This preserves richer intermediate temporal information across scales. Evaluated on 15 years (2010-2024) of closing prices from 10 global equity indices using strict chronological out-of-sample testing with the final 365 trading days held out, VertiFuseX achieves 30-54% MAPE reductions and over 40% improvements in MAE and RMSE versus LSTM-based baselines, and outperforms seven state-of-the-art models across 33 metric-dataset comparisons. Ablation studies confirm penultimate-layer fusion drives these gains over final-layer fusion and decision-level ensembling. Gradient-based saliency analysis shows consistent emphasis on mid-range dependencies at lags 9-15 days. Economic validation via algorithmic trading simulation under extreme market regimes shows reduced maximum drawdowns and superior risk-adjusted returns. With 675k parameters, a 2.6 MB memory footprint, and 1.5 ms/sample inference latency, VertiFuseX offers a lightweight, interpretable, deployment-ready framework for robust financial forecasting.

Retrieval-Augmented Generation (RAG) has empowered Large Language Models (LLMs) to tackle knowledge-intensive tasks. However, navigating global, heterogeneous knowledge bases (large-scale knowledge graphs and text corpora) for complex reasoning remains a challenge. Existing methods typically employ reactive, graph-driven exploration strategies, which blindly follow graph topology without adapting to the question context or evolving exploration progress, and lack deep bidirectional synergy between graph and text. To address these limitations, we propose CoG (Cognition on Graph), a cognitive-inspired, training-free framework for adaptive knowledge exploration. Drawing inspiration from human problem-solving, CoG performs a continuous plan-explore-reflect cycle, where it proactively formulates investigation plans, performs dual-source retrieval, and dynamically reflects on progress to adjust strategies. Crucially, it establishes deep bidirectional synergy between structured graph and unstructured text, where entities extracted from text dynamically guide graph exploration to bridge knowledge gaps. Extensive experiments on seven multi-hop QA benchmarks demonstrate that CoG significantly outperforms state-of-the-art methods while achieving superior exploration efficiency. Our code and datasets are available at https://github.com/zhougengxian/CoG.

Editing is a core component of video production, requiring creative planning and decisions under multiple constraints. Here, we report methods for agentic tooling for automated video editing across three tasks varying in editorial goal, complexity and creativity, namely scene previews, video summaries and cinematic trailers. We evaluate the outputs and discuss implications for automation and agency.

Vision encoders for Cell Painting are typically ranked by a single evaluation, commonly replicate mean average precision (mAP). We introduce CP-BG-Bench, a paired-view evaluation framework that holds the central cell fixed across four matched views (raw crop C, segmented S, and density-augmented variants CD and SD), ablating or augmenting surrounding pixels as a controlled intervention. Instantiating the framework on three datasets (JUMP-CP, RxRx1, RxRx3-core) and three encoders (DINOv3 ViT-B/16, OpenPhenom, SubCell) under four community-standard protocols (replicate mAP, scIB batch integration, CellProfiler feature prediction, cross-batch perturbation recall), we find that the four protocols rank the same encoders systematically differently, with disagreements decomposing along three axes: cell versus background, morphology versus context, and within-study versus across-batch. The largest effect: on RxRx3-core, SubCell with segmented inputs retains 94% of crop replicate mAP but only 32% of crop R@10, so the within-study signal preserved under segmentation is largely non-transferable; density augmentation recovers 84% of the within-study C-to-S gap but only 8% of the cross-batch gap. Segmented views predict CellProfiler features as well as or better than crops on two of three datasets, inverting the replicate-mAP ranking, and the C-to-S gap varies by an order of magnitude across datasets while remaining similar across encoders, indicating that background-driven gain is set by experimental design rather than by the encoder. Single-metric ranking of Cell Painting encoders is therefore sensitive to the protocol used, and protocol disagreements are interpretable as projections onto the three axes the paired-view design exposes. We will release the paired-view datasets, reconstruction pipelines, 36 trained checkpoints, aggregated embeddings, and the full evaluation suite.

Recovering unmanned aerial vehicles (UAVs) in maritime environments is challenging due to wind turbulence and ship-deck motion, making it a valuable test case for alternative control and learning approaches as conventional landing approaches often become unreliable. We study simulated mid-air capture of quadrotor UAVs by a ship-mounted robotic arm, learning robust cooperative control policies with Heterogeneous-Agent Proximal Policy Optimization (HAPPO) Reinforcement Learning. We train with HAPPO using a curriculum and an adversarial wind agent (HARL-AC) in NVIDIA Isaac Lab, and compare the obtained control policies against those generated through curriculum-based domain randomization and a benchmark trained on a single sea state. In-distribution evaluation on sea states $0/4/5$ shows comparable success for HARL-AC and domain randomization of up to $97.5\%$. On out-of-distribution sea states $7/8/10$, HARL-AC generalizes better, achieving up to $16\%$ higher median success rate at sea state 10, and substantially lower crash rates of up to $14\%$ compared to the domain randomization policy. Furthermore, we show that the adversarially trained policy shows more cautious behavior, slightly increasing timeouts by $<3\%$, but yields safer recovery behavior in severe, unseen conditions.

Test-time scaling improves code generation by spending additional inference budget (e.g., calls or tokens) on direct sampling, feedback-conditioned repair, and reasoning-guided implementation. Search-based methods can allocate this budget adaptively, but two challenges remain. First, tree-structured search treats each generation history as a separate state even when trajectories converge to the same program, duplicating evaluation and preventing statistics from being shared. Second, sampling, repair, and reasoning have complementary and state-dependent payoffs, making online allocation among them difficult under a finite budget. To address these challenges, we propose an adaptive graph search method with heterogeneous actions (GraphAHA). GraphAHA organizes the test-time code generation in a typed directed acyclic graph. Equivalent programs are merged into a single code node, allowing their downstream search statistics to be reused across all discovery paths. Hierarchical Thompson sampling then selects whether to generate a new state or follow an existing successor and, for generation, chooses among the type-valid sampling, reasoning, implementation, and repair operations. Evaluated on LiveCodeBench and CodeContests with Qwen2.5-Coder and DeepSeek-Coder, GraphAHA achieves the best score in 18 of 20 cases. For Pass@1 measured using visible tests, it outperforms the strongest baseline for both models on both benchmarks by 4.1 percentage points on average, demonstrating more effective use of a fixed inference budget.

Objective Prediction models are commonly trained using objectives such as Bernoulli negative log-likelihood (NLL), although downstream clinical decisions may depend on specific risk thresholds. We introduce Smooth Net Benefit ($σ$NB), a differentiable approximation of Net Benefit designed to align model training with threshold-specific clinical utility. Materials and Methods We evaluated $σ$NB as a training objective for logistic regression, generalized additive models (GAMs), and XGBoost with three Hessian implementations. Experiments used the Framingham cardiovascular risk dataset and 44 TabZilla datasets comprising 72 dataset-threshold combinations. Results $σ$NB training did not consistently improve Net Benefit in Framingham. Across the TabZilla benchmark, mean standardized Net Benefit for logistic regression increased from 0.5669 with NLL to 0.5765 with $σ$NB (mean difference 0.0096, 95% CI -0.0001 to 0.0193). For GAMs, mean standardized Net Benefit decreased from 0.5921 to 0.5625 (mean difference -0.0296, 95% CI -0.0721 to 0.0129). For XGBoost, NLL achieved 0.6745 compared with 0.6723--0.6735 across $σ$NB implementations. In logistic regression, $σ$NB gains were positively associated with the performance advantage of XGBoost over NLL-trained logistic regression. Discussion The effect of $σ$NB was context dependent, with modest gains concentrated in logistic regression and little benefit for more flexible model classes. This suggests that decision-focused optimization may be most useful when limited model flexibility leaves greater scope for improvement. Conclusion Our results do not support $σ$NB as a general replacement for NLL training, but support further investigation of decision-focused objectives in settings where conventional likelihood-based training may not adequately capture decision-relevant structure.

Offline-to-online reinforcement learning reduces interaction cost for real-world robot learning but suffers from persistent value estimation instability. Existing methods address this through pessimistic regularization, lower-bound calibration, and architectural normalization, but an overlooked source of instability lies in the entropy formulation: the standard log-entropy term can become negative, destabilizing policy updates. We introduce SCQ (Sigmoid-Bounded Conservative Q-Learning), which replaces this term with a sigmoid-bounded formulation that stays strictly positive. SCQ retains conservative Q regularization and return-based lower-bound calibration, stabilizing policy optimization without sacrificing exploration. We evaluate SCQ on D4RL (Minari) benchmarks under both single-demonstration and standard dataset settings, as well as on simulation and real-world visual tasks. SCQ matches or exceeds baseline performance while exhibiting more stable training dynamics across state-based and visual benchmarks, and transfers to four real-robot platforms including manipulation, wheeled, quadruped, and humanoid systems. A direct clipping intervention that removes negative log-probability contributions, together with gradient-matched positive-score controls, indicates that positivity rather than a particular score shape alone drives much of the improvement. Project website: https://scq-rl.github.io.

Agentic pipelines for structured-query generation are rapidly expanding, but it is unclear which part of the loop produces the gain. We use LAST-CQ -- a five-agent, training-free, execution-grounded Text-to-Cypher framework -- as an instrumented testbed, running three counterfactuals over 2,471 live-database queries and six backbones spanning three vendor scale tiers. Removing correction is worth between 3.1% aggregate execution-BLEU against the single-pass system and 12.3% against a no-refinement counterfactual (up to 80.7% for the weakest backbone). Replacing schema-grounded, LLM-synthesised feedback with raw database error strings costs almost nothing (20.9% vs. 19.9% naive exact match; <0.2% end-to-end; equivalent within $\pm 0.075$ set-F1 by two one-sided tests). Spending the same call budget on parallel sampling degrades quality by 10-11%. What works is detecting failure and routing it to a retry, not the feedback sophistication or number of samples. LAST-CQ itself recovers 91.7% of queries that fail under single-pass generation, while a query that succeeds first time still costs exactly one LLM call. We also show that n-gram overlap on serialised results is not a bound in either direction: it over-scores against set equivalence on 65.9% of results while under-scoring against judged semantics. Finally, we calibrate our LLM judge against blind human labels and find it optimistic by 9 points.

Coding agents increasingly read repository knowledge from SKILLs --- plain \texttt{.md} files versioned alongside the code. Recent work synthesizes these files automatically, by optimizing the document against a benchmark. A bare repository comes with no benchmark, and the synthetic tasks prior work builds are small enough that a capable agent saturates them with no document at all. We mine harder tasks --- merged pull requests of the repository, reverted at a single frozen base commit; and score a candidate document by whether the same agent does better with it than without it. On three Kotlin repositories, the documents GEPA finds raise this score by $4.9$pp on average, and the ones SkillOpt finds leave it where it started, $0.1$pp above the seed. The GEPA gain matches what prior work reports with the same optimizer, and at the dataset size a single repository supplies it cannot be separated from the agent's run-to-run variance; settling that would take more tasks than one repository's history yields. The documents themselves read better than the score: a maintainer of one repository found in them knowledge one only gets by working in the project.

Robots are increasingly used in unstructured environments. The need for them to safely grasp unknown objects without damaging them becomes crucial. Humans achieve this by sensing and quickly responding by adjusting their grasping force. Similarly, effective grasp acquisition in robots requires compliant interaction strategies that can adapt to uncertain object properties and adjust to any instabilities during manipulation. We present a geometry-aware force/torque-based contact estimation method for a coarse position-controlled gripper, combined with an adaptive admittance controller for safe grasp acquisition. The desired contact forces are estimated online to keep stable contact with objects of unknown properties. This enables compliant and stable grasps while avoiding excessive forces. Experiments with objects of different sizes, shapes, stiffnesses, and weights show that the proposed algorithm not only prevents slippage but also applies minimal force to safely grasp an object without causing excessive deformation.

Applying an imitation learning policy to a new manipulation task usually requires collecting new demonstrations and retraining the model, which makes sample efficiency a practical concern. Pretraining on large-scale robot datasets is effective in this respect, but such datasets are costly to collect and train on, while data augmentation techniques typically require a new round of data generation and retraining for each task. A complementary question is what useful prior can be provided to a policy at negligible cost before any task-specific data are collected. In this study, we construct a geometric visual pretraining dataset in which each scene contains only a plane, an object, and a hand, and trajectories are generated automatically. The scenes contain neither textures nor backgrounds; pretraining primarily exposes the policy to the geometric relationship between the hand and the object. Furthermore, representing the hand as a cube avoids tailoring the dataset to a specific robot morphology. We evaluate this geometric prior using ACT on three simulated robots across five manipulation tasks each, as well as on three real-world robot tasks. Across many of these robot--task combinations, fine-tuning from the geometric prior achieves higher success rates in the early stages of training than training from scratch while using only a small number of task demonstrations. These results suggest that even highly simplified geometric scenes can provide a useful initialization that transfers across robots and to real-world tasks when task data are limited.

We consider the problems of formal verification and synthesis in parametric Markov decision processes (MDPs) with general measurable state and action spaces. The heart of our approach is a parameter flattening transformation, which allows us to transform parametric MDPs into semantically equivalent non-parametric MDPs. Building on this transformation, we introduce the novel notion of parametric supermartingale certificates, which generalize the traditional supermartingale certificates---used for non-parametric MDPs---to the parametric setting. We use our parametric supermartingale certificates to design algorithms for verification and approximate synthesis in polynomial arithmetic parametric MDPs. This leads to the first verification and synthesis algorithms for parametric MDPs with general state and action spaces. We implement our algorithms and experimentally evaluate them on several continuous parametric random walk benchmarks.

Foundation models for human brain recordings are usually evaluated by signal reconstruction, future-state prediction, and transfer to downstream tasks. However, these benchmarks do not establish whether a transferred model remains valid when the system is actively perturbed. We define perturbational validity as the preservation, after limited system-specific adaptation, of the conditional distribution of future trajectories given the current state and a controlled input, and we evaluate it at three levels: time-series accuracy, dynamical-structure similarity, and responses to perturbations excluded from calibration. We demonstrate the framework in an oracle-drift simulation of stochastic bistable systems. Two otherwise identical multilayer perceptrons were trained on drift evaluations from passive or input-driven trajectories. Next, for each held-out system the shared weights were frozen and only a three-dimensional embedding was adapted, with a correctly specified cubic model fitted from scratch as comparator. With two to five system-specific evaluations, perturbational pretraining yielded lower errors in recovering controlled flow, landscape geometry, finite-run occupancy, response distributions, and dose-transition curves. The advantage was reproduced across five independent runs, persisted under full-network adaptation of the passive model, and was attributable to input excitation rather than transition-state coverage: excitation alone lowered controlled-flow error 1.94-fold relative to coverage alone, in five of five runs. The cubic model became competitive as calibration grew, showing that the benefit is specific to few-shot transfer. This controlled demonstration does not test recovery of dynamics from noisy or partially observed brain recordings. It shows why passive prediction should be complemented by prospective evaluation under controlled inputs.

AI coding assistants are becoming co-authors of production software, yet their evaluation centers on functional correctness, leaving open whether their code differs from human code in the quality dimensions dominating lifecycle cost. We compare human-written and AI-generated code at scale: 787,562 function pairs across Python, Java, and C, each human function mined from open-source repositories paired with implementations generated from its docstring by three AI assistants (OpenAI GPT models, DeepSeek-Coder, Qwen2.5-Coder). We characterize structural complexity and statistical naturalness, and map static-analysis findings onto Orthogonal Defect Classification for defects and the Common Weakness Enumeration for vulnerabilities, making authors and languages directly comparable. AI-generated code is structurally compressed and stylistically templated: roughly half the size and branching of human code, clustering apart at the style level. Defect profiles differ in kind: human code concentrates issues of mature codebases, AI code repetitive boilerplate; security is language-dependent, with LLMs producing more, and more severe, findings in Python and Java but fewer high-severity memory-safety findings than humans in C. Once size is controlled for, complexity metrics carry little signal, while naturalness separates authors. Finally, we release CQBench, a benchmark of 27,346 issue-prone tasks with baselines and an evaluation pipeline for quality assurance and security testing.

Handwritten text recognition resources are often small and distributed across collections that differ in language, script, document structure, and annotation format, making joint page-level training difficult. We propose ExpertHTR, a unified vision-language framework that addresses this problem through complementary supervision and conditional model capacity. Structural annotations from heterogeneous datasets are first organized through a common Page-Region-Line representation and used to construct four related training tasks for complete transcription, physical-line coverage, text localization, and localized recognition, without requiring additional manual labels. Building on a jointly trained dense model, ExpertHTR introduces a sparse Mixture-of-Experts architecture with an always-active shared branch and conditionally routed full-MLP experts. Sparsegen allows the number of active routed experts to vary with the hidden representation, while routing regularization reduces persistent concentration on a small subset of experts. Experiments on seven heterogeneous handwriting benchmarks show that complementary supervision consistently improves training with page transcription alone, while joint multi-source training provides further gains on most datasets. The proposed sparse expert model further improves the dense baseline on six of the seven sources. The final unified model also substantially outperforms the evaluated general-purpose OCR and vision-language systems on most benchmarks and achieves state-of-the-art performance on the IAM paragraph-level benchmark, while specialized HTR systems remain stronger on several challenging collections.

Personalized marketing can increase customer engagement, satisfaction, and conversion. While existing personalization approaches have become effective at matching the right product to the right customer, the visual representation of advertisements remains generic and only weakly tailored to the individual. Prior research shows that generative artificial intelligence can improve the creation of personalized advertisements, particularly for text, and that image generation models can support scalable advertisement production. However, little research has examined how detailed customer information can be systematically translated into fully AI-generated, personalized advertising imagery at scale on a technical level. To address this gap, we propose AdMan, a multi-agent pipeline that transforms customer data into personas, generates personalized advertisement images conditioned on product reference images, and applies an LLM-based judge agent for automated quality control. We implement the pipeline with two different model configurations and evaluate it across four products, using six celebrity personas for qualitative inspection, and 100 real customer profiles, producing 1745 advertisements. The evaluation combines a qualitative expert focus group and a quantitative artifact-rate assessment. The results show that the pipeline can generate photorealistic and personalized advertisements. At the same time, performance varies substantially by product complexity and model configuration. Our findings extend the literature on AI-based personalized advertising by demonstrating the feasibility and current limitations of fully automated image generation for advertising.

Large language models (LLMs) process long contexts, including long documents and lengthy conversations, but face token-level memory usage that increases proportionally to input length. Although model optimization and lossy prompt compression are widely used, these methods still fail to solve the long-context recall problem beyond pretrained and size-constrained context windows. This paper proposes a long-context recall method that maintains near-constant GPU memory usage as context length increases, without additional training. The main idea is to reconstruct facts using parameter activations in the LLM's feed-forward layers, which store residual vectors representing facts from the source document. Utilizing residual vectors allows the LLM to deterministically reconstruct query relevant facts without referencing the original document, preserving high fidelity and reducing memory usage without fine-tuning weights. Experimental results show that the proposed method enables answering single-fact questions in two-million-token story contexts where previous methods fail.

Generative artificial intelligence (GenAI) is changing how work is organized and performed. Real estate marketing is a prime example of this, yet evidence of GenAI in real estate agents' day-to-day practice remains scarce. In this work, we report on our insights from a German-based empirical study with eleven semi-structured interviews. GenAI is already utilized across different activities, with marketing communication being the most prominent. Concrete use cases are emergent and unevenly adopted, with writing exposé texts being the only widely established one. Interaction is predominantly human-in-the-loop: GenAI drafts, structures, and retrieves, while real estate agents curate, verify, and decide. Constraints stem less from model capability than from integration with listings and documents, data availability, and compliance in sensitive tasks. The study contributes a grounded map of existing and potential use cases and identifies tentative practical implications for adoption.

Earthmoving tasks such as excavation, backfilling, or embankment construction require deliberate repositioning of deformable soil. For these tasks, human operators use all shovel faces, while autonomous systems so far are limited to excavation and dumping. Current methods often rely on heuristic models but do not incorporate soil mechanics. We address this shortcoming by using Reinforcement Learning in a GPU-parallelized Material Point Method particle simulation. Our controllers are conditioned on material state such as shape and compactness, enabling skills that use multiple contact faces of the tool and displace material both inside and outside of the shovel. To use the same learned weights across machines, our policies operate in a normalized end-effector space and are deployed through a calibrated machine interface. We evaluate this calibrated transfer on an 11.5t hydraulic excavator and a 500g tabletop robot. We validate performance through autonomous construction of a 42m long, 2.1m high embankment in 45min, executing 201 individual policy strokes without failure, retry, or operator intervention. In a direct comparison, the autonomous controller matches an expert operator's progression speed and produces a higher, more consistent embankment. Additional qualitative backfilling and compaction experiments demonstrate the material-state awareness and calibrated transfer across machines.

Advanced large language models (LLMs) with long context windows can substantially reduce input truncation in document-level machine translation (DocMT). However, direct Doc2Doc translation remains prone to n-gram repetition and progressive quality degradation. A common remedy is to segment the document into finer-grained chunks. Nonetheless, conventional rule-based chunking approaches fail to handle the length distribution mismatch between training and inference. To address this, we introduce Fixed-Range Chunking (FRC), utilizing dynamic programming to partition documents into chunks within a predefined length interval. By consistently applying FRC during training and inference, the input documents of any length are mapped to the same length distribution, substantially reducing train-test length mismatch. Centered on FRC, we propose a lightweight dual-boundary matching algorithm for chunk alignment, alongside four distinct training strategies. Experimental results show that FRC-based fine-tuning substantially improves 7B LLMs over direct Doc2Doc fine-tuning and outperforms existing DocMT methods on IWSLT2017. We further construct GlobVDoc, a 10-language test set independent of mainstream DocMT training sources, and show that FRC improves out-of-distribution document translation.

Audio-based Multimodal Large Language Models (MLLMs) can generate detailed natural-language descriptions of complex acoustic scenes, yet it remains unclear which parts of the input audio support each generated token. This is particularly challenging because acoustic evidence is distributed across time and frequency, and concurrent sound events may overlap temporally while occupying different spectral regions. We introduce STAG, to our knowledge the first post-hoc framework for token-level spectro-temporal grounding of captions generated by audio-based MLLMs. STAG estimates the temporal support for each generated token using target-token-specific vocabulary projections of the encoded audio representations, measures frequency-band relevance through controlled spectral occlusion, and combines the two signals into a spectro-temporal relevance map. We evaluate STAG against ten post-hoc explanation methods across four grounding benchmarks, where it achieves the best event-localization performance on every dataset, and apply it to eight audio-language backbones without parameter updates. Counterfactual deletion further shows that removing the identified evidence selectively reduces confidence in the corresponding event and frequently removes it from the regenerated caption. These results provide behavioral support for the faithfulness and selectivity of the explanations.

Large language model agents are expected to continuously adapt to new tasks and environments over their lifetime by reusing past experience. However, existing memory-based agents struggle to transfer reusable experience across environments and suffer from catastrophic forgetting as experience accumulated. To address these challenges, we propose LifeMem, a lifelong learning framework that enables agents to transfer knowledge across multiple environments. During learning, LifeMem clusters accumulated interaction trajectories based on underlying workflows to extract reusable skills. When solving a new task at inference time, the agent recalls relevant skills and trajectories to guide actions. To validate our method, we conduct experiments across 10 environments and over 13k tasks with 2k newly annotated interaction trajectories. Results show that LifeMem enables effective experience reuse in lifelong learning, achieving both reduced forgetting on learned tasks and superior cross-task transfer. Further analysis reveals that task streaming impacts learning, while consolidating structurally similar trajectories within memory boosts performance.

Russian state propaganda spreads across many languages and online spaces. Yet, most computational work examines only one such space, usually social media, in one or two languages, and analyses sources rather than content. We introduce SWARM (Search-Web documents Annotated for Russian propaganda, Multilingual), a dataset of 2,183 search engine results across nine languages and diverse web domains (e.g., news, blogs, government sites), each annotated by trained coders for whether it supports a recurring Russian propaganda narrative. We benchmark a source-based blocklist, supervised classifiers, and zero-shot LLMs against these labels. The blocklist misses most propaganda-supporting documents, because such content is not confined to flagged "propaganda" outlets but also appears on mainstream ones. Content-level analysis helps, though how much depends on the model: the strongest LLM reaches a positive-class F1 of 0.73, whereas the supervised classifiers reach only about 0.5, with the smaller LLMs over-predicting support, mistaking topical relevance for endorsement. Detecting search-borne propaganda thus requires per-language, content-level evaluation, which we hope SWARM and our evaluation code enable.

While Reinforcement Learning from Human Feedback (RLHF) is the standard paradigm for aligning large language models with human preferences, its effectiveness in pluralistic settings has been called into question. Notably, recent work by Gölz et al. (2025) demonstrated that the \textit{distortion} -- defined as the multiplicative gap between the average user utility of the RLHF policy and the optimal average utility -- can scale exponentially with the Bradley-Terry temperature parameter $β$ when users have heterogeneous preferences. In this work, we present a fine-grained analysis of the distortion of RLHF with reward clipping and demonstrate that such exponential degradation is not a fundamental property of the algorithm but rather a consequence of distribution mismatch between the distribution generating preference data ($μ$) and the KL reference policy ($π_{\mathrm{ref}}$). To this end, we establish tight upper and lower bounds on the distortion of RLHF across multiple regimes of the KL regularization strength. We show that in a representative regime, under the Bradley-Terry model, the distortion is $\tildeΘ(βB + β)$, where $B$ is an upper bound on the log density ratio between $μ$ and $π_{\mathrm{ref}}$. In particular, when there is no distribution mismatch (i.e., $μ= π_{\mathrm{ref}}$), RLHF achieves the optimal distortion of $O(β)$ up to a constant. Our results suggest that, to reasonably maximize average utility with RLHF, it is preferable to use on-policy sampled preference data or to fine-tune before RLHF on data from a source close to $μ$.

Robot foundation models achieve strong in-distribution performance but often degrade under visual distribution shifts. When learning to generate actions from pretrained visual representations, models may exploit task-irrelevant visual cues that correlate with demonstrated actions within the training distribution. Such vision-action shortcuts can undermine generalization when these correlations change under distribution shifts. Mitigating these shortcuts requires constraining how visual information is used for action generation while preserving task-relevant spatial information. We propose Latent Interface Training (LIT), a framework-agnostic two-stage strategy that first establishes a spatial-goal-conditioned action prior without images, then constrains visual conditioning through a pose-supervised latent interface. Stage 1 trains the action expert to generate action chunks conditioned on language, robot state, and each demonstrated chunk's terminal SE(3) end-effector pose, learning goal-directed action generation independently of visual cues. Stage 2 introduces a latent interface that aggregates visual and semantic representations and serves as the pretrained action expert's only visual conditioning pathway. The interface is supervised to reconstruct the terminal pose previously used to condition Stage 1, encouraging it to retain the goal-relevant spatial information needed for action generation. Across four vision-language-action and world-action architectures (Pi0.5, MolmoAct2, FAST-WAM, and ImageWAM), LIT improves overall LIBERO-Plus success by 3.87-10.70 percentage points while preserving or improving average LIBERO success. Real-world evaluations show 13.30-16.70 percentage-point gains in success aggregated across three tasks under unseen camera configurations, lighting variations, and distractors.

Recent advances in explaining time series (TS) models have produced methods that identify which past values a prediction depends on. However, most existing methods return a single importance vector, assuming that every predicted step depends on the same past values. In this paper, we show that this assumption does not hold, as different forecast steps depend on different past values. Motivated by this observation, we propose Horizon-Resolved eXplanation (HRX), which adds a horizon axis to the explanation, so that every forecast step receives its own importance map. HRX is a simple yet effective plug-in framework with three components: 1) an estimator that reads these maps out of any differentiable forecaster without modifying the TS backbone, 2) an evaluation protocol that validates the horizon axis by measuring how much a single forecast step changes when the inputs an importance map ranks highest are removed, and 3) a rank criterion that predicts in advance whether the axis is worth resolving on a given TS. We further show that this step-wise dependence is low-dimensional, as the explanations of all steps are built from a few shared maps whose number does not grow with the forecast length. Extensive experiments across various backbones and datasets show that the improvement comes from the horizon axis and holds for estimators of previous explanation methods. Code is available at https://github.com/seunghan96/HRX.

Shape control of deformable linear objects (DLOs) is challenging for imitation learning because deformation behavior varies with material properties such as stiffness and elasticity, so a single policy must generate different action sequences for different objects even when the goal shape is identical. We propose a diffusion policy conditioned on material labels that are estimated online during manipulation. A recurrent estimation network predicts the material label of the grasped object from the time series of multi-view images and robot joint states, and the predicted label conditions the diffusion policy at every inference step. We collected 480 real-robot demonstrations covering four DLO materials and three groove-placement tasks, and compared per-material specialist policies, a task-conditioned policy without material labels, a policy conditioned on ground-truth material labels, and the proposed policy. Conditioning on ground-truth material labels improved the average success rate from 45.8% to 60.0% over the task-only policy, and the proposed policy reached 60.8% without any prior material information, matching the policy given ground-truth labels. A post-hoc analysis shows that the estimator extracts material-related information from the manipulation observations and that the diffusion policy responds to the resulting conditioning signal, while the one pronounced failure case is associated with persistent confusion between two similar materials.

Full-duplex spoken dialogue models support low-latency turn taking, interruption handling, and backchanneling, yet a key capability remains underexplored: steerability, the ability to reliably shift conversational behavior along attributes such as tone, persona, speaking rate, and voice style in response to user instructions. We introduce a taxonomy of text- and audio-based steerability that identifies substantial gaps in current full-duplex models. To address this gap, we introduce SteerDuplex, a Moshi-based full-duplex speech model fine-tuned on natural conversations and synthetic dialogues targeting instruction following, vocal delivery, reasoning, and duplex interaction. We further apply two-stage reinforcement learning (RL) with hybrid rewards, combining verifiable interaction checks and judge-based semantic feedback to improve timing and response continuity. To evaluate full-duplex spoken steerability, we introduce SteerBench, a benchmark with 390 spoken prompts and 1,067 human-authored binary audio and text rubrics spanning tone, persona, style/accent, and speed/length. On SteerBench, supervised training improves audio-steering average pass rate by 44.5 percentage points over the strongest evaluated open baseline. On Audio MultiChallenge, task average pass rate improves by 7 points over its strongest evaluated open baseline. RL further raises source-clean interruption response from 72.5% to 82.5% and reduces synthetic pause barge-in from 26.5% to 9%. Steering and aggregate task scores remain comparable or higher, while reward probes reveal reward hacking through incomplete responses. Our model and benchmark support systematic research on spoken steerability, with reward analysis showing why timing gains must be evaluated alongside response completeness.

A 3D scene graph groups objects into rooms. When a robot is asked to fetch an object from the kitchen, that grouping is what tells it where to look. An object recorded in the wrong room is not retrievable by a query naming the correct room. We introduce Progressive Boundary Closure, which recovers room layer from a monocular RGB video. A SLAM front end and an open-vocabulary segmenter supply a structural point cloud, camera trajectory and object tracks. The cloud is rasterised into a top-down map, rooms are recovered from it, and each object takes the room holding most of its extent. The difficulty lies in the map itself. Walls are recorded only where the camera looked, so a gap in the boundary may be a doorway or a stretch of wall that was never observed; nothing distinguishes the two. Prior methods treat both as passages, merging rooms that should remain separate. We observe that both require the same treatment: a room should not extend across either, so both are closed and need not be distinguished. Such an opening closes under a small amount of boundary growth, and few sightlines cross it, so points in different rooms rarely see one another. We use the first to recover rooms and the second to assign objects to them. Rooms are obtained by Progressively thickening the boundary inward and freezing each free-space region once it becomes enclosed, so every opening seals at its own scale rather than at a radius fixed in advance. Camera poses are used as seeds, which removes the sampling heuristic and makes the segmentation deterministic. Over 10 floors of 6 HM3D-Semantics scenes, scored against HOV-SG on identical top-down maps, we recover 74 rooms for 72 annotated regions (HOV-SG: 44), raising room F_1 from 0.741 to 0.890 at IoU 0.25 at some cost in precision, and object-to-room ARI from 0.488 to 0.696 (p=0.002, ahead on every floor).

Large Language Model alignment typically relies on learned proxy reward models, which significantly increase the memory footprint during training and are notoriously prone to instability and reward hacking. While offline methods like Direct Preference Optimization (DPO) bypass the reward model, they lose the ability to perform online exploration. If no optimization constraints are applied, this can lead to format collapse in bounded, continuous spaces. To resolve this, we propose Direct Preference Density Alignment: An alternative framework that removes the need for a learned proxy reward model while strictly preserving the benefits of online reinforcement learning. We leverage large-scale user data (approximately 90,000 samples) to construct non-parametric preference density maps, establishing an empirical reward surface. In addition to removing the reward model, Direct Preference Density Alignment enables the combination of the online structural grounding of Group Relative Policy Optimization (GRPO) with the targeted offline refinement of DPO. We show that this GRPO+DPO combination achieves the highest performance, and in a blind audio equalization listening test, enables a 1.5B-parameter model to achieve perceptual parity with a carefully prompt-engineered GPT-4o mini baseline, using only a fraction of the inference compute.

While multimodal reasoning has advanced rapidly, solving complex geometry problems critically hinges on active visual assistance, such as constructing auxiliary lines, spurring the rise of Visual Chain-of-Thought (VCoT). However, existing evaluations typically assess visual generation quality and final answer accuracy in isolation, failing to examine whether intermediate visual aids are geometrically valid, effectively utilized in subsequent reasoning, or causally responsible for task success. To bridge this gap, we introduce GeoVAD-Bench, a diagnostic benchmark that pairs a fine-grained five-dimensional trajectory diagnosis covering perception, auxiliary quality, utilization, deductive reasoning, and final correctness with controlled No-Aux, Auto-Aux, and GT-Aux intervention settings to systematically isolate intermediate error modes, the causal gains of visual aids, and the resulting autonomy gap. Our findings reveal that while high-quality auxiliary aids offer substantial theoretical gains for geometric problem solving, autonomous generation is frequently hampered by compounding errors across geometric perception, faithful visual manipulation, visual-state grounding, and deductive reasoning. Guided by these diagnostic insights, we establish a specialized data construction pipeline encompassing geometric perception, diagram editing, and interleaved visual-textual reasoning trajectories, and develop a progressive SFT and multimodal RL training framework. The resulting model, GeoWeave-8B, outperforms the base model by +25.3% in final geometric accuracy and achieves a +30.4% gain in process average across the four intermediate diagnostic dimensions.

Multi-task learning (MTL) requires navigating unavoidable trade-offs among competing objectives. This paradigm is frequently formulated as multi-objective optimization (MOO), where the scalarization is favored to reduce an MOO problem to a single objective. We empirically find that existing merit-function-based scalarization approaches are sensitive to the relative scales of different objectives in practical MTL, where task losses commonly differ by orders of magnitude. The optimization process often favors objectives with larger scales even though the underlying Pareto optimal solutions remains invariant to rescaling (i.e., multiplying an objective by a positive constant). To address this issue, we propose Scale-Invariant Merit-function-based Scalarization (SIMS) for MTL. Specifically, SIMS adopts a transformation-induced merit function to convert the MOO problem of MTL to a single objective that renders optimization invariant to the magnitudes of losses. Theoretically, we prove that the requirement for scale invariance uniquely determines this transformation to be logarithmic. We further show that this general transformation-induced merit function preserves weak Pareto optimality and admits a smooth surrogate with controllable approximation error. Extensive experiments on representative multi-task benchmarks demonstrate that SIMS consistently outperforms existing scalarization methods and achieves state-of-the-art performance.

A compact controller can coordinate stronger experts by selecting whom to consult, formulating requests, and integrating their responses. We study whether learning from both the controller's decisions and the experts' reasoning and code improves its generation after expert removal. We introduce \textsc{Rivet} for \emph{collaboration internalization}: expert-augmented reinforcement learning applies a shared outcome signal to controller decisions and returned expert spans, and verified trajectory internalization consolidates complete successful interactions through format-aware supervised training. The deployed controller generates reasoning, code, and interaction structure with local Python execution and no external LLM. Across seven competition-mathematics benchmarks, RIVET-1.7B and RIVET-4B achieve average accuracies of $28.25\%$ and $44.16\%$; Stage~II improves RIVET-4B's accuracy after expert removal by $6.49$ points, and GPQA-Diamond results provide evidence of generalization to scientific reasoning. Ablations show gains from ordinary trajectory supervision and additional format weighting, supporting the effectiveness of training on the content and structure of verified collaborations.

AI is beginning to make substantive contributions to LLM inference optimization. Existing AI optimizations are predominantly profiling-based. Profiling-bound feedback confines the search to the capabilities and performance of an existing software stack, preventing a fundamentally better architecture of LLM inference systems from being identified. To enable the AI-driven LLM inference system architecting loop, we argue that a general workload representation, a verifiable mutation space, and an implementation-independent evaluator are required. We present the RoofLang domain-specific language (DSL) that provides these features. In our evaluation, RoofLang reveals that DeepSeek V4-series models could achieve 3.5-39.5$\times$ higher peak decode throughput than other representative models. This gap is disproportionate to their total parameter counts and arises largely from compact KV-cache designs that support larger batches and reduce memory traffic. A persistent optimizer agent further discovered several new architectures that improved both throughput and interactivity of DeepSeek V4 Pro on NVIDIA B300 by 6.23-50.1%.

Quantized Mixture-of-Experts (MoE) services can hold several pre-materialized instances of one base model, but quantization damage varies sharply across requests and bitwidths. Because instance materialization and replica counts consume memory and require slow reconfiguration, we treat them as upstream provisioning decisions and study routing within a fixed resident pool. Within this fixed-pool boundary, we route each request to maximize modeled throughput under a class-level expected quality-degradation budget and measured instance capacities. To predict this request-specific risk, we introduce FWP (Fragility-Weighted Perplexity), computed from prompt tokens on a reference-instance prefill and calibrated to candidate-instance degradation. Underlying FWP is an exact two-expert affinity--fragility decomposition and a conditional multi-layer top-$k$ expansion whose bias, interaction, route-change, separability, and higher-order terms remain explicit. Using these calibrated risks, a window-level linear program yields a signed reduced-reward score that is KKT-consistent with the LP optimum under optimal prices and primal-feasible tie allocation. On 88 extended Qwen prompts, complete W2, W3, and W4 instances quantizing all 6,144 expert blocks incur mean $Δ$NLL of $0.9437$, $0.1832$, and $0.0513$. Under the same population and $τ=0.1513$, FWP allocation reaches a $1.284\times$ offline model-based multiplier versus $1.253\times$ for request-agnostic mixing and $1.000\times$ for static W4, an incremental $2.5\%$ relative FWP gain.

Dexterous manipulation requires coordinated multi-finger control and effective tactile feedback, yet learning these capabilities remains challenging due to the lack of large-scale real-world data and the difficulty of extracting effective representations from sparse tactile signals. We build a robot platform and teleoperation system to collect a 200-hour bimanual dexterous manipulation dataset with synchronized visual, tactile, and language annotations, comprising 10,576 trajectories across 65 tasks, 69.5% of which involve dexterous multi-finger manipulation. We further propose STAR, an integrated training recipe for vision-tactile-language-action (VTLA) models that addresses the spatial, temporal, and informational sparsity of tactile signals through visual-tactile joint pre-training, sparse-global tactile token representation, and sparse future tactile prediction. Trained on this dataset, STAR achieves a 61% average success rate across four real-world tasks with 100 post-training trajectories per task, demonstrating dexterous performance under task-specific post-training.

Clinical de-identification relies on accurately identifying personally identifiable information (PII). However, manually annotated datasets are costly to construct, while existing synthetic alternatives often provide limited details about their generation process or rely on relatively simple synthesis strategies. We introduce Meddies-PII-Dataset, a corpus of one million synthetic clinical documents spanning seventeen languages and nine PII labels. The documents are generated using attribute-conditioned prompts and validated through thirteen deterministic gates that enforce structural and annotation consistency. To evaluate the dataset's utility, we train Meddies-PII-Model, a BIOES token classifier, and compare it with existing PII extraction systems using exact-match entity-level F1. Meddies-PII-Model achieves the highest performance among the evaluated systems on all reported benchmarks, with a mean F1 of 0.827 across fifteen external benchmarks, compared with 0.658 for the strongest baseline. Upon acceptance, we will publicly release the dataset, benchmark suite, model, generation framework, and evaluation code to support research on multilingual clinical de-identification.

We demonstrate that a general-purpose agent can directly drive a physical robot throughout task execution without any task-specific or environment-specific training. We introduce Agent as Policy (AGP), which places task planning and execution under the agent's control. Given a task and a robot interface, the agent interprets visual evidence, writes executable programs, issues motion commands, and revises its actions in response to physical outcomes. This brings the agent's reasoning and programming capabilities into continuous interaction with the physical world. We study AGP across multiple real-world manipulation tasks spanning precision manipulation, dynamic motions, and deformable objects. These include assembly from human videos, block construction from goal images, die reorientation, targeted throwing, and bimanual towel folding. AGP achieves success rates of 100%, 100%, and 80% on three block construction configurations. These findings establish a path for general-purpose agents to act as robotic policies, extending their autonomy to physical manipulation through runtime reasoning, programming, and interaction.

Real-world Earth observation (EO) agents must translate high-level scientific questions into executable workflows to acquire observations, prepare data, perform domain computations, and derive conclusions from runtime evidence. Existing EO agents typically start from supplied observations, while benchmarks typically provide prepared inputs or candidate answers, leaving full-chain open-world EO execution largely untested. We present Earth-Agent-Pro, an execution-adaptive Plan-and-Execute framework using expert-authored skills to constrain planning and runtime tool use. Workflow-centered structured memory records planned steps, accepted evidence, and their dependencies, enabling repair of only the affected workflow suffix when runtime evidence invalidates a step. Separate large language model adapters use sequence-level supervised fine-tuning for planner workflow composition and node-level group relative policy optimization with locally verifiable rewards for executor tool-argument grounding. Earth-Bench-Pro instantiates 248 expert-curated task cores as 744 questions under three matched regimes. Its 248 Open-World Execution questions span RGB imagery, spectral observations, and remote sensing products, pairing high-level requests with runtime data requirements, executable trajectories, and open-ended answers grounded in execution evidence. With a shared GPT-5 backbone, Earth-Agent-Pro achieves 66.13% LLM-as-Judge accuracy, exceeding ReAct by 20.95 points in this metric and 24.44 points in Tools-In-Order. Joint adapter tuning raises Qwen3.5-9B LLM-as-Judge accuracy from 38.31% to 50.00%, an 11.69-point gain over the untuned configuration. Planning-only evaluation and execution with the reference workflow show that the adapters improve workflow composition and argument grounding, respectively. Code and datasets will be released soon.

Category-level in-hand manipulation of articulated objects is a formidable yet underexplored challenge for dexterous robotic hands. This difficulty stems from two core bottlenecks: first, controlling an object's internal degrees of freedom is tightly coupled with maintaining grasp stability on a free-floating base; second, acquiring diverse object models and functional grasps at scale is highly labor-intensive, yet vital for generalization given the system's sensitivity to initial configurations. In this work, we present ArtManip, the first category-level articulated in-hand manipulation method that generalizes across object instances and diverse initial grasps. For initial configuration construction, we develop an automated pipeline that procedurally generates diverse articulated objects and synthesizes task-oriented functional grasps. For policy learning, we propose a robust two-stage training strategy that incorporates articulation physics randomization, reward curriculum, and latent representation distillation to handle complex contact and joint dynamics during deployment. Extensive experiments across four object categories demonstrate that our policy generalizes to unseen instances and varied configurations in simulation, and achieves zero-shot transfer to 12 real-world objects featuring diverse shapes and joint mechanics.

Large language models are being organized into multi-agent systems with specialized roles, but whether such specialization produces distinct forecasts and whether subsequent synthesis improves utility remains unclear. In this study, we carried out a live, prospective evaluation over the final 56 matches of the information-dense 2026 FIFA World Cup, keeping a frontier foundation model constant while assigning two primary forecasting agents contrasting specialist roles: a quantitative specialist focusing on structured performance statistics and a news specialist focusing on current injuries, tactics and information from press conferences. Their forecasts were then reviewed by a separate critic before being combined by a meta-agent, resulting in a sequential four-agent model. Forecasts from the betting market served as an external benchmark. The news specialist obtained the highest mean probability-weighted Top-3 utility and matched the betting market in Top-3 exact-score hits. Nevertheless, the two specialist forecasters agreed on at least two of the three scorelines in 50 out of 56 matches, and the meta-agent never generated more than one scoreline outside the specialists' forecast set. These findings show that rapidly changing, unstructured information can provide a valuable forecasting signal alongside structured statistics, whereas adding critic and meta-agent stages does not necessarily create complementary information or improve on the strongest specialist.

Test case synthesis is crucial for evaluating and ranking programs generated by large language models (LLMs). However, constructing high-quality test cases remains challenging because reliable expected outputs are often difficult to obtain. We propose Confidence-Gated Transductive Test Generation (CoTT), which first uses an efficient inductive procedure and invokes transductive generation only when inductive confidence is low. This adaptive design improves output reliability while allocating extra computation only when needed. On code reranking benchmarks, CoTT outperforms prior baselines across the reported metrics while reducing cost relative to applying transductive generation to every input. These results show that confidence-based allocation of test-time computation provides a favorable efficiency-effectiveness trade-off with a single efficient LLM.

We consider an agent acting against a general non-Markovian environment. The agent maintains its agent states, but is free to choose a transition kernel across those states and optimize its state-feedback control policies. We study the bi-level agent design problem that optimizes the transition kernel and the policy it induces, given said kernel with offline data of observations and actions obtained via a behavioral policy. For general environments, we show that a soft $Q$-learning algorithm converges almost surely to the fixed point of a soft Bellman equation defined by the stationary averages that the behavioral policy and the chosen kernel induce, and we delineate what separates the resulting policy from an optimal one. In partially observed Markov decision problems, we analyze convergence properties of parametrized transition kernel design via zero-th order and Bayesian optimization techniques.

Comprehensive benchmark suites are essential for improving large language models (LLMs), but many widely used benchmarks are redundant, making evaluation unnecessarily expensive. Although recent benchmark compression methods (BCMs) can mitigate this cost, many strong BCMs rely on large collections of per-sample evaluation results from numerous LLMs to identify representative samples. Building such collections is also expensive unless they are already public, making these methods difficult to extend to newly released benchmarks. To address this challenge, we present ZipBench, a simple and low-cost BCM with theoretical error and rank-consistency guarantees. ZipBench evaluates only a small set of anchor LLMs, synthesizes pseudo evaluation results to broaden coverage, learns compact sample representations, and selects a small yet representative subset. Building on it, we create ZipBench Zoo, a collection of compact versions of 100+ benchmark proxies spanning text, multimodal, and agent tasks. These benchmark achieve mean absolute errors of 0.002--0.02 and average Spearman correlations of ~0.98 with the full benchmarks. Overall, ZipBench reduces the cost of both LLM evaluation and compact benchmark construction, lowering the barrier to broad LLM research for compute-constrained researchers. The code has been released in https://github.com/MilkThink-Lab/ZipBench.

We introduce AMDKernelVault, an open HIP and Triton kernel corpus and training framework for recent AMD CDNA GPUs. Existing LLM-based kernel agents are largely CUDA/NVIDIA-centric and often depend on repeated frontier-LLM calls for generation, reflection, and optimization. To address this gap, we develop HIPKernelGen and TritonKernelGen, agent-driven pipelines that transform PyTorch references into HIP or Triton kernels, compile and validate candidates under ROCm, and latency-profile them on AMD hardware. The corpus contains 62,153 execution-verified HIP kernel samples, 2,377 production-grounded ROCm Libraries QA entries, and 39,893 Triton kernels. We further train Qwen3-8B with supervised fine-tuning and execution-aware reinforcement learning as a demonstration of the corpus's utility. Under fixed evaluation budgets, it achieves the highest correctness among the compared models on PyTorch-to-HIP (34.0% Pass@1), TritonBench-G (33.2% Corr@3), and ROCmBench (41.94% Corr@3), but does not uniformly lead compilation or speed metrics. The corpus and documentation are available at https://huggingface.co/datasets/amd/AIG-Datasets, and the associated training and kernel-generation code is available at https://github.com/AMD-AGI/hip_kernel_llm_lab.

False wake-up activations remain a persistent challenge in conversational AI. Speech phonetically similar to a device's wake word can produce a syntactically valid and semantically coherent ASR transcript that the assistant incorrectly executes. Most existing systems make a single intent decision in isolation, without a mechanism to learn from recurring errors over time or adapt to individual users through personalized learning. We introduce the Feedback-Driven Adaptive Self-Correcting Inference Layer (ASCIL), a complementary post-ASR correction framework that re-evaluates wake-up intent before response generation by fusing acoustic embeddings, linguistic cues, device context, and patterns from past misclassifications. ASCIL interprets implicit signals, including hesitation, disengagement, and silence, and explicit signals, including cancellation and repetition, as automatically inferred, noisy behavioral indicators of potential misclassification. These signals drive online pattern updates without manual annotation, whereas the intentional/unintentional reference labels used for offline evaluation are human-annotated. It generalizes from prior errors, applies corrective adjustments at inference time, and continuously updates in parallel with natural-language execution. Evaluated on a proprietary dataset of 3,667 interactions with human-annotated intentional/unintentional reference labels spanning 14 acoustic and contextual conditions, ASCIL achieves 54.27% relative error reduction on a session-disjoint subset constructed from baseline failures, and up to 24.39% relative error reduction at threshold 0.90 on the issue-tagged evaluation slice. These gains are achieved while improving intentional acceptance rates, with a median added latency below 60 ms in the reported benchmark.

As enterprises modernize legacy monolithic systems to microservices, Large Language Models (LLMs) are heavily utilized for automated code translation. However, traditional vector-based Retrieval-Augmented Generation (Standard RAG) struggles to capture topological relationships. It fetches isolated chunks that sever inheritance chains, leading to high compilation failure rates. This paper introduces a Hierarchical Context-Resident Graph (HCRG) methodology to resolve these limitations. Our pipeline uses tree-sitter for Abstract Syntax Tree (AST) extraction, maps architectural edges into a Google Cloud Spanner Property Graph, and serializes this structure into a Gemini Context Cache for topological, parent-first code translation. We shift evaluation from naive text-overlap to a custom 7-metric Software Engineering framework. Traditional metrics like CodeBLEU (which scored 91% for both methods) effectively masked Standard RAG's structural failures behind syntactically plausible but broken code. Empirically, Graph RAG decisively mitigates dependency loss: API hallucination rates dropped from 56.4% to 16.2%, Dependency Resolution Quality improved from 34.8% to 65.9%, and Parent-Child Consistency rose from 26.7% to 45.5%. However, Graph RAG introduces specific trade-offs. The dense global context causes defensive over-engineering by the LLM, reducing Cyclomatic Complexity Consistency from 71.6% to 46.7%, and slightly degrades Docstring Preservation (67.0% to 61.0%). Ultimately, while trading code complexity for reduced hallucinations, Graph RAG provides a substantially more viable, architecturally sound path for automated enterprise codebase modernization.

Open-ended reinforcement learning often relies on rubric-based rewards for tasks without directly verifiable answers. Yet the policy and reward system form a dynamic feedback loop: as the policy optimizes the current reward, an initially useful reward system may become unreliable due to reward hacking or reduced response discriminability. The reward system should therefore evolve rather than remain fixed during training. Existing dynamic-rubric methods adapt evaluation criteria, but reward failures can also arise from scoring mechanisms or signal composition. We introduce EvoRS, a self-evolving RL framework that evolves the reward system from on-policy experience, representing it as an executable Reward-DAG. Specifically, an agentic designer updates this system from on-policy rollouts and reward traces to maintain train-time reliability. Across writing and roleplay, EvoRS achieves the best quality under all three judges, outperforming the policy by \(2.107\) and \(4.767\) points, respectively, while reducing reward hacking and coverage failures and preserving reward informativeness. Ablations confirm that a comprehensive fixed reward system cannot remain reliable in open-ended tasks and must evolve throughout training.

Vision-Language Models such as CLIP enable effective few-shot medical anomaly detection (AD) via strong image-text semantic alignment. However, their globally contrastive pretraining lacks explicit spatial supervision, limiting precise lesion localization. In contrast, Vision Foundation Models (VFMs) such as DINO learn spatially coherent patch representations via self-distillation and local-to-global consistency, better capturing fine-grained anatomical structures. Leveraging this complementarity, we propose Spatial-FAD, a spatial-aware few-shot medical AD framework that improves lesion localization by combining VFM spatial priors with CLIP semantics. Specifically, we introduce a VFM-enhanced adapter that injects a structural affinity prior derived from DINO into CLIP features. This structure-guided refinement encourages visual embeddings to better adhere to lesion boundaries while maintaining semantic alignment. To address the loss of spatial detail from patchification and the limited input resolution of CLIP, we adopt a sliding-window aggregation strategy. This generates high-resolution, spatially dense embeddings to further enhance localization granularity. Moreover, we introduce a prototype-enhanced support memory scheme to efficiently exploit the few-shot support set. This module stores compact prototypes for normal and abnormal patterns, reducing memory costs while boosting performance by fusing patch-to-prototype and image-text similarities. Extensive experiments on three benchmark datasets, including Liver CT, Retinal OCT, and Brain MRI, demonstrate that Spatial-FAD significantly outperforms state-of-the-art methods, especially in lesion segmentation. Notably, in the 4-shot scenario, our method achieves an average improvement of over 11.4% in Dice score and 1.8% in AUC. Code is available at: https://github.com/JuzhengMiao/Spatial-FAD.

Sensitive attributes such as age, income, and occupation can be inferred from user-generated content by aggregating indirect cues across many ordinary posts. LLM-based profilers can perform this aggregation automatically and with high accuracy, which makes large-scale personal attribute inference a major privacy threat. Existing LLM-based profilers, however, offer limited insight into which specific posts, concepts, and relationships made an inference possible, which is key to targeted privacy mitigation, i.e., redacting or rewriting only the few posts that actually leak an attribute, rather than perturbing entire histories. We introduce GraphProfiler, an auditable LLM-based profiler that represents each user's post history as a source-linked personal knowledge graph where nodes and edges trace back to the originating post and resolves attribute predictions to cited graph records and source texts. GraphProfiler reaches 86.7% attack success rate on the eight-attribute SynthPAI benchmark, within two points of strong text-only baselines, and 84.6% on PANDORA, while citing supporting evidence for over 98% of predictions. Our controlled ablation experiments provide evidence that the cited posts contribute to attack success, as removing them reduces the attack success rate substantially more than removing an equal number of random posts.

Social-deduction games such as Werewolf are increasingly used to evaluate LLM agents, but existing evaluations often rely on final game outcomes. We propose a belief-shift evaluation benchmark in Werewolf for analyzing communication skills through belief updating. Using LLM-played games, we annotate suspicion and accusation messages and measure how an observing village-side model's beliefs change after each message. We evaluate 40 open-weight LLM configurations on 1,224 annotated messages. Our results show that larger models better distinguish true wolves from villagers based on game history, but accusations still strongly influence their beliefs. Models become more suspicious of the accused target and less suspicious of the accuser, especially when the accuser is trusted, even if the accuser is wolf-aligned. Larger models better resist accusations from accusers they already distrust. Overall, our findings suggest that current open-weight LLMs up to 120B parameters still struggle to integrate accusation content with source trust in strategic communication. Our benchmark and code are available at https://rlg.iis.sinica.edu.tw/papers/werewolf-accusation-benchmark.

A world model asked what happens if an object is pushed at several speeds produces several futures. If the model has the object in mind, those futures agree about it: each implies the same mass and friction. The consistency checks now used to vet world-action models ask whether a model's futures agree with each other, and none of them knows any physics. We show that this is not enough, and what to do instead. IMPLY reads the physics each rollout implies by inverting a simulator and scores a set of rollouts by how well one object explains all of them, anchored to two calibration pushes the model has observed. In a controlled setting, self-consistency gives a perfect score to a model that ignores the object and always predicts a typical push; anchoring exposes it (AUROC 0.70 versus 1.00). On a real model, V-JEPA 2-AC adapted to the scene, the same thing happens. Given its own calibration pushes the model tracks the object (per-object correlation with the truth 0.91); given another object's, it does not (0.05). Self-consistency cannot tell these apart, preferring the right evidence on 52% of objects, chance level, while anchored disagreement prefers it on 73% and correlates 0.92-0.99 with the rollouts' error. Used to choose among candidate rollout sets, it comes within 0.003 of an oracle that sees the truth. A model that has internalised the wrong object is exactly as self-consistent as one that has internalised the right one; consistency has to be anchored to evidence.

LLM-as-a-judge protocols are commonly debiased by instructing judges to ignore presentation cues such as citation formatting, source labels, and evidence-display style. We show that this intervention can suppress bias while damaging the resolution of the measurement instrument. We introduce TraceJudgeBench, a diagnostic benchmark for auditing citation-like artifacts in RAG and agent-workflow evaluation, covering content-equivalent pairs, citation ablations, correctness conflicts, human-validated soft and moderate quality gaps, prompt-strength ladders, decoupled judging, and a controlled workflow-ranking probe. Across GPT-5.5, Claude Sonnet 4.6, and DeepSeek V4-Flash, stronger anti-citation prompts reduce worse-cited wins from up to 50.5% to 0%; yet some operating points already convert validated moderate-gap decisions into Tie before the strict stress-test endpoint, while correctness-conflict accuracy remains at or above 93.0%. A second, 50-pair FinQA moderate-gap construction reproduces the qualitative frontier, and open-weight Qwen2.5-14B-Instruct-AWQ and Gemma-3-12B-IT runs reproduce the central HotpotQA frontier. TRACE-style decoupling recovers 96.5-100.0% better-plain resolution across the reported settings. Human validation separates three meanings of Tie: correct equivalence Tie, calibrated soft-boundary Tie, and resolution-destroying Tie on validated quality gaps. We frame debiasing as a measurement intervention whose bias suppression, resolution retention, Tie cost, and protocol cost must be reported jointly. The supplementary artifact contains benchmark splits, prompts, raw judge outputs, validation summaries, and analysis.

Adaptive RAG often uses retrieval-time signals to decide whether another retrieval, reranking, or multimodal step should run. We ask whether these signals add routing value once the query itself is already known. Across document, audio, and video RAG, we compare matched query-only and query+retrieval routers while holding the optional actions, router family, training procedure, and evaluation fixed. On the held-out final evaluation, adding the tested retrieval signals does not produce a reliable routing improvement over the query-only baseline. Some retrieval signals are associated with whether a later step will help, but that predictability does not consistently lead to bet- ter RUN/SKIP decisions. The main lesson is therefore methodological: retrieval-state features should not be credited with routing value unless they improve over a matched query-only control. Our results do not show that routing or retrieval state is generally useless; they show that the incremental value of retrieval signals must be demonstrated rather than assumed.

Long-running LLM agents require memory mechanisms that maintain coherent internal states across interactions. We study a lifecycle-labeled memory setting in which write episodes provide lifecycle metadata during training, and phase-aware readout is used during evaluation. This setting reflects the need to distinguish information that should remain influential across future interactions from information that should affect only the current context. A mismatch between these lifecycles can cause temporary information to overwrite durable knowledge, leading to behavioral drift in persistent agents. Within this setting, we introduce \textbf{LifeFuse-Mem}, a lifecycle-aware neural memory framework that separates information according to its temporal commitment. LifeFuse-Mem uses dedicated memory components and lifecycle-aware updates to allow stable and transient knowledge to evolve locally without converting temporary context into durable state. On the controlled anti-overwrite benchmark, LifeFuse-Mem improves acquisition-controlled retention and reduces temporary overwrite; on two public long-memory benchmarks, it remains broadly competitive. These results suggest that explicit lifecycle signals can help diagnose and mitigate overwrite in compact online memory.

Due to the highly deformable nature of garments, training a generalizable policy for robotic T-shirt folding and unfolding remains a significant challenge. In this work, we present a large-scale synthetic dataset for robotic T-shirt folding and unfolding, covering 6 robotic embodiments, 1K T-shirts, 1K environmental assets, and 120K episodes with rich annotations, which can be used to train a wide range of manipulation policies. We first follow the FoldNet pipeline to generate a large-scale dataset of physically simulatable T-shirts with diverse appearances and annotated semantic keypoints. Based on these semantic keypoints, we then generate manipulation demonstrations for different robotic embodiments through a unified rule-based framework. We use these demonstrations to train visuomotor policies, and experimental results demonstrate that models trained solely on our synthetic data can achieve over 90\% end-to-end task success rates when directly deployed to unseen real-world environments and previously unseen T-shirts from arbitrary initial configurations. Project URL: https://pku-epic.github.io/FoldNetXX/.

Reinforcement learning is now the standard way to train large language model agents on long-horizon tasks, where dozens of interdependent actions precede a single sparse reward. Critic-free, group-relative methods such as GRPO suit this regime, but they broadcast one trajectory-level scalar to every step and cannot say which decision drove the outcome. GiGPO recovers a step-level signal by grouping time steps that share an anchor state, yet it merges the step- and episode-level estimates under one fixed weight, spending the same resolution on a pivotal branching decision as on a routine, near-deterministic transition. We argue that the right resolution is state-dependent, and propose GACA, a critic-free estimator whose granularity follows an uncertainty-based criticality proxy. GACA scores every step by the negative log-likelihood its own rollout already records, then blends the two advantages with a per-step weight that grows with that score, so the gradient places more weight on the fine-grained signal at above-average NLL and on the episode-level signal below it. We derive an exact risk decomposition for the implemented mixture and show that sufficiently small modulation improves on fixed mixing under positive directional alignment. A separate conditional result bounds local action-value variation using expected NLL, while an error-projection analysis characterizes when mixing adds value beyond scalar uncertainty reweighting. On ALFWorld and WebShop, GACA improves task success over GRPO and GiGPO at both 1.5B and 7B scales.

Lux AI Season 3 requires agents to act under partial observability, randomized episode level dynamics, and a best of five match structure that rewards both tactical execution and fast adaptation. We present HORIZON, a hierarchical agent that combines symmetry aware spatial perception, dual memory belief tracking, relic centric graph attention, information gain driven exploration, and an opponent conditioned policy mixture. HORIZON separates short horizon control from cross match meta reasoning, while auxiliary belief and world model objectives stabilize learning. Trained with PPO in a large scale JAX simulator, the resulting agent explicitly infers hidden game parameters and opponent style. Experiments show consistent gains in match win rate, episode win rate, adaptation gain, and league rating over strong recurrent and feed forward baselines.

Reinforcement learning with verifiable rewards is typically performed on-policy, keeping training data close to the current policy but limiting learning to trajectories that the policy can discover itself. Off-policy methods such as supervised fine-tuning, on the other hand, can leverage external knowledge beyond the base model's capabilities, but may suffer from large distribution shift. The key challenge is thus to expand exploration without sacrificing learnability. In this work, we introduce Minimal Intervention Reinforcement Learning (MInTRL), which expands the exploration frontier through sparse, local interventions in otherwise on-policy rollouts. During generation, a judge-intervention policy periodically reviews the current policy's output, replaces erroneous suffixes with short corrections, and immediately returns control to the policy. During training, MInTRL adopts a sequence-level advantage-regression objective that eliminates the need for importance sampling. We show that sparse, local interventions can substantially improve coverage beyond finite-budget on-policy sampling while preserving the overall on-policy nature of the resulting trajectories. Across math and code benchmarks, MInTRL consistently outperforms standard on-policy and off-policy baselines. Ablations show that MInTRL remains effective with self-intervention and across different judge policies, while performance peaks at moderate intervention intensity, highlighting the importance of intervening minimally. These results establish minimal intervention as an effective paradigm for enhancing on-policy RL.

Neural architecture search (NAS) evaluates candidate networks, but fully training enough architectures to rank an entire space is expensive. Zero-cost proxies score architectures at initialization, yet their ranking quality varies across search spaces. Learned predictors reduce evaluation cost but typically require fully trained labels or partial-training features for individual candidates. We introduce $\textbf{RiPPLE}$, $\underline{\textbf{R}}$anking v$\underline{\textbf{i}}$a $\underline{\textbf{P}}$refix-$\underline{\textbf{P}}$ropagated $\underline{\textbf{L}}$abel $\underline{\textbf{E}}$xtrapolation, which treats partial training as a source of labels for a small coverage set of anchors. RiPPLE trains these anchors to an early prefix, extrapolates their learning curves to surrogate labels, and propagates the labels over label-free architecture features. The early-training signal remains a label on the anchors rather than a per-candidate feature. Feature, readout, and encoding rules are selected without held-out accuracy and reused across search spaces. We evaluate the method on twelve benchmark cells from four search-space families and on the larger DARTS space. The results examine ranking quality, label efficiency, architecture selection, and the roles of readout, coverage, and propagation. RiPPLE provides a whole-space ranking from a fractional anchor-training budget, with comparisons interpreted under their respective evaluation and cost protocols.

Large language models (LLMs) are rapidly evolving from conversational assistants into agentic AI systems that reason, plan, invoke tools, maintain persistent memory, communicate with other agents, and execute multi-step tasks. At the same time, modern models exhibit substantially stronger native safety alignment than earlier generations on which many jailbreak attacks and defenses were originally studied. This shift raises a fundamental question: \textit{which established jailbreak-security findings remain valid in the era of modern LLMs and agentic AI?} We address this question through a Systematization of Knowledge (SoK) that reframes jailbreak security around the full agentic execution pipeline. We develop unified taxonomies of attacks and defenses spanning user interaction, planning and reasoning, memory, tool use, and inter-agent communication, and introduce a security--utility--efficiency evaluation framework that separates native harmful-prompt safety, adversarial jailbreak robustness, and agent-level security outcomes. We further conduct a controlled empirical study of representative attacks and defenses within a common agentic framework. Our results reveal three important gaps. First, strong native alignment does not imply robustness to adversarial jailbreaks. Second, defense effectiveness is highly model-, attack-, and component-dependent and can come at substantial cost in over-refusal, utility, and latency. Third, low final-response attack success can mask severe intermediate compromise: planning, memory, and tool interactions may remain unsafe even when the final response is successfully filtered. These findings motivate a shift from response-centric jailbreak defense toward cross-layer, execution-aware security that protects agent state, component transitions, and external actions while preserving practical utility and efficiency.

Foundation models, including large language models, vision-language models, and time-series foundation models, are increasingly deployed on embedded and edge platforms for CPS and IoT applications, where energy, latency, and memory are as critical as task accuracy. Existing benchmarking tools evaluate model capability in isolation, reporting accuracy assuming sufficient compute, while hardware profiling tools remain platform-specific and mutually incompatible. As a result, users lack a unified workflow for making deployment decisions across heterogeneous devices. We present HoliBench, a modular benchmarking and deployment toolkit that jointly characterizes accuracy, latency, and energy across platforms from single-board computers to GPU servers. Its platform abstraction layer calibrates cross-device measurement, and the toolkit supports multiple model modalities, inference engines, concurrencies, and existing evaluation harnesses. An interactive interface exposes constraint-aware configuration selection over a design space that is profiled once and reused across studies. Using HoliBench, we characterize 20 models across 7 device types, 3 quantization levels, 8 inference backends, and over 30 tasks, surfacing tradeoffs that existing tools miss: quantization reduces latency only on hardware with low-precision support, accuracy gains show diminishing returns relative to energy, and for autoregressive workloads, average inference power is approximately constant across output lengths. We further find that single-model profiles compose under sequential co-resident execution. In a multi-model CPS deployment, standalone profiles predict combined-pipeline latency and power within 1.2% and 2.5%, enabling deployment exploration without exhaustively profiling every pipeline configuration. We release HoliBench as open-source infrastructure for deployment-aware evaluation of foundation models.

Learning from trial and error is a promising way to improve language agents on complex tasks such as computer control. Reflexion introduced verbal reinforcement learning, which turns failed trials into text that guides later attempts without updating model parameters. We introduce VRL-Bench, a harness for fair evaluation of trial-and-error learning under finite trial budgets. Across three models on MiniWoB and WebShop, we evaluate updates from several prominent verbal-memory methods spanning Reflexion and later work: each improves observed success over memory-free retry in some settings but reduces it in others. Replay experiments show that using reflection can reduce success rates, revealing a trade-off between exploiting experience and continued exploration. We propose VEX$^2$, a verbal exploration--exploitation scheduler that uses a language model to jointly select policies and allocate the remaining trial budget. VEX$^2$ is the only evaluated update to achieve positive observed success-rate gains over retry in all six settings.

Cognitive Diagnosis Models (CDMs) play a pivotal role in personalized online learning. Traditional CDMs rely on discrete, ID-based embeddings to represent students, exercises, and concepts. This paradigm diverges from the nature of learner cognition, where knowledge is not stored and retrieved as isolated symbols. As a result, CDMs suffer from semantic limitations when new exercises or concepts appear. In this paper, we propose a Process-aware Language Cognitive Diagnosis (PLCD) framework that uses language-derived structures as cognitive priors and response records to calibrate student posterior states. PLCD leverages large language models (LLMs) to construct concept schemas and cognitive process graphs, and uses target-conditioned semantic memory to retrieve historical responses that are relevant to each target exercise. A process-grounded Language-to-Cognition Mapper with DA-MoE experts and process-level contrastive learning then maps the textual evidence into a unified cognitive space. Experimental results show that PLCD not only outperforms traditional baselines in predicting student performance but also exhibits strong cognitive transfer capabilities. These results connect the computational power of LLMs with the psychometric goal of measuring latent knowledge states, suggesting that structured language priors calibrated by response records can improve cold-start robustness and cognitive grounding.

This paper presents a novel approach to hexapod locomotion by evolving each leg's gait independently through a decentralized evolutionary algorithm. Using the Webots simulator and the Mantis hexapod robot, we optimize individual leg controllers without centralized coordination, allowing emergent behaviors to drive the development of efficient, coordinated locomotion. Our decentralized method is benchmarked against cooperative coevolution, demonstrating improved efficacy in generating stable and adaptive gaits while showing interesting emergent coordination. By enabling independent evolution of leg controllers, this method reduces the complexity of gait optimization and highlights the potential of decentralized strategies for scalable and adaptive robotic systems.

This work investigates the control strategies of complex whole-body robot teleoperation that coordinate active perception, bimanual manipulation, and navigation. We developed a hybrid control framework, combining the free-form and constrained control, for the whole-body teleoperation of the TIAGo mobile manipulator. We conducted a user study to explore people's control strategies under different task constraints such as limited time and low tolerance of errors. Our results highlight the effective use of coordinated control in improving task efficiency and reducing the risk of reaching individual joint limits. We discuss our results and their implications for designing future whole-body robot teleoperation systems.

Autonomous driving requires more than recognizing what is present in a scene: a planner must determine how road structure, surrounding agents, and their motion states should influence a future maneuver. Existing learning-based planners can capture these influences through latent scene features and trajectory decoders, but the relationship between environmental factors and candidate actions often remains implicit. This limits the ability to inspect, diagnose, or refine how scene context affects the safety of a predicted trajectory. Classical safety fields provide an explicit spatial representation of this relationship, but their risk shapes and relative weights are prescribed in advance and do not adapt to each scene. We introduce READ, a framework that learns an explicit, planning-aligned risk representation from complementary geometric and behavioral constraints. READ instantiates this representation as a continuous spatiotemporal field, enabling differentiable queries along candidate trajectories. The learned field connects scene understanding with action selection by encouraging predicted trajectories to align with low-risk regions, while retaining a differentiable interface for trajectory evaluation and refinement. READ integrates with both end-to-end planners and Vision-Language-Action models. Experiments on NAVSIM show consistent gains across matched end-to-end backbones and strong performance in a VLA setting; READ also achieves competitive results on NAVSIM v2. These results establish learned spatial risk as an explicit, adaptable representation for safe planning.

We present ORQA, a method for testing occupation-level knowledge in large language models. Prior methods either map abstract LLM skills to occupations via task definitions or utilize expert knowledge which is difficult to obtain at scale and expensive. ORQA complements both of these methods by connecting O*NET occupations to trusted occupation-specific websites (such as regulatory agencies, licensing bodies, professional organizations, and government publications) and converting these into source-traceable question-answer pairs. A combination of an automated pipeline and human review produces a set of high quality questions about occupations. The question set created via our method covers 116 occupations from all 21 major groups in the SOC, with 480 questions sourced from 187 different websites. Each question is designed to probe a real-world skill question that is relevant to the occupation in question. We test 15 state-of-the-art frontier and open-weight models via this method. Claude Opus 4.6, GPT-5.4 and Claude Sonnet 4.6 all perform the best at approximately 58-62% while smaller open-weight models achieve approximately 33-41% performance. Performance varies significantly across occupations. Healthcare-related occupations achieve the highest performance (78%) while Office and Administrative Support achieve approximately 40%. Performance on individual occupations (e.g. Sheet Metal Workers and Fish and Game Wardens) is essentially zero. We also find that open-ended questions and weighting by wage bill do not significantly affect the ranking of models on this benchmark. We believe that leveraging existing trusted occupation-specific information to test LLM knowledge in professional domains may be a scalable and useful method for evaluating occupation-level AI performance in the future. Results and data are available at orqabench.org.

In the VTaC benchmark 71% of ventricular-tachycardia alarms are false, but silencing a real one can delay recognition of a dangerous arrhythmia. We reframe alarm reduction as three-way triage (retain, suppress, or defer) and bound the decision this analysis treats as harmful: among suppressed alarms, the fraction that were genuine stays below a user-set budget with 95% confidence, under i.i.d. event sampling. Alarms sharing a waveform record are dependent, so the clustered analysis is a sensitivity check. On the official split a 5% budget certifies in all three seeds, suppressing 74.8% of false alarms while silencing 1.5% of genuine ones, at AUROC 0.953 and Challenge Score 83.33, numerically comparable to the strongest of the eleven published systems. Our central finding measures what multiplicity costs: the correction charges for every candidate, so a finer grid can certify strictly less. Under held-out calibration the 885-cell grid we declared certifies 1 of 15 fold-runs, while choosing the grid on a separate selection partition certifies 8. We project the calibration volume each budget needs, making an uncertifiable budget a design parameter. Finally, adding a learned reliability dimension to the policy grid did not sharpen the certified frontier.

SNAP-KG is a framework for assigning newly arriving entities to semantic communities in a growing knowledge graph (KG) using only their raw features, with no graph access and no retraining at inference time. It was evaluated on five multi-view benchmarks and a 2.4M-node OGB-WikiKG2 KG. In each of these datasets, at least one graph view is homophilous, meaning that connected nodes usually belong to the same class, and SNAP-KG performs well on all of them. This paper asks what happens outside that setting. We extend the evaluation to three heterophilous graphs (Texas, Wisconsin, Chameleon) and measure the edge homophily of every view. When no homophilous view is available, clustering quality drops sharply for both SNAP-KG and the transductive baselines used in its original evaluation. What decides this is the homophily of the relation, not the number of relations. Multi-view fusion still helps, but only when at least one homophilous relation provides a reliable foundation. The homophily assumption is therefore shared by the whole method family, not specific to SNAP-KG. We argue that heterophilous multi-view clustering is a separate research problem, outside the scope of this work. As future work, we outline how a heterophily-aware teacher could be distilled into SNAP-KG's projector to serve both homophilous and heterophilous KGs.

Long-term conversational agents must answer user queries by recalling information from extended dialogue histories, yet directly using the full history is costly and often unreliable, while compressed memory units may lose fine-grained evidence needed for question answering. Motivated by the reconstructive view of autobiographical memory, we propose CueMem, a cue-guided framework that treats extracted memory records as retrieval cues rather than self-contained evidence and reconstructs query-relevant dialogue context from their source turns. During memory construction, CueMem extracts fine-grained memory cues from dialogue turns and links each cue to its source turn. At query time, it retrieves query-relevant cues, maps them to source-turn anchors, and expands from these anchors over a turn graph that captures temporal proximity and semantic relatedness, reconstructing a compact evidence context from the original dialogue for LLM answer generation. Experiments on LoCoMo and LongMemEval show that CueMem consistently outperforms representative long-term memory baselines. Further analyses show that graph-based context reconstruction helps recover supporting dialogue evidence while reducing query-time input tokens and latency compared with the full-history LLM setting. These results highlight retrieval cues as an effective alternative to self-contained memory evidence for long-term conversational question answering.

Humanoid robots must traverse cluttered obstacle fields using onboard proprioceptive and visual observations, yet existing methods usually process multimodal observations without explicitly considering their different characteristics: proprioceptive observations are low-dimensional but governed by highly nonlinear robot dynamics, while egocentric visual observations are high-dimensional, noisy, and redundant. We propose DWMP (Dual World Model Policy), a framework that provides the actor with separate but complementary world-model representations for humanoid obstacle traversal. A Koopman-based dynamics world model lifts proprioceptive observations into a latent space where their temporal evolution is approximately linear, making the dynamics features easier for the actor to learn from. An RSSM-based visual world model compresses egocentric depth observations into compact stochastic states while preserving obstacle-related geometry. The student policy receives the fused latent representation for action generation, combining linearized proprioceptive dynamics with compressed visual perception. Experiments in simulation and on a Unitree G1 humanoid robot show that DWMP improves obstacle traversal performance over baselines and supports real-world deployment under randomized obstacle layouts.

Existing agent benchmarks mainly evaluate final task success or tool-call correctness, providing limited insight into whether agents can reliably diagnose and recover from intermediate execution failures. This limitation becomes particularly critical in multi-turn parallel tool-use scenarios, where errors may propagate across dependent branches and trigger cascading failures. We introduce ParaRecover, a process-level benchmark for evaluating error localization and recovery in multi-turn parallel tool-use agents. Built upon a fine-grained taxonomy of 14 error types covering planning dependencies, tool selection, and argument matching, the benchmark comprises 10,626 instances spanning two difficulty levels. To enable finegrained, process-oriented evaluation, we further propose the SDE rubric, which measures structural integrity, diagnostic reasoning, and evolutionary strategy during agent execution.Experiments across more than ten mainstream LLMs reveal that even state-of-the-art models still struggle with multi-turn error propagation,implicit tool-use failures, and precise replanning. Moreover, we demonstrate that the SDE rubric provides effective supervision signals for improving agents' reflective recovery capabilities. Our data and code are available at https://github.com/gbw206/ParaRecover.

In large-scale networked dynamical systems, local faults or disturbances may propagate through the interconnections and compromise the stability of the entire network. Partitioning the network into disjoint subsystems, or islands, limits this propagation by isolating unstable portions of the network from those that remain stable. This paper studies a two-stage controlled islanding and reconnection problem. In the first stage, the network is partitioned into stable and unstable islands. In the second stage, selected stable islands are reconnected to recover connectivity lost during islanding while satisfying stability requirements. We derive Lyapunov-based sufficient conditions for both stages. To address the coupling between the islanding and reconnection decisions, we construct an augmented graph on which the bases of a graphic matroid induce exactly the admissible islanding and reconnection pairs. We further show that the resulting optimization problem admits an equivalent nonincreasing supermodular formulation. Based on this structure, we develop a local-search algorithm with counterexample refinement and establish a performance guarantee for its inner loop with fixed sample sets. Numerical studies on linear networked systems illustrate the proposed procedure, examine its sensitivity to fault locations, and compare the proposed method with a mixed-integer linear programming benchmark in terms of solution quality and computational performance.

Affective computing has advanced wearable state inference, but on-device reasoning about whether, when, and how to intervene remains challenging. We present Affective Agent, a three-layer reference architecture for personalized intervention reasoning under uncertainty on wearable-class hardware. It combines a compact sub-billion-parameter language model with physiological evidence, context, and user history to decide whether, when, and how to intervene, without cloud dependency or per-user retraining. The architecture is organized into three interacting layers (perception, personalization, and reasoning), adapting to individual users through host-managed structured memory evolution rather than per-user weight updates. We instantiate Affective Agent in indoor environmental quality control and evaluate it on held-out, simulator-generated longitudinal scenarios spanning physiological variation, context, signal quality, and intervention history. Results show that memory-driven personalization and two-pass structured reasoning improve intervention decisions within this synthetic evaluation. By moving the decision layer on-device, this work demonstrates a path from wearable state inference toward closed-loop, personalized intervention on wearable-class hardware.

Traditional large language models (LLMs) are scoped to individual user sessions, limiting their knowledge to a single conversation and preventing them from learning user preferences that evolve over time. Existing agentic memory systems address this limitation but generally operate at the individual-user level, restricting the public knowledge that could be shared across users to improve downstream responses. We introduce AIM (Agentic Interoperable Memory), a unified, privacy-aware memory framework that enables multi-agent, multi-user LLM systems to persistently manage private and shared memory. AIM dynamically classifies information as private, scoped to one user and inaccessible to others, or public, accessible to all users. It enforces index-level access controls so that private memories are retrievable only by their owner, protecting sensitive data while allowing beneficial shared knowledge to improve coordination and consistency. We also introduce MUMBench (Multi-User Memory Benchmark), a dataset of multi-user interactions containing private and shareable information across four domains. To our knowledge, MUMBench is the first public dataset designed to evaluate multiple memory operations, including retrieval, creation, update, and deletion, in a multi-user environment. Across three independent runs on MUMBench, AIM achieves 96.0% visibility classification accuracy, 58.8% strict operation accuracy, and 70.5% state-aware operation accuracy.

A central question in LLM reasoning is whether reinforcement learning (RL) instills genuinely new capabilities or merely reshapes how existing knowledge is expressed during inference. Building on the distribution-sharpening hypothesis, which holds that RL reallocates probability mass toward high-reward trajectories already latent in base models, we ask: can we unlock those latent paths without costly RL fine-tuning? We present Decision-Flow Sampling (DF-Sample), a training-free, data-free inference-time framework that constructs a hierarchical reasoning tree, scores terminal nodes for quality, and back-propagates utilities to inform each intermediate branching decision. Unlike conventional sampling strategies that make purely local step-wise choices, DF-Sample performs explicit global trajectory evaluation before committing to a path, recovering high-quality but low-probability reasoning chains that standard decoding overlooks. On GPQA, DF-Sample achieves 45.6% accuracy, surpassing power sampling (38.9%) and GRPO (39.9%), showing that a training-free method can outperform a trained one. Across three models and four benchmarks, DF-Sample consistently outperforms baselines, indicating substantial latent reasoning potential in pretrained base models.

Collecting high-quality robot data remains a fundamental challenge for training robot foundation models. Task and motion planning (TAMP) offers a scalable way to generate demonstrations, but our experiments show that raw TAMP trajectories provide surprisingly little benefit when used to fine-tune pretrained vision-language-action (VLA) models, despite successfully solving the target tasks. We hypothesize that this failure arises from a behavioral distribution mismatch between planner-generated trajectories and the data used to pretrain the VLA. To address this mismatch, we introduce DATAFARM: Distribution-Aligned Task And motion planning for Fine-tuning A Robot foundation Model, an approach that incorporates the pretraining distribution directly into TAMP trajectory generation. DATAFARM aligns generated trajectories with the pretraining data in robot joint configurations, motion style, and temporal execution profiles. We evaluate DATAFARM on three tabletop manipulation tasks that TAMP can perform and a cloth-folding task beyond the capability of TAMP. DATAFARM achieves an average success rate of 56.7%, substantially outperforming raw TAMP (8.3%) while approaching human teleoperation (61.7%). On Deformable Object Manipulation, which is outside the fine-tuning distribution, the fine-tuned model retains 85% success, compared with 90% for the pretrained model. These results show that aligning planner-generated demonstrations with the pretraining distribution can make TAMP an effective source of data for VLA fine-tuning. Website and code: https://prpl-group.com/datafarm/

Small models are made more capable through distillation from a larger one that shares their tokenization scheme. However, do distilled byte and token models behave similarly in terms of scaling trends as compute and data increases? To enable this comparison, we introduce two variants to efficiently convert token logits to Byte Logits: 1) approximate: Marginalize-It, and 2) exact: End-Of-Token. We then present the first large scale study of overtraining decoder-only dense transformer models varying two dimensions simultaneously: the tokenization scheme (Tokens, Bytes, Bytes w/ eot) and the training objective (Distillation vs. Cross-Entropy), sweeping layer-parameter-matched models with roughly 1 billion parameters up to 1 trillion bytes of data. Across eight benchmarks spanning three categories: Multiple Choice QA, Language Generation, and Machine Translation, we find that Token-1B models outperform byte models (End-Of-Token-1B and Bytes-1B) in the low-FLOP regime but eventually plateau; byte models start worse yet surpass Token-1B models with more compute, reaching a higher downstream task performance ceiling. Extrapolating the average top-1 error vs. validation BPB scaling laws predicts that, asymptotically, distilled End-Of-Token-1B outperforms distilled Token-1B by up to 4%. They are also far more data efficient, matching the performance of distilled Token-1B using only one-sixth of the training data. Moreover, by operating over a small vocabulary of 256 bytes instead of on the order of 100K tokens, they circumvent the need for top-k truncation during logit dumping, while also reducing logit storage costs to roughly one-fifth. Finally, our downstream performance scaling laws predict that our distilled End-Of-Token-1B models asymptotically surpass the Llama 3.2-1B, Gemma-3-1B-pt, and Gemma 2B models on averaged downstream tasks by up to 6.5%, 8.1%, and 2.1%, respectively.

Vision-Language Navigation (VLN) in unseen indoor environments is useful in real-world robotics, where an agent must follow natural-language instructions, locate objects, and answer spatial questions without a pre-built map or fixed object vocabulary. Multimodal vision-language models (VLMs) provide strong open-vocabulary grounding and zero-shot reasoning, but struggle to emit reliable metric quantities such as range, bearing, and comparative spatial relations directly from images. Existing approaches address this by folding geometry into hand-engineered pipelines or asking models to output waypoints, requiring changes to the control stack for different robots, tasks, or vocabularies. We introduce AnchorVLN, an open-vocabulary VLN system built on a simple rule: the VLM proposes semantics; geometry decides metrics. It is realised as EMBODIED-NAV-MCP, a Model Context Protocol (MCP) server driven by a VLM agent through a compact set of callable tools. Since no tool accepts distance in metres or bearing in radians, the schema enforces the semantic-geometry boundary without modifying the downstream autonomy stack. We benchmark both tasks of the CMU Vision-Language Navigation Challenge 2026: 30 instruction-following questions over 15 scenes and a frozen 45-question object-reference set. The full system achieves 64.4 percent on instruction following, dropping by 13.3 percentage points without controller modeling (t = 2.77). On object reference, geometric anchoring clears the challenge overlap threshold on 10 of 45 questions, versus 0 of 45 for direct coordinate estimation, reducing median center error from 3.37 m to 2.48 m.

World models let agents plan by predicting the consequences of their actions, but changes in the environment can make them inaccurate. We study the problem of adapting a world model to an unknown test-time environment, drawn from a known environment family, using only a few episodes of interaction. Existing approaches trade off computational cost against expressivity, i.e., the range of models a method can produce. For example, in-context learning is computationally cheap but limited in expressivity, and gradient-based adaptation is expressive but computationally expensive. We present CLAW (Context-conditioned Low-rank Adaptation of World models), which addresses this tradeoff by using a hypernetwork to generate low-rank (LoRA) adapters at test time. During pretraining, we simulate adaptation to a variety of environments and jointly train the hypernetwork and base world model. At test time, we freeze the base model and use a forward pass of the hypernetwork to generate adapters from a small batch of test-time transitions. We evaluate CLAW in locomotion and manipulation environment families that vary in dynamics, embodiment, and reward. We show that, using only seconds of test-time data, CLAW outperforms gradient-based adaptation and in-context learning during online adaptation. We also show that CLAW avoids overfitting in data-scarce regimes, that its advantage comes from the expressive adapters rather than context conditioning, and that pretraining the hypernetwork jointly with the base model outperforms training it post hoc.

Retrieval-augmented generation (RAG) can improve knowledge-intensive question answering, but the first design choice is easy to overlook: how should the source corpus be partitioned into retrievable units? Fixed-size chunks often return long passages whose relation to the question is only implicit. We introduce EAR, an Entity-Aware Partitioning approach for multiple-choice question answering (MCQA). EAR extracts normalized surface anchors from the question, answer options, and corpus; retrieves local windows around matching corpus anchors; and can attach a larger parent passage through an extractive summary. We evaluate EAR on a cleaned Massive Multitask Language Understanding (MMLU)-style subset of 153 questions selected by an automatic corpus-support heuristic and using decontaminated public textbook text. Across same-protocol top-k = 3 and top-k = 8 sweeps with Mistral, Gemma, and DeepSeek, EAR entity-window reduces retrieved words by 37.5-40.2% relative to chunks. Observed accuracy changes are +5.2, +1.3, and -3.9 points at top-k = 3, and +5.9, -3.3, and -4.6 points at top-k = 8; none of the entity-window differences is statistically significant. The scoped contribution is methodological: EAR provides a compact and inspectable retrieval unit, while its rule-based anchor extractor remains domain-specific and requires separate validation before transfer.

Designing effective chemotherapy regimens is hindered by tumor heterogeneity and drug resistance, which complicate the deployment of patient-specific model-based optimal control across diverse populations. We develop and compare closed-loop deep reinforcement learning (DRL) dosing policies with continuous (TD3) and discrete (DQN) action spaces trained on a high-dimensional heterogeneous tumor model. The DRL policies are benchmarked against a Pontryagin's Maximum Principle (PMP)-derived open-loop benchmark. We assess generalization under parametric heterogeneity using a 100-patient virtual cohort with plus or minus 10 percent uniform perturbations in growth and drug-sensitivity parameters. Across this cohort, TD3 achieves higher average tumor reduction, while DQN yields tighter inter-patient dosing consistency, revealing a clear efficacy-consistency trade-off in this study. Our simulations assume full observation of all tumor subpopulations; translation to sparse and noisy clinical measurements will require partial-observability formulations and/or state estimation. Overall, the results show that simulation-trained DRL can learn state-dependent feedback dosing policies that complement open-loop optimal control benchmarks.

Large language models (LLMs) can generate biomedical hypotheses, but it remains unclear whether they truly reason from scientific evidence or simply produce convincing-sounding ideas. To study this, we combine three major biological databases: the Kyoto Encyclopedia of Genes and Genomes (KEGG), Rhea, and UniProt, into a unified biochemical knowledge graph and construct a benchmark of 550 paths connecting enzyme sources to rare disease endpoints, yielding 13,200 hypotheses from six LLMs under four conditions varying the biological information each model receives: source enzyme only, full biological path, or source and disease endpoint only. Hypotheses are scored using an expert-derived five-criterion rubric on a 1-5 scale per criterion. We find that models given both the source and disease endpoint often produce the highest-scoring hypotheses, showing that LLMs can generate compelling ideas from minimal information. However, these hypotheses are less grounded in the evidence. In contrast, models given the full biological path generate hypotheses more consistent with known mechanistic relationships. We call this evidence-disciplined reasoning. To confirm this effect, we shuffled intermediate path steps while keeping endpoints fixed. Evidence grounding dropped significantly (delta = -0.793, p < 0.001), confirming models genuinely used path structure during reasoning. Our findings show that knowledge graphs support hypothesis generation in two ways: they identify biological endpoint pairs absent from the literature, and their mechanistic paths guide how LLMs reason between them.

The climate literature has grown faster than review teams can read it. That gap matters most for a concept like the environmental social tipping point, the threshold at which a small change triggers rapid, self-reinforcing change in a social system. Evidence of this kind of shift is usually contained in one or two paragraphs within a longer document. As a result, existing text mining tools-which categorize entire documents by topic or highlight isolated claims-leave an expanding set of important evidence without any systematic method for discovery or organization. This paper presents an open and modular transformer-based framework that detects and structures social tipping point evidence at the passage level. The framework joins five components into a single deployable workflow: a DistilBERT boundary splitter for segmentation, an iteratively augmented RoBERTa classifier for detection, a Mistral 7B model that rewrites each detected passage for clarity, a LLaMA 3.2 3B model that rates the passage against five published social tipping point criteria, and a Milvus vector store for semantic retrieval. The system is wrapped in a Streamlit interface backed by MinIO object storage. Evaluated on a 163-passage benchmark labelled by GPT-4.1 and a 51-passage set reviewed by experts, the splitter surpassed three competing methods on a nine-metric composite score (6.137). The tuned RoBERTa model achieved 71.4 percent accuracy with a Cohen's kappa of 0.337 on the full benchmark, and 87.5 percent accuracy with a kappa of 0.742 on passages with labels, outperforming both a climate-focused model and untuned language models.

This work develops a Lyapunov framework for a broad class of arbitrary-order sliding-mode algorithms in multi-agent systems. We introduce abstract homogeneous chains, a class of nonlinear error systems characterized by common convexity and homogeneity properties. For this class, we establish global finite-time stability for arbitrary order, construct a homogeneous Lyapunov function, and derive a recursive optimization-based gain-proposal procedure. The framework addresses several gaps in existing dynamic average consensus and distributed differentiation results: it provides a recursive numerical optimization-based gain-proposal procedure for EDCHO at arbitrary order, extends REDCHO convergence from local to global, and provides arbitrary-order numerical gain-proposal rules for leader-follower distributed differentiation, previously available only at first order. It also provides a new arbitrary-order observer for multi-leader affine formation tracking with global finite-time convergence.

Diffusion-based robot policies have become widely used in robotic manipulation, where they are typically trained with behavior cloning. However, policies trained purely from demonstrations are limited by the quality and coverage of the available data. Reinforcement learning can further improve the performance of these pretrained policies through interaction. A common approach is to use policy-gradient methods that formulate diffusion-policy fine-tuning as an outer environment MDP together with an inner denoising MDP. However, existing methods typically assign the same environment-level credit to all denoising steps used to construct an action chunk, without distinguishing which intermediate decisions contributed most to the final return. We introduce Denoising Intermediate Advantage (DIA), a policy-gradient method that learns a value function over partially denoised actions and uses it to construct a denoising- level advantage for each step of the generative process. DIA com- bines this inner credit signal with the standard environment-level PPO advantage, providing state-dependent credit throughout the denoising chain. Across Robomimic, FurnitureBench, Franka Kitchen, and D3IL, DIA consistently improves final performance over existing diffusion-policy fine-tuning methods. Beyond final reward, DIA reaches successful states more efficiently and can shift farther from the pretrained behavior distribution, enabling it to discover more effective and efficient task-level strategies and subtask sequences that baseline methods fail to reach.

Inference-time power sampling via Sequential Monte Carlo (SMC) can substantially improve large language model (LLM) reasoning without requiring post-training. However, many existing SMC approaches rely on equal-weight resampling, which can aggressively prune low-weight trajectories, discarding potentially correct reasoning paths and degrading the genealogical diversity of the search space. To address this, we introduce Chopthin-Consensus Power Sampling (CCPS). Our method applies the Chopthin resampler to LLM decoding: rather than equalizing weights and forcing unnecessary particle duplication, it enforces an upper bound on the ratio between the largest and smallest weights and carries the unequal weights forward. This targeted intervention preserves a richer set of distinct reasoning paths, keeps the weighted SMC approximation unchanged in conditional expectation, and guarantees a lower bound on the post-resampling effective sample size (ESS). To fully exploit this enriched population, we employ a semantic-majority selection mechanism that merges token-identical final trajectories, clusters semantically equivalent answers, and returns the answer supported by the largest number of distinct trajectories. Evaluating across three open-weight models and five reasoning benchmarks, we show that Chopthin increases oracle coverage in 13 of 15 settings. Combined with semantic-majority selection, CCPS matches or exceeds the final-answer accuracy of the Power-SMC baseline in 14 of 15 settings, delivering absolute gains of up to 10.6 percentage points. These findings demonstrate that diversity-preserving resampling and diversity-aware selection are complementary mechanisms for training-free LLM reasoning. Code is available at github.com/MinooAhmadii/chopthin-consensus-power-sampling.

Question-answering often requires reasoning across multiple connected facts rather than retrieving a single isolated relation. Knowledge graphs (KGs) provide a structured way to represent such facts, but training large language models (LLMs) only on isolated KG head-relation-tail triples may limit their ability to learn the surrounding context needed for multi-hop reasoning. In this work, we propose a context-augmented training framework for multi-hop question-answering. Although generally applicable, we validate the framework in the context of disease-specific KGs, extracted using a reliable KG extraction framework called GraphMERT, for Gastroparesis and Diabetes. For each primary KG triple, we attach supporting triples extracted from the same source text chunk to form a context graph (CG). This creates two supervision settings: KG-grounded supervision, which uses only the target KG triple or path, and CG-grounded supervision, which uses the target KG triple or path together with supporting context triples. We train the Qwen3-14B model using supervised fine-tuning (SFT) under both settings, producing KGModel and CGModel variants. To strengthen the lower-hop factual foundation of the models, we introduce an LLM-judged, history-aware adaptive repair pipeline that identifies unresolved one-hop failures, continually fine-tunes on targeted repair examples, and removes or quarantines problematic noisy triples. This repair stage enables the models to reach 100% accuracy on the cleaned retained one-hop validation sets. Finally, we employ reinforcement learning (RL) using lower-hop question-answer items and evaluate generalization on harder 3-hop, 4-hop, and 5-hop tasks. Across both diseases, context-augmented supervision consistently improves multi-hop performance over KG-only supervision. RL initialized from repaired SFT checkpoints yields larger and more stable gains.

Self-improving agent workflows create an audit problem when the same controller can change both its behavior and the conditions under which that behavior is judged. We present GuardrailLoop, a simulation-based testbed that makes three operational contracts jointly testable: preservation of human-defined policy, compute accounting at every recorded execution prefix, and recovery of a specified scientific state after crashes. A hash-pinned policy fixes goals, scope, evaluation identity, budget, and release conditions; machine-directed evolution is restricted to a code-owned feature catalog and bounded knobs. The contribution is an executable boundary and an evaluation protocol that separates useful adaptation, state recovery, and repeated execution. In a paired 50-seed 2 x 2 study, round-stage growth changes target attainment by +1.00 and restricted mean compute to target by -56.97 simulated GPU-hours (95% paired-bootstrap interval [-58.91,-54.70]); idle growth has zero measured utility effect. Across 240 enumerated crash injections, all runs recover the defined outcome, but only 210 preserve the normalized trace: 30 pre-commit crashes repeat a planner call. Resource-drift, kill-switch, integrity, and output-guard matrices satisfy their specified checks. These findings show why successful outcome recovery is insufficient evidence of exactly-once execution. They establish conformance within one calibrated deterministic testbed, rather than general safety or real-world self-improvement.

Comparing and selecting task-oriented LLM agents increasingly relies on a low-cost offline evaluation gate: persona-driven LLM user-simulators converse with each candidate, an LLM-as-a-judge scores the transcripts, and the higher-scoring agent is promoted. We introduce GAUGE, a reusable offline protocol that measures whether this gate's ranking matches a grounded verifiable reward across 25 agents from six providers on the $τ^2$-bench and SimulatorArena benchmarks, separating two kinds of evaluation validity that release practices conflate: ranking validity and construct validity. First, a satisfaction-success gap: satisfaction carries essentially no information about task success, as conversations rated satisfied by our blind panel are decorrelated from actual success, with 57.5% of them failing the customer's task, a pattern consistent across five rater populations, both benchmarks, and every subjective dimension we rated. Second, while the gate's ranking is robust across the broad capability span, it loses resolution among the near-equal strong agents: this decision-disagreement rate jumps from $<$1% on wide-reward pairs to 31% on close pairs. The gate is thus human-validated yet mis-anchored. As a remedy, we propose a calibrate-then-trust cadence in which a judge-free completion bit is a zero-cost tripwire for truncation regressions.

This paper presents a battery-aware predictive trajectory-planning and control framework for multirotors operating under spatially localized disturbances. Candidate trajectories are evaluated through closed-loop vehicle--motor--battery propagation, allowing disturbance-induced control demand, electrical energy, battery evolution, and terminal-voltage-dependent actuator capability to enter the planning process. % A reduced-order battery model is numerically benchmarked against an independently implemented Simscape equivalent-circuit reference, with a power NRMSE of $0.64\%$ and a cumulative-energy discrepancy below $0.7\%$. % In a $150$-s, $640$-m mission containing three disturbance regions, the selected trajectory reduces electrical energy consumption by $7.46\%$ and position-tracking RMSE by approximately $72\%$ relative to the disturbance-aware fixed-reference baseline. % Planner ablations show that battery-dependent terms are nonbinding at nominal SOC but alter the selected trajectory under a depleted-battery stress condition. % Execution with multiple feedback controllers further demonstrates that controller selection changes the tradeoff among tracking accuracy, energy consumption, and actuator utilization. % The results demonstrate the benefit of accounting for predicted closed-loop energetic and battery--actuator consequences during trajectory selection.

Draft-verify-revise is a common LLM orchestration pattern for scaling inference-time compute. One LLM drafts, a second critiques the draft and provides feedback, and a third uses that feedback to revise the draft into the final output. As context cascades between stages, LLMs at different stages can resolve a context-dependent expression such as "previous" differently. When that happens, the expression undergoes a deictic shift, a change in what it refers to. This phenomenon was studied with a synthetic dataset of 10 base examples, each rendered in three conditions. Holding the shared components constant, the conditions varied whether the draft stage LLM (the assistant) or the verify stage LLM (the grader) resolved the expression correctly, and how much independent reasoning the revise stage LLM (the meta-evaluator) needed to determine which reading was correct. Six models from three providers were tested across 21 reasoning effort configurations using e-values for sequential testing, in a primary experiment and an ablation experiment that removed error classification labels from the grader's feedback. A separate LLM analyzed the meta-evaluator's stated rationale for each wrong verdict. Balanced accuracy (the unweighted mean of sensitivity and specificity) ranged from 0.156, below chance, to near-perfect. GPT-5.2 rose from 0.156 without reasoning to 0.942 at its highest reasoning effort level, while Gemini 3 Pro stayed above 0.94 at every level. Gemini 3 Pro at low reasoning effort outscored GPT-5.2 at xhigh reasoning effort for roughly 5% of the cost per trial. When the meta-evaluator erred, it tended to rely on surface cues rather than operational reasoning. Context engineers implementing draft-verify-revise pipelines should be wary of deictic shifts and make the intended referent explicit at each stage.

Prompt-policy editing offers a practical way to improve agents that synthesize executable workflows without updating the underlying model. However, persistent prompt editing has two coupled properties. First, edit locality does not imply effect locality: an edit confined to one policy segment can ripple through downstream execution, altering behavior beyond the edited segment. Second, edit effects are composition-sensitive: edits that work in isolation can interfere after composition, causing one or both to lose their benefit or become harmful. Persistent adaptation must therefore support two distinct decisions: identifying where the policy should change from execution feedback, and determining whether the resulting edit remains safe to persist after composition. To address these challenges, we introduce RIPPLE (Replay-Informed Persistent Policy Localization and Editing), which separates where an edit is made from whether it remains safe after composition. It diagnoses failed trajectories, maps each actionable failure to a predefined policy segment, and restricts the correction to that part of the policy. RIPPLE then evaluates candidates against the same iteration-start policy to compare their isolated gains, before replaying promising edits after previously accepted updates to expose downstream effects and interactions. Only edits that remain safe under composition are retained. We evaluate RIPPLE on Flow-HO, a synthetic held-out benchmark for executable workflow synthesis. RIPPLE improves validation success by up to 23.1% and yields positive gains on two additional frozen language-model backbones, while maintaining edit efficiency and low execution cost. Targeted interaction analysis further demonstrates both properties: a segment-local tool-use edit changes downstream resource resolution and validation, while an edit beneficial in isolation becomes harmful after composition.

Factual hallucination in closed-book question answering is often treated as a coverage problem: a model fails because the relevant fact is absent from its internal memory. This view misses a second source of error. Even when a fact has been observed, finite memory may force it to be stored only approximately. We study this effect through a simple coverage--compression model of factual recall. We consider an unstructured question-answering task with $N$ possible queries and $K$ possible answers. A learner observes $M$ training facts, compresses them into at most $B$ bits, and answers uniformly drawn test queries without retrieval. For a uniformly random ground-truth mapping, we prove $\mathcal{E} \geq \frac{M}{N}δ^\star\!\left(\frac{B}{M}\right) + \left(1-\frac{M}{N}\right)\left(1-\frac{1}{K}\right)$, where $δ^\star(r)$ is the inverse rate-distortion function of a uniform $K$-ary source under zero-one loss. The two terms separate compression distortion on observed facts from missing coverage on unobserved facts. The bound gives a compact way to reason about selective memory, forced compression, structure, retrieval, abstention, and long-context organization. We study the predicted signatures with theory-implied simulations and controlled fact-injection probes in modern language models that vary fact load and effective trainable memory. The result is not a complete theory of hallucination, but an information-theoretic account of a separable failure mode: lossy recall of observed facts under finite memory.

Development and humanitarian organizations produce and support surveys, administrative registries, and other data resources to inform research, policy, and operations, yet systematically identifying where these datasets are referenced remains difficult. Such references are dispersed across research papers, project documents, humanitarian reports, and other unstructured text, limiting both the ability to trace data use and to identify potential gaps in data availability or dissemination. We present a weakly supervised framework for adapting dataset extraction to forced displacement and Fragile, Conflict, and Violence (FCV) documents without first constructing a large manually labeled training corpus. A lightweight model trained on general research literature generates candidate dataset mentions from unlabeled domain documents, which a frontier large language model (LLM) reviews in context, validating or rejecting candidates and correcting their extraction boundaries. The resulting annotations are supplemented with targeted synthetic and contrastive examples and used to fine-tune the lightweight model for large-scale extraction. We evaluate the resulting model on an independent gold-standard benchmark of 1,706 text passages spanning research, humanitarian, and operational documents. Across the full benchmark, the model achieves 74.1\% precision and 70.5\% recall at the mention level; among passages containing dataset references, precision reaches 89.5\%. At the passage level, the model achieves 88.2\% accuracy and 88.6\% specificity in distinguishing passages with dataset references from those without them. These results demonstrate a practical approach for constructing domain-specific supervision when labeled data are limited, and provide a technical foundation for larger-scale analysis of data use and potential gaps in the displacement data landscape.

Slender rod insertion arises in precision manufacturing, where millimetre scale diameter and tight clearances demand accurate perception and control. Conventional peg-in-hole methods assume a rigid object whose tip pose is fixed relative to the gripper. This assumption breaks down for a high aspect ratio rod, which can bend during manipulation, making its tip motion dependent on the rod configuration, grasp, material properties, and contact. We present RodForesight, a learning framework that factorises the task into two stages: 1) coarse approaching, which uses visual servoing to map diverse initial configurations into a compact near hole hand-off region; and 2) predictive insertion, which performs fine alignment and completes the insertion. It is worth noting that the two stages can be wrapped into an end-to-end design. During insertion, a diffusion policy generates candidate action chunks, while an action conditioned world model predicts their effects on rod-hole alignment. This pre-execution evaluation enables RodForesight to select the best action chunk based on predicted tilt and radial errors before execution. Experiments investigate the performance of different stages and the end-to-end setting, where RodForesight improves the success rate from 88.9% to 96.7%, compared to baseline methods such as diffusion policy.

Checking Quran recitation from an ASR transcript requires distinguishing unresolved mistakes from repetitions, repairs, opening formulas and accepted spelling differences. We report a completed human annotation of 100 production recording cases: 348 scored units and 162 localized events across ten combined labels. An executable evaluator scores labels and word positions together. A plain diff reaches label-aware F1 0.525 and localization F1 0.826; adapted production cleaner/alignment components reach 0.518 and 0.786, with exact-span F1 0.505 for both. Correcting the adapter's word coordinates recovers all five annotated repetition events, showing why annotation interfaces must be checked before interpreting baseline failures. In a preliminary pilot, eight single 20-minute runs across three coding agents and eight models span label-aware F1 0.143 to 0.892: seven land far above every baseline, and one collapses below the naive diff from a missing normalization step. Across the six, 970 of 972 gold-event instances draw an overlapping prediction, so what remains is not detection but convention: span extent, and the labels whose boundary is stipulated by adjudication rather than visible in the text. Seven of 162 events defeat all six same-day runs, five of them one orthographic rule, and the strongest run still misses the same ones. No run annotated before building, so the pilot measures the algorithm half of the task only.

Mobile manipulation requires perceptual evidence at different spatial scales for base motion and arm control, while the two action modalities remain kinematically coupled. Existing policies often employ specialized action generation for different subsystems but condition heterogeneous action branches on a shared perceptual representation, leaving subsystem-specific perception-action correspondence implicit. We present MoPA, a framework that aligns perceptual conditioning with mobility and manipulation while preserving coordination at the action level. Dual Perceptual Streams employ two mutually masked query banks to extract separate perceptual representations from a shared vision-language context. Perception2Action Adaptation jointly updates each query bank and its corresponding action stream at every layer of a structured Mixture-of-Transformers decoder, while enabling information exchange between the two action streams. Coupled conditional flow matching learns a joint vector field for coordinated generation of both action chunks. On the ManiSkill-HAB benchmark, MoPA achieves state-of-the-art performance across all three task suites. Across four real-world tasks, MoPA achieves a mean full-task success rate of 76.3%, outperforming the best baseline by 12.5 percentage points. Ablation studies and further analyses validate the effectiveness of the proposed design. Website is available at: https://mopa-policy.github.io/.

This work considers multi-agent coordination with arbitrary information networks among the agents using a game-theoretic approach. A system designer aims to assign local utility functions to the agents to guide their actions toward a desired system objective. The performance of the assigned local utilities is measured by the well known pure price of anarchy (pPoA) metric that equals the ratio of the system objective at the worst pure Nash equilibrium of the corresponding game to the optimal system objective. Our aim is to derive the utility functions which optimize the pPoA-based performance guarantees for any given information network and system objective. We develop a linear program that derives the optimal pPoA for any arbitrary information network and arbitrary system objective. Our work is the first to solve optimal utility design for arbitrary networks; our techniques generalize previous approaches which considered only the full-information setting. For supermodular objective functions, we prove that counterintuitively, a fully communication-denied utility design is optimal irrespective of the original information network. For submodular system objectives, an exhaustive numerical analysis suggests that the optimal utility design is robust to communication failures even for this case. When the system objective is weighted maximum coverage, the marginal contribution utility design provably optimizes the pPoA for a wide variety of information networks of interest.

Vision-Language-Action (VLA) models show high robotic manipulation capabilities via a two-stage design: a Vision-Language Model (VLM) stage followed by an Action Diffusion Transformer (ADiT) stage. Since robots must meet strict Service-Level Objectives (SLOs) for safety, VLA inference is inherently latency-critical. Meeting these SLOs requires high-end GPUs, yet weight, cost, and power constraints preclude integrating such GPUs on-robot. Prior works offload VLA inference to edge servers that serve many robots on VLA models. However, current VLA systems lack support for multi-request, multi-model execution on a multi-GPU server under SLOs, while existing serving systems for multi-stage models are optimized for throughput and stage disaggregation across separate GPUs, which are ill-suited for the millisecond-scale stages of VLA models. We design Robion, the first VLA serving and management system for multi-robot, multi-model requests on multi-GPU edge servers that meets SLOs. Our serving engine disaggregates the VLM and ADiT stages within a GPU via two streams, dynamically restricting the SMs on VLM stream so ADiT always finds SMs to run alongside it, and co-locates multiple models by sharing these streams across them, prioritizing requests by least remaining SLO time. Our management engine enables flexible model placements on multi-GPU servers, and integrates an intelligent traffic controller that maximizes per-model batching under the chosen placement while bounding each GPU's load to meet SLOs. For individual models, Robion serves on average 6.7$\times$ and 1.5$\times$ higher robot load within 98% SLO attainment over vLLM-Omni, the most widely used multi-stage serving system, and Monolithic, which runs VLM and ADiT as a single pipeline, respectively. In a large-scale experiment of serving 8 different models on a 4-GPU server, Robion can serve up to 64 robots within 98% SLO attainment.

Reliable relative localization between aerial and ground robots is a key requirement for tightly coordinated heterogeneous teams. This can be difficult to do using conventional frame-based cameras and fiducial markers because they are sensitive to motion blur, lighting variations, and payload constraints. This paper presents EVPeriscope, an event-based perception system that enables detection, localization and control of a quadrotor using an upward-facing event camera on a ground robot by detecting the high-frequency visual signature of its propellers. This system allows the quadrotor to function as an extended perception system for the ground robot when onboard sensors exhibit degradation or occlusion. We demonstrate the capabilities of this marsupial ground-aerial system via experiments in challenging field conditions with wind speeds of up to 15 mph, in both daylight and at night. We show that the system supports localization and closed-loop navigation through dense foliage where the ground robot's sensors are occluded. Our control system for the quadrotor operates at 200 Hz entirely with onboard sensing and computation. More details and experiment videos can be found on the project page: https://ongdexter.github.io/evperiscope.

Spatial reasoning depends not only on metric properties such as distance, angle, and shape, but also on topological relations that remain invariant under continuous deformation. Cognitive science identifies these relations as foundational to spatial understanding, yet foundation-model evaluations largely focus on metric or viewpoint-dependent relations. We introduce MindTopo, a benchmark of topological intuition across five properties grounded in cognitive science and formal topology: continuity, separation, order, enclosure, and knots. MindTopo evaluates each property at two cognitive levels. Reasoning asks a model to identify topological relations or infer how they change. Planning instantiates a foundation model as a closed-loop agent whose policy selects environment actions. MindTopo contains 11,030 instances across 13 procedurally generated task types with controllable difficulty. We benchmark 14 MLLMs and study agent configurations augmented with image and video generation, including 3 video generative models in planning settings. Every MLLM performs better on reasoning than on planning, and the best-performing model remains far below observed human performance. On Qwen3-VL-2B-Instruct, supervised fine-tuning and reinforcement learning improve reasoning more than planning. Generated observations retain local cues and reach plausible endpoints, but audited rollouts do not reliably follow environment dynamics or preserve topology across transitions. Our website is at https://mind-topo.github.io/

In this technical report, we propose Pelican-Sim 1.0, a general world model simulator for embodied intelligence that predicts future observations from visual context and robot actions to support downstream learning and decision making. The model incorporates four key design features: (1) Unified action representation: a 28-dimensional action value space covering most mainstream embodiments, keeping one model valid across heterogeneous devices. (2) Action-visual injection: URDF- and camera-rendered action videos bridge actions and pixels, giving markedly better controllability across embodiments, scenes, and tasks (PSNR +0.904 over alternative fusion baselines). (3) Sparse mixture-of-experts (MoE): sparse MoE layers add capacity for heterogeneous dynamics and absorb the action modality while reducing inter-modality conflict (FVD -6.530 vs. the dense backbone). (4) Efficient rollout generation: causal adaptation and few-step distillation yield a four-step autoregressive simulator, achieving a 5.67-fold speedup over the 35-step model. Benefiting from these designs, we train on approximately one million real-world and simulated trajectories and obtain large gains in action controllability and video quality: PSNR improves over the strongest evaluated baselines by 4.636 on AgiBotWorld Beta, 2.080 on RoboMIND, and 10.343 on RoboTwin, with the adapted EWMBench DYN score up 0.426 on RoboTwin. Relying on this, four downstream applications on RoboTwin succeed: 500 generated trajectories added to 50 demonstrations per task raise policy success from 70% to 93%; policy evaluation reaches a Pearson correlation of 0.994 across five checkpoints; and relative success gains reach 47.7% for action selection and 20.3% for policy improvement. Qualitative generalization across trajectory, scene, object, embodiment, and viewpoint shifts highlights its potential as a general-purpose world model simulator.

As Speech Large Language Models (speech-LLMs) become increasingly multilingual, Arabic remains significantly underrepresented, highlighting the need for dedicated infrastructure to train and evaluate Arabic speech-LLMs. To address this gap, we introduce Nuha-Speech, a comprehensive initiative to develop general-purpose Arabic speech-LLMs spanning dataset construction, model training, and systematic evaluation. Specifically, we constructed a large-scale Arabic Speech Question-Answering (SQA) corpus comprising over 1.5 million training samples to allow instruction tuning over a broad range of core speech tasks. Then, the corpus was used for supervised fine-tuning based on Qwen-Omni model variants at different scales. Finally, we designed an evaluation framework featuring diverse tasks and tailored metrics. Through this work, we aim to establish foundational infrastructures for Arabic Speech-LLMs under constraints imposed by limited Arabic speech resources.

Large language models generate fluent text that can contain unfaithful claims -- a phenomenon known as hallucination. We present a multi-signal detection pipeline combining fine-tuned DeBERTa-v3 classification, Monte Carlo (MC) Dropout uncertainty quantification, and temperature-scaled calibration for response-level hallucination detection. Evaluated on the HaluEval benchmark, our pipeline achieves F1=0.915 and AUROC=0.977 on general-domain tasks, with per-task F1 scores of 0.97 (QA), 0.96 (Summarization), and 0.82 (Dialogue). MC Dropout inference further improves accuracy to 93.2%. A context ablation study confirms the model performs genuine entailment reasoning rather than exploiting surface patterns, with summarization F1 dropping 24% when knowledge context is removed. Learning curve analysis reveals that 25% of training data captures 77% of full-data performance. Beyond detection, we apply Direct Preference Optimization (DPO) to a Qwen2.5-0.5B generator, reducing its hallucination rate from 85.5% to 37.7% (55.9% relative reduction) as measured by our detector. Cross-domain evaluation on the SciFact biomedical benchmark shows that general-domain training transfers poorly (F1=0.52), motivating domain-specific fine-tuning. PubMedBERT fine-tuned on SciFact achieves F1=0.63 and AUROC=0.81, demonstrating that domain-matched pre-training is the strongest adaptation strategy. Code and models are available at https://github.com/varunteja99/hallucination-detection-nlp

Many biological discovery problems require experiments to be selected sequentially under constrained budgets. CRISPR screening is a prominent example, as exhaustive perturbation testing is often infeasible and candidate perturbations must instead be prioritized over multiple experimental rounds. Despite the importance of this problem, existing benchmarks for adaptive hit discovery remain limited in scale and diversity. Here, we introduce AssayBench-Loop, a large-scale benchmark for adaptive hit discovery comprising 1,389 CRISPR screens across five phenotype categories. Beyond enabling systematic evaluation, its scale makes it possible to learn acquisition strategies across historical experiments. Building on this resource, we introduce AssayLoop, a sequential experimental design framework combining AssayFormer, a transformer-based amortized acquisition policy trained across historical screens to adapt from experimental feedback, with LLM-derived biological priors through an adaptive handoff. In this view, completed experiments become training data for learning how accumulated evidence should guide what to test next, while LLMs provide prior biological knowledge to seed the search. We further introduce AssayLLM, showing that the same principle can be extended directly to an LLM through task-specific post-training. On temporally held-out screens, AssayLoop achieves a 5.67-fold enrichment over random selection and recovers 27.7% of hits after assaying approximately 5% of the candidate library, outperforming existing adaptive-design methods and standalone LLMs, and AssayFormer alone. Performance improves with increasing historical training data and transfers to phenotype categories excluded from training. These results demonstrate the value of learning acquisition policies across historical experiments and combining them with broad biological priors for efficient adaptive hit discovery.

Recent advances in Vision-Language-Action (VLA) models have improved robotic manipulation, yet observation-to-action learning remains limited by a fundamental transition realizability gap, manifested in three tightly coupled problems: (i) Transition ambiguity. Visually similar current observations may correspond to different manipulation phases and imply different subsequent transitions. (ii) Prediction--execution mismatch. A visually plausible predicted future observation does not necessarily correspond to a physically realizable transition. (iii) Experience--realization mismatch. A historically executable action pattern may not necessarily realize the intended transition in the current scene and therefore requires context-aware adaptation. Accordingly, we propose UniMPA, a Unified Memory-Prediction-Action model that addresses these problems through a shared action-grounded transition interface. (i) UniMPA introduces Persistent-Selective Future Prediction to resolve transition ambiguity by modeling the intended future state evolution. A persistent latent stream continuously tracks task-level progress, while a transition-critical pixel stream selectively resolves fine-grained interaction changes through memory-grounded prediction. (ii) To assess the physical executability of the anticipated transition, the predicted transition queries a temporal Visual-Action Memory Bank. The bank retrieves historically realized visual-action experience, grounding future prediction in executable evidence. (iii) To adapt executable experience to the current scene, an Action-Visual Memory Bank retrieves visually grounded action prototypes from historical action evolution. Prototype-Biased Flow then shifts the flow source toward a historically supported action manifold for context-aware refinement.

Recursive self-improvement (RSI) enables AI systems to turn experience and feedback into persistent changes that improve both their capabilities and the process of future improvement. We first use the Headroom-Closed Index (HCI) to reveal the problems of existing LLMs, then introduce the RSI concept and its development roadmap: from improvement-execution autonomy, improvement-strategy autonomy, experience-acquisition autonomy, and environment-adaptation autonomy, to recursive meta-improvement. Next we examine RSI across scenarios (e.g., scientific discovery, embodied intelligence, software engineering), highlighting their distinct requirements and development speeds. Drawing on diverse industry practices and preliminary empirical evidence, we connect RSI research with practical systems and identify key challenges to achieving genuine RSI.

Agent-based model predictive control (AMPC) has recently been proposed as a distributed scheme that collaborates with all agents to achieve optimal holistic performance. However, its optimality highly depends on the prediction accuracy that requires all agents or their contributions to be known, which is too idealistic for actual implementation. This research proposes a novel practical hybrid control scheme - learning agent-based MPC (LAMPC), combining the model-based AMPC approach and data-based learning methods to improve the holistic vehicle performance for multi-agent systems. The Gaussian process regression (GPR) enhanced by an online data management strategy serves as the learning core to predict unknown contributions. A novel multi-step prediction mechanism leverages the GPR learning potential along the horizon. The predicted mean, representing the learned unknown contributions, completes the system model in the MPC for more accurate control. Meanwhile, a stochastic framework is formulated to guarantee control safety and feasibility using soft chance constraints based on the prediction variance. Both simulations and experiments show that, with the learning capability, LAMPC outperforms the traditional AMPC. LAMPC can achieve higher tracking performance in well-learned scenarios and always guarantee constraint satisfaction even in less-learned scenarios. Moreover, the proposed hybrid control scheme is efficient for real-time implementation and is flexible to any control agent topology.

A language model normally begins training with random word embeddings: whatever 'banana' means must be learned from training corpora. I implement St. Augustine's picture of word learning, meaning by ostension, for a small masked language model (DeBERTa) trained on 10M words: before training, visually grounded tokens receive embeddings derived from the image regions they label; other tokens start random. Visual initialization leaves a measurable imprint that lasts until the end of training. At the same time, the effect remains invisible under most BabyLM benchmarks, which probe abstract grammatical knowledge: visual initialization does not affect performance there. The only zero-shot exception is object-property knowledge (COMPS, Misra et al. 2023), where seeding helps in every configuration. To follow up on this result, I build a corpus-tailored version of the Visual-Property Swap benchmark (Lin et al., 2026), which tests color, material, size, and shape knowledge, with per-item training frequency and seeded status. Here, vision-seeded models have a persistent, seed- replicated advantage, confined to the seeded words. As a causal test, I show that synthetic grounding of previously unseeded words transfers the advantage to exactly those words. Function words and abstract vocabulary also receive strong visual seeds and retain them throughout training, and the training objective draws on them: held-out mask-prediction loss falls for these words in every seed. However, no benchmark I run registers this. What evaluation would pick this up remains an open question.

Speech large language models (SpeechLLMs) offer reduced latency and retain paralinguistic nuances that are typically lost in cascaded automatic speech recognition (ASR) and text-based LM architectures. However, they continue to lag behind text-only LLMs on complex reasoning tasks, while real-time spoken interaction imposes strict latency constraints. Although prior works employ Chain-of-Thought (CoT) and concurrent reasoning to enhance reasoning capabilities without inducing prohibitive delays, an inherent accuracy-latency trade-off persists. In this paper, we investigate whether a streaming SpeechLLM can dynamically revise its reasoning traces on the fly. We introduce RetroThinker, a multi-stage post-training framework that equips the Moshi model to self-verify and forward-correct CoT steps during inference. RetroThinker combines supervised fine-tuning (SFT) on curated retrospective thinking data with length-based direct preference optimization (DPO) to optimize retrospective during early reasoning (i.e., reasoning concurrently while the user speaks). Evaluated on the GSM8K benchmark, RetroThinker significantly improves the accuracy-latency trade-off over non-retrospective baselines, achieving an 11% absolute accuracy gain at a comparable latency.

We consider a game theoretic approach to solve multi-agent coordination problems with submodular maximization objectives. It is known for such problems that the Nash equilibria for the corresponding game are always within 50% of the optimal, but that the equilibria which achieve this worst-case bound are not stable. To exploit this instability, we propose a family of algorithms which we call Truncated Noisy Best-Response (TNBR) Algorithms. These algorithms are flexibly characterized by agents asynchronously and stochastically selecting actions from a neighbourhood of their best response payoffs. We compute bounds on the recurrent classes of TNBR algorithms' associated Markov chains. Our bounds fall into two categories: first, "Performance" bounds ensure that TNBR algorithms always have a high-value recurrent state; second, "Safety" bounds ensure that TNBR algorithms never have arbitrarily-bad recurrent states. Furthermore, these two types of bounds are linked by a waterbed-like effect: every game with a poor Safety guarantee necessarily has a favorable Performance guarantee.

Language identification in code-mixed text, largely observed in social media, is highly essential when users frequently switch between multiple languages within a single utterance. Accurately identifying the languages of code-mixed tokens becomes an urgent necessity. Traditional language identification models, designed for monolingual text, are not well suited for token-level language identification in code-mixed settings. We formulate the task as a sequence labeling problem and fine-tune contextual transformer-based models MuRIL and XLM-RoBERTa best suited for Indian languages. We evaluate these systems on three different data configurations (Hindi, Gujarati, and Bengali) to predict language labels for individual tokens. We release a benchmark for language identification in code-mixed tokens with manually annotated test sets. We propose two approaches of code-mixed generation using parallel sentences of three languages. The trained models demonstrate the effectiveness of contextual embeddings for token-level language identification in multilingual social media text. For reproducibility and to facilitate future research, we publicly release our fine-tuned models.

Cardiovascular screening models trained on national health surveys routinely report areas under the receiver operating characteristic curve (AUROC) near 0.89. We asked whether that accuracy reflects learning or target leakage, whether tabular foundation models change the answer, and whether the properties deployment requires survive joint examination. We benchmarked ten classifiers spanning linear, tree-ensemble, neural, glass-box, and tabular foundation classes for prevalent myocardial infarction in 442,067 respondents of the 2022 Behavioral Risk Factor Surveillance System across five feature tiers of decreasing leakage risk. Each was audited for discrimination, calibration, fairness at an explicit screening threshold, conformal coverage, explanation faithfulness, and inference cost, then applied -- models and thresholds frozen -- to 430,755 respondents of 2023. Removing two post-diagnostic features cost every model 0.049-0.051 AUROC, collapsing the field into a 0.0045-wide band. The glass-box explainable boosting machine was non-inferior to every alternative within a pre-specified 0.005 margin while scoring the cohort roughly 104 times faster than the strongest foundation model. One threshold detected 75.4% of women's infarctions against 89.0% of men's; editing the model's shape functions reduced the gap to 0.010. Marginal conformal prediction gave 0.86 coverage to men and 0.82 to adults over 60; Mondrian calibration repaired every stratum. Frozen models transported within 0.002 AUROC. Reported headroom in this literature is a property of the feature set, not the learner. Transparency cost nothing measurable and made fairness repair and uncertainty conditioning directly auditable. Evaluation practice, not model capacity, is the binding constraint.

This paper investigates optimal control for nonlinear systems with unknown dynamics, input constraints, disturbances, and adversarial signals. The objective is to develop a learning-based control method that allows the designer to prescribe the desired convergence time in advance. An integral reinforcement-learning framework is proposed to avoid requiring exact knowledge of the system dynamics while ensuring that the control input always satisfies the actuator constraints. Current and recorded data are combined to train the critic without requiring persistent excitation. The learning gain is selected directly from the prescribed convergence deadline. Lyapunov analysis is then used to establish practical predefined-time convergence of the coupled state-critic system in the presence of disturbances and adversarial channels. The effectiveness of the proposed method is validated through the stabilization control of a two-link robot manipulator.

Automatic speech recognition (ASR) systems and audio language models (audio LMs) now report low error rates on monolingual benchmarks, but their behavior on code switched speech in low resource, diacritic rich languages remains poorly characterized. We present a switch aware evaluation of eleven modern systems (six ASR models and five audio LMs) on English Yoruba code-switched speech, using a deterministic 2000 utterance evaluation set and a shared scoring pipeline. Beyond word error rate (WER), we report switch localized diagnostics: a switch entry token error rate (SETER), windowed switch point error rates, language specific error rates, and a diacritic insensitive WER. Our central finding is that aggregate WER hides code switching behavior. The best system by WER (an ASR model) is statistically indistinguishable from a leading audio LM on WER, yet the audio LM is significantly better on every switch localized metric. Across faithful systems, Yoruba token recognition collapses (error 0.97 for almost all systems) while English tokens are recognized far better, and errors concentrate sharply at switches into Yoruba. Several generative audio LMs fail as exact transcribers, producing translation, verbosity, and prompt leakage that are strongly prompt dependent. We release manifests, metric implementations, and evaluation scripts to support reproducible, switch aware benchmarking for African code switched speech.

This paper addresses predefined-time resilient consensus of leaderless second-order nonlinear multi-agent systems under denial-of-service (DoS) attacks, motivated by coordination requirements in safety-critical applications. The agents are subject to bounded external disturbances and communicate over a strongly connected directed graph whose links are simultaneously disabled during attacks. We develop a switching sliding-mode protocol with the objective of reaching an invariant manifold of position and velocity agreement. The protocol uses relative position and velocity information during attack-free intervals and local velocity feedback during communication blackouts. A time-scaling function remains constant during each blackout and resumes evolving when communication is restored, accounting for the time available for consensus. Under bounds on attack duration and frequency, we derive sufficient gain conditions through a Lyapunov analysis. We show that, despite bounded disturbances, the agents achieve position and velocity consensus by a realistic settling time equal to a prescribed convergence duration plus the cumulative attack duration up to the realistic settling time. The prescribed convergence duration is independent of the initial conditions, and the realistic settling time reduces to that duration in the absence of attacks.

Dexterous in-hand manipulation of a grasped object with an anthropomorphic hand is an unsolved frontier for robot dexterity. The contact-richness and highly dynamic nature of object-hand interactions tend to require extensive modeling or data-collection efforts for learning-based approaches. Modern simulators used for reinforcement learning (RL) cannot fully replicate the required contact complexity, while collecting dexterous demonstrations for imitation learning (IL) remains an open problem. In this research, we present an embodied control approach based on real-time task Jacobian estimation of the combined hand and object system on the physical robot. Using only the CPU on a laptop, the proposed controller begins in-hand pen writing after approximately 18 s of initialization and continues to adapt online, without an analytic hand--object kinematic/contact model, simulation training, or precollected task demonstrations. We demonstrate that the same estimator/controller formulation works on three anthropomorphic robotic hand systems (one physical, two simulated) to show human-like, in-hand articulation of a grasped pen by an embodiment-independent formulation. Sub-millimeter in-plane precision (mean 0.6 mm across runs) is achieved across letters and shapes written in the air and on paper on a physical robot. To our knowledge, this is the first demonstration of an anthropomorphic hand writing arbitrary single-stroke trajectories with a grasped pen through purely in-hand motion, and it showcases an alternative to compute- and data-heavy approaches such as RL and IL for achieving dexterous manipulation through computationally simple and data-efficient algorithms.

Large language models are superseded every few quarters; clinical evidence takes years. We asked whether medical research is keeping pace with the systems it evaluates. PubMed returned 11,628 records for January 2023 to June 2026 across fourteen clinical domains, growing 45-fold; 2.5% used a randomised, controlled or prospective design. Evaluation lag, from a study's newest named model release to its own publication, widened from 1.33 to 6.08 quarters. Because discontinued models age mechanically, we benchmarked this against a counterfactual holding model composition fixed: migration to newer systems offset only 56% of the drift (95% CI 50-65). Randomised trials evaluated models a median 4.6 quarters older than other designs (P = 3 x 10^-19), yet among studies naming a model still under development no design differed from any other; 62% of randomised trials evaluated a discontinued family. Rigour and currency are in tension, and that tension reflects model selection rather than research timelines.

Allowing large language models (LLMs) to retrieve information from a set of trusted documents can increase reliability and reduce hallucination. However, recent work has demonstrated that retrieval-augmented generation (RAG) can have unintended side effects on the overall safety of the generated responses, when prompted for harmful or dangerous content. A clearer understanding of the mechanisms leading to this result is needed, as increasing numbers of end users turn to RAG to incorporate corporate documents and knowledge bases into LLM-based systems. We introduce RAG-Safety-Bench, a benchmark to measure the safety impact of RAG on LLM models. By removing the confounding effect of retriever quality, and cleanly separating the problem into four conditions -- non-RAG, RAG with an oracle document containing the answer to the harmful request, RAG with documents related to the harmful request but without the specific answer, and RAG with random, safe documents -- the benchmark isolates the impacts of different factors in the observed safety degradation. We report results across five open-source LLMs, showing an inverse relationship between benign and unsafe capability, strong evidence that baseline safety guardrails do not lead to downstream safety guarantees in the RAG case, and model-specific support for previous findings that even benign documents can lead to unsafe generation in retrieval-enabled systems.

Imitation learning for dexterous hands is bottlenecked by the difficulty of collecting contact-rich demonstrations that transfer faithfully to the robot. Prior wearable-exoskeleton systems record only on the human side and retarget via open-loop mappings calibrated in free space, which degrade under contact. We present SEED-UMI, a framework in which both the human and the robot wear the same exoskeleton: joint encoders become a physically shared measurement, and wrist cameras mounted to the exoskeleton observe the same outer mechanism during both human data collection and robot policy rollouts. This turns retargeting into paired cross-embodiment supervision and lets policies train directly on raw exoskeleton-centric wrist images, without segmentation or inpainting. On five contact-rich tasks, SEED-UMI achieves 3.0x greater data collection efficiency than exoskeleton-based teleoperation and a 70.0% average rollout success rate.

Large language model serving costs scale directly with output sequence length, yet standard preference alignment often inflates response verbosity without improving utility. We study whether the parameterization of post-training updates affects generation length: low-rank subspaces alter sequence length without modifying the alignment loss. We present LOCUS, a method that selects a task-aware low-rank adaptation subspace to minimize output-token cost subject to a utility constraint. Within this subspace, post-training retains the native preference objective with a frozen backbone. Across Anthropic HH-RLHF dialogue preferences, we evaluate two $\sim$3B decoder backbones, Pythia-2.8B and Qwen2.5-3B, against protocol-matched full-parameter DPO and DrDPO branches and the released SamPO checkpoint. LOCUS reduces continuation length by up to 39.84\% on Pythia-2.8B and by 14.87--17.58\% on Qwen2.5-3B while updating only 0.24--0.28\% of model parameters, with no material change in the internal preference diagnostic.

Collective intelligence depends not only on the capabilities of individual members, but also on how those members are organized. Yet artificial multi-agent systems are typically assembled using fixed organizational structures, even when the physical tasks they perform impose fundamentally different coordination requirements. Here we show that principles from human organization theory can be operationalized to organize large, heterogeneous collectives of embodied artificial agents. We introduce ORCH (Organizing Roles and Coordination Hierarchies), which constructs task-specific hierarchical organizations by combining pooled interdependence for work that can proceed concurrently with sequential interdependence for work governed by prerequisite relationships. Across 25 wildfire-response missions spanning reconnaissance, rescue, transportation, resource management, containment and suppression, we evaluated teams of up to 50 heterogeneous agents using eight large language models. Organizations constructed using these principles consistently outperformed four representative embodied multi-agent approaches across mission outcome, execution efficiency, exploration and computational resource use. Human-designed ORCH organizations improved final score by 63.97% and execution efficiency by 74.29% on average relative to the four prior frameworks. Organizations generated automatically by language models improved these measures by 43.63% and 52.53%, respectively. These advantages persisted across missions and underlying language models. Notably, collective performance was not monotonically determined by model scale. Analysis of long-horizon missions showed that hierarchical organization enabled teams to preserve concurrent activity within specialized groups while coordinating ordered transitions between mission phases.

This paper details the Eloquence team's approach to Task 2 of the 2nd MLC-SLM challenge at Interspeech 2026, which involves multilingual Multiple-Choice Question Answering (MCQA) across 21 languages. Three approaches are explored. First, we fine-tune Voxtral-Mini-3B via LoRA with cross-lingual data augmentation, ASR transcript augmentation and timestamp-aware audio cropping, achieving 0.72 macro-accuracy on evaluation Phase 2. Second, we apply multimodal in-context learning (ICL) to the frozen Voxtral-24B model to correct a strong label bias, reaching 0.81, our best result. Third, a training-free retrieval system based on a three-layer voice-anchored memory combining acoustic identity, semantic content, and a knowledge graph achieves 0.68. All three systems substantially outperform the official baseline.

Post-training quantization compresses large language models (LLMs) by storing their weights at reduced precision, and each quantized weight introduces an error into the hidden states. Naively, these errors should accumulate with depth and corrupt next-token prediction; randomly initialized models accumulate these discrepancies rapidly, whereas quantized pretrained models accumulate much less hidden-state error and largely maintain downstream task performance, even though they were never trained with quantization noise. This raises the question we address: why does post-training quantization work? Comparing full-precision and quantized forward passes, we identify two mechanisms that characterize pretrained quantization robustness. First, the error a layer newly introduces tends to oppose the error it inherits from the layer's input. The two cancel partially such that the discrepancy between full-precision and quantized passes grows slowly. This counteracting residual interaction develops during pretraining. Our quantitative analysis identifies it as a major factor slowing hidden-error growth. Second, LM-head geometry preferentially preserves the scores and probabilities of high-ranked tokens, which typically represent the model's most confident predictions. Together, these mechanisms explain why quantization error that passes through numerous layers can still produce only small output changes, and we verify the findings across models and quantization settings.

While existing strategies for Flight and Maintenance Planning for the defense sector generally address idealized conditions, real-world planning often involve non-nominal initial fleet states and complex inspection schemes. To address these challenges, we propose a multi-year planning strategy that maximizes long-term fleet availability. The formulation incorporates multiple competing objectives capturing the tight coupling between aircraft usage and maintenance, while enforcing cyclic inspection requirements and limited maintenance-dock capacity. To enhance practical applicability, we introduce three complementary strategies: a receding-horizon approach to reduce computational burden, a Bayesian Optimization routine to automatically tune cost-function coefficients, and an aircraft-specific weighting scheme that adapts model priorities based on fleet-condition statistics. Results on three realistic benchmark scenarios demonstrate the effectiveness of the approach.

On-Policy Self-Distillation (OPSD) has emerged as a popular paradigm for large language model (LLM) self-improvement, allowing models to act as their own teachers by leveraging privileged information such as ground-truth solutions. However, recent findings indicate that OPSD can severely degrade the performance of LLMs on complex reasoning tasks: By forcing the student to imitate an artificially confident reasoning trace conditioned on privileged information, OPSD inadvertently suppresses expressions of uncertainty and penalizes the exploratory, self-corrective behaviors required to solve challenging problems. To address this, we introduce Negative Self-Distillation (NSD), a new framework that optimizes LLMs by diverging from flawed reasoning rather than imitating privileged solutions. Instead of relying on ground-truth answers or external supervision, NSD uses the model itself to generate a question-specific negative condition (eg, acting as a ``careless reasoner'') and pushes the student's distribution away from this self-generated negative teacher. Naively applying unlearning objectives to achieve this divergence is problematic, as flawed reasoning tokens are confounded with basic linguistic tokens; indiscriminately penalizing both risks catastrophically degrading the model's foundational language capabilities. We resolve this by designing a dynamic gating mechanism that automatically identifies and isolates reasoning-critical tokens, ensuring gradient updates target only behavioral flaws while preserving the model's linguistic priors. Empirically, NSD consistently outperforms OPSD and other label-free, self-bootstrapping reinforcement learning (RL) baselines.

Vision-Language-Action (VLA) and World-Action Models (WAMs) have demonstrated strong capabilities in general-purpose robotic manipulation, yet their generated actions may violate hard physical constraints and therefore be unsafe or infeasible for deployment. Existing safety approaches either optimize statistical safety objectives without deterministic per-step guarantees or correct unsafe actions only during inference, creating a mismatch between policy training and execution. We introduce ActSafeGuard, a differentiable and training-aligned safeguard layer for flow-matching based policies. ActSafeGuard integrates hard action feasibility into policy learning, not merely treating safety as an inference-time external component. Through an analytical ray-scaling operator design, ActSafeGuard enables boundary-aware gradients to guide the model to naturally learn constrained manifolds. Extensive experiments on multiple standard foundation backbones ($π_{0.5}$ and Fast-WAM) across various tasks demonstrate that ActSafeGuard consistently achieves a $100\%$ step safety rate while fully preserving or even boosting task success rates, providing a scalable and minimally invasive solution for safe embodied AI deployment.

We present a robust contact-aware control framework for aerial writing on an underactuated platform. The framework combines nonlinear model predictive control (NMPC) for accurate end-effector position and normal-force tracking at small reference penetration depths, with consistent performance across controller tunings, with whole-body incremental nonlinear dynamic inversion (INDI) for robustness to frictional and aerodynamic disturbances during contact. The proposed controllers are validated on a quadrotor-based aerial manipulator with a rigid, single-link, one-degree-of-freedom (DoF) arm in simulation and real-world experiments. The aerial writing experiments span vertical and inclined surfaces, multiple reference forces, different friction conditions, and wind disturbances. The results demonstrate that robust simultaneous five-DoF end-effector pose and contact-force tracking is achievable on a standard underactuated quadrotor with a simple, rigid, single-link arm, without requiring a fully actuated platform, a complex arm, or dedicated force/torque sensing.

Underactuated robotic hands achieve adaptive and robust grasping with a reduced number of actuators, but predicting the stable equilibrium pose of the grasped object remains a significant challenge. This paper introduces a quasi-static analytical approach to assess passive stability in underactuated multi-finger hands. A novel three-finger hand architecture integrating a differential spring-loaded slider mechanism is introduced, enabling versatile and adaptive grasping. The study focuses on how the differential mechanism influences the overall grasp behavior and analyzes the effect of object size on the stable equilibrium configurations for two canonical grasp types: cylindrical and spherical.

This paper develops a critic-free policy iteration (PI) method for continuous-time linear zero-sum games. The central idea is to characterize the saddle-point policies directly in the joint policy space, rather than treating the quadratic value matrix as an iterative variable. A policy game Riccati equation (PGRE) is introduced whose unknowns are policy gains only. Its solutions are shown to be in one-to-one correspondence with the symmetric solutions of the game algebraic Riccati equation. Based on a stabilizing anchor, PI is performed directly in the actor space. The actor-space Jacobian is nonsingular at every stabilizing policy, and the resulting policy sequence coincides with that of simultaneous PI. For unknown dynamics, a data-driven algorithm uses a single batch of data and nullspace projection of endpoint increments to eliminate the value matrix, yielding an actor-only regression. A necessary and sufficient rank condition for unique policy recovery is established and shown to be iteration-invariant, demonstrating that critic identifiability is unnecessary. A power systems frequency-regulation example verifies convergence and policy recovery, while scalability tests demonstrate substantial reductions in computational and memory requirements.

Mainstream robotic policies often adopt a Markovian formulation, but many complex real-world manipulation tasks are inherently non-Markovian, requiring long-horizon memory beyond the current observation. Existing memory mechanisms often rely on language summaries, growing visual windows, or their combinations, and may therefore lose fine-grained visual evidence or face a trade-off between history coverage and execution efficiency. We introduce MaP-WAM, a Memory-as-Plans framework that decomposes memory-dependent world-action modeling into memory-grounded planning and plan-conditioned execution, and uses long-term multimodal episodic context as planning-time evidence rather than repeatedly conditioning the executor on the full history. MaP-WAM represents memory as completed segment records containing language instructions and sparse visual context, and converts this episodic memory into compact plans comprising the next segment-level language plan and corresponding visual guidance. A World-Action-Progress (WAP) model executes each plan over an unknown duration by jointly predicting action chunks and corresponding execution progress at inference time, calibrating predicted progress through plan-observation alignment for adaptive segment transitions and closed-loop context updates. MaP-WAM keeps the executor context length fixed, while structured attention further enables key-value caching in both planning and execution. MaP-WAM achieves state-of-the-art performance on RMBench with an 83.3% success rate and attains 78.0% success on real-robot tasks, while maintaining approximately constant executor inference latency as task history grows.

Humanoid locomotion across complex terrain demands forward-looking exteroception to anticipate obstacles, yet this signal is unreliable in real-world deployment, failing partially and intermittently. Existing perceptive policies often assume that depth observations remain clean and in-distribution, while recent attempts to unify perceptive and blind control typically route or switch between separate sub-policies, leaving recoverable information in partially corrupted depth unexploited. We instead propose CAP, a single-stage humanoid locomotion policy that recovers this signal with a perceptive world-model encoder trained as a learned denoiser to reconstruct clean depth from a corrupted input, together with a co-active proprioceptive variational encoder that supplies depth-free body-state information. A coupled training recipe pairs a depth-noise curriculum on the world-model input with world-model feature dropout on the policy-facing latent, exposing the policy to failures across the entire perception-quality spectrum. In simulation, CAP matches or improves upon perceptive baselines when depth remains informative, and degrades more smoothly than a binary-switching baseline as perception worsens. On the Unitree G1, controlled trials and indoor-outdoor deployments demonstrate perception-robust locomotion under intermittent occlusion, real-sensor corruption, and outdoor depth artifacts.

Efficient language learning requires methods to reduce the reliance on large data and computational resources. We investigate structural transfer: First training models on non-language data to induce useful priors for natural language. This approach is a form of weight initialization for multilingual language modeling. We evaluate transfer via next-token-prediction loss, weight shifts in the model, and downstream linguistic benchmarks. Several symbolic data types - notably music, probabilistic grammars, and cellular automata - yield lower language-modeling loss than random initialization. These gains coincide with smaller weight shifts during subsequent language training, suggesting that structural transfer positions models in a more favorable region of the parameter space. However, a lower loss does not translate consistently into better downstream linguistic performance, and transfer from non-language data is less efficient than additional language data. We conclude that non-language data can serve as a partial substitute for language data for the training objective of next-token prediction but does not reliably support broader linguistic generalization.

The development of autonomous driving demands comprehensive testing in mixed-traffic scenarios involving vulnerable road users (VRUs), where purely artificial agents often fail to capture authentic human social negotiations. While human-in-the-loop (HITL) simulators enable safe investigation of these interactions, existing multi-agent platforms struggle with the network latency and synchronization constraints required for high-fidelity haptic feedback. To resolve this, we present CARLAverse, an open-source, multimodal simulation ecosystem. Extending modular hardware abstraction, CARLAverse integrates driving (DrivoCARLA), cycling (CycloCARLA), and pedestrian (WalkoCARLA) simulators into a shared virtual environment. Its core methodological contribution is a distributed physics architecture: latency-critical ego dynamics and high-frequency force feedback are computed locally on client nodes, while a central CARLA server orchestrates non-player character (NPC) physics and global traffic. By decoupling haptic control loops from network bottlenecks, CARLAverse enables scalable, cross-institutional HITL experiments without compromising physical immersion. Code and documentation: https://git.ieem-ka.de/simulator-environments/carlaverse

Maritime Air-to-Sea links are dominated by a line-of-sight ray and a sea-surface reflected ray whose destructive combination produces deep fades. Existing mitigation strategies optimize Unmanned Aerial Vehicle position or trajectory but leave attitude unexploited. This paper treats the full three dimensional attitude as a physical-layer control variable that shapes the two-ray interference through antenna phase-center displacement. Under small-angle and far-field assumptions, the constructive-interference condition reduces to an affine constraint in the Euler angles and admits a closed-form family of minimum-norm attitude candidates. A differentiable soft-minimum rule yields a smooth reference tracked by a constrained Nonlinear Model Predictive Control controller on a fully-actuated tilting multirotor. The proposed scheme increases cumulative throughput by 11.4% over a pitch-only benchmark and 22.2% over a zero-orientation baseline, while preserving trajectory tracking and respecting actuator limits.

Reviewer comments naturally relate to specific parts of the reviewed paper, yet grounding these comments to the underlying evidence is difficult due to long multimodal documents. Existing benchmarks do not capture this setting and largely focus on explicit, information-seeking queries. We introduce ReGround, a large-scale dataset for reviewer comment grounding that links 10,267 reviewer comments to 16,274 evidence in the original anonymous submission of 3,656 papers. We build on a simple observation: author rebuttals often include explicit references to content of the submission used to address reviewer comments, providing a high-precision annotation source. We cast grounding as a retrieval task and evaluate a wide range of retrieval methods. Results show that retrieval over the entire paper content performs poorly, evidence-type inference is a major bottleneck, and multimodal evidence provides complementary signals that text alone misses. Our dataset exposes grounding reviewer comments as a difficult and practically important problem for scientific document understanding.

Reliable robot deployment requires online failure monitoring, yet existing monitors mainly derive risk from proxy signals or train dedicated monitoring components. We ask whether the internal predictive states of a frozen pretrained robotic world model already contain directly decodable failure information. Failure-Aware Readout from World Models (FARM) trains only a 33,985-parameter supervised readout over frozen VLA-JEPA predictive states, producing step-wise failure scores and causal trajectory risk. Five-fold out-of-fold evaluation across seven source tasks reaches 85.68/88.59 pooled AUROC/AUPRC, and FARM gives the best Seen performance among 15 matched baselines on the 10-task benchmark. Across four real-robot populations on PIPER X, SO-101, and Franka, fixed-readout transfer and readout-only adaptation test deployment shifts without updating the predictive backbone. FARM also discriminates failures from partial causal histories and adds 0.2256 ms mean CUDA latency once the frozen state is available. These results support frozen predictive world-model states as reusable features for causal, transferable, and low-overhead execution monitoring.

Skill adaptation frameworks based on reinforcement learning often require restrictive assumptions to maintain stability, such as fixed observations or tightly controlled exploration schedules. In cluttered and dynamic environments, however, unrestricted exploration can lead to unsafe behaviour and unstable learning, particularly when task-relevant observations lie near obstacles or involve moving objects. In this work, we present Dist-GPRL, a distance-aware and safety-guided reinforcement learning framework for structured robot skill adaptation. Building upon Gaussian Process (GP)-based skill parameterisation, our framework sequentially adapts overlapping local windows of sparse trajectory via-points rather than modifying the complete skill at every policy step. Raw policy outputs are correlated through the GP covariance structure, producing temporally coherent trajectory updates while reducing the action-space and credit-assignment difficulties associated with global trajectory adaptation. Safety is incorporated through two complementary forms of guidance. A safe-subspace prior derived from the Hausdorff Approximation Planner (HAP) biases policy exploration toward feasible regions, while dynamically updated distance field clearance and gradient rewards provide local obstacle awareness. A trajectory-kinematics similarity regulariser further preserves the demonstrated velocity and acceleration characteristics during adaptation. We evaluate the framework on two dynamic object-manipulation tasks in simulation and transfer the learned policy to real-world robot execution. Experimental results demonstrate higher task success, lower collision frequency, and more stable learning than the baselines, while preserving the kinematic characteristics of the demonstrated skill.

We study asynchronous replanning in a two population linear quadratic mean field game in which the populations may begin from different beliefs and hence use different plans. Each population observes its own aggregate trajectory and a public record of implemented revisions, while its continuation best response depends on the opponent's current plan. We identify the information required for replanning as the aggregate state at the end of the initial observation interval together with the opponent's active continuation plan. For linear observations, recoverability of this state-plan pair is characterized by a kernel inclusion, and a bounded factorization quantifies sensitivity to observation error. In particular, the required pair may be recoverable even when the full hidden belief is not. Once initialized, the public event record and the common best-response map recursively determine subsequent opponent plans, and the resulting local algorithm reproduces an ideal benchmark on every finite opportunity prefix; implemented revisions alternate as a consequence of best-response persistence. For finite populations, we derive an eventwise linear recursion for sampling errors, obtain finite prefix error bounds, and prove record matching for an autonomous deadband rule under a positive decision margin. Finally, we separate unique solvability of mutual continuation responses from stability of alternating responses, and show that at a pre-terminal Zeno accumulation, spectral stability together with a moving-boundary estimate yields convergence of the continuation plans to the equilibrium restarted from the actual limiting state.

Large language models exhibit complementary strengths, motivating routing methods that dispatch each query to the most suitable model. Although existing routers are effective in single-turn settings, they do not directly transfer to multi-turn dialogue, where routing performance critically depends on how historical context is segmented, retained, and incorporated into the current prompt. This introduces two fundamental challenges: preventing information loss and information confusion during context construction, and evaluating routing quality without conflating model selection with prompt construction quality. In this paper, we propose SWRouter, a Similarity-Contractive Window Router for multi-turn large language model routing. SWRouter combines a similarity-based context segmentation mechanism for prompt construction with a dual-metric evaluation framework that decouples construction accuracy from router performance. Experiments on multi-turn dialogue benchmarks demonstrate that SWRouter consistently surpasses strong baselines, achieving a 16.26% improvement in evaluation accuracy over the best individual large language model and an additional 8.22% gain over the Conv-ID Context baseline. Our results highlight that multi-turn large language model routing requires a joint design of context construction and evaluation, rather than a direct extension of single-turn routing methods.

Large language models (LLMs) are increasingly used for machine translation, yet their outputs often contain additional text beyond the translation itself, such as language labels, explanations or bilingual repetitions, which we term translation noise. Despite its prevalence, this problem lacks dedicated benchmarks and systematic study. We analyze over 790,000 translation outputs from 12 LLMs across 22 language pairs (LPs) and identify 12 recurring noise patterns, which we group into formatting and content noise. Building on the observed patterns, we construct TransClean, a controlled benchmark of 9,900 pairs of noisy and clean translation outputs, comprising 8,800 synthetically generated instances and 1,100 manually curated authentic instances. We evaluate two extraction approaches on the TransClean benchmark: 1) a span-based extraction method leveraging translation quality estimation models for span detection, and 2) an LLM-based extraction method that prompts an LLM to isolate the translation. Our benchmark and analysis provide the first systematic framework to evaluate and improve the cleanliness of LLM translation outputs.

State-of-the-art retrieval-augmented generation (RAG) methods exploit document structures to acquire sufficient evidence, but often incur substantial token costs. To reduce structural-context tokens without compromising high RAG accuracy, we present {\sf VikingRAG}, a directory-aware semantic data management system that tightly integrates semantic and structural access to support structural-context-efficient, evidence-gap-driven multi-round retrieval. To further reduce token overhead of multi-round interaction, we materialize agentic multi-round retrieval traces as experience edges, and reuse these edges for similar queries, avoiding repeated multi-round exploration. To additionally reduce token costs when agentic multi-round retrieval is unnecessary, we introduce an adaptive escalation strategy that answers from one-round experience-augmented retrieval when the evidence is sufficient, and invokes agentic multi-round retrieval only otherwise. Experiments on real datasets show that the base system {\sf VikingRAG} matches high accuracy of state-of-the-art methods while consuming only 11.6\%--51.9\% of their tokens. With retrieval-trace reuse and adaptive escalation, token costs drop to 5.1\%--32.5\% while maintaining competitive accuracy and practical document-storage performance, showing the utility of this work for emerging AI knowledge bases.

Aerial robot swarms have the potential to transform time-critical safety, security, and search-and-rescue operations. By coordinating multiple robots, they can rapidly survey disaster sites, map collapsed or GPS-denied environments, and search cluttered areas faster than a single robot, reducing response times and minimizing risks to first responders. Realizing this potential, however, requires robust autonomous swarm navigation, which remains an active research challenge. Progress is further constrained by existing platforms, as commercial drones are often closed-source or lack the onboard computational resources needed for agile, vision-based collective flight. Moreover, developing, deploying, and maintaining software across multiple aerial robots requires significant engineering effort. To address these challenges, we present SwarmNxt, an open-source software platform built on the open-source OmniNxt drone hardware. SwarmNxt provides an end-to-end toolkit, including detailed hardware assembly instructions with a video tutorial, automation tools for parallel software deployment and swarm-wide updates, and a ROS 2-based framework for autonomous navigation. The platform integrates state-of-the-art control, planning, and depth estimation into a single ROS 2 multi-agent system, providing an open research infrastructure for physical swarm experimentation. We validate SwarmNxt through two real-world experiments: a six-drone swarm performing decentralized planning with high-speed inter-drone collision avoidance, and a four-drone swarm executing collective flight with onboard depth estimation in an obstacle-filled environment. Both experiments were run indoors with global position from external motion capture; perception, planning, and control run onboard.

Human-to-humanoid retargeting has largely been studied on legged platforms, while comparatively few wheeled-humanoid systems support coupled locomotion and manipulation from general human motion. Building on GMR's configurable general-motion retargeting and BeyondMimic's physically simulated R1 Pro learning framework, we present a reproducible pipeline that converts multi-dataset SMPLX motion into executable loco-manipulation behavior for the Galaxea R1 Pro wheeled humanoid. The robot has a planar three-wheel base, a serial torso, and two arms but no leg joints, so human lower-body motion must be redistributed across base motion and torso posture without sacrificing manipulation-relevant arm geometry. Our pipeline combines canonical body-shape preprocessing, planar-base normalization, morphology-aware differential inverse kinematics, shoulder-rooted hierarchical arm retargeting, and continuous torso substitution for bending and squatting. A reference-twist-driven planning layer then decodes planar base motion into continuous three-wheel steering and rolling commands subject to hysteresis, kinematic continuity, acceleration, and actuator-rate limits. Finally, a 21-dimensional BaseDecode policy is trained in Isaac Lab with directional joint-limit scaling, focused upper-body tracking, and a staged wheel-contact reward. The resulting system provides a complete bridge from human motion data to physically trackable wheeled-humanoid loco-manipulation rather than a visualization-only retargeter; quantitative policy comparisons remain scheduled for a later revision.

This paper describes our system for Task~2 of the second Multilingual Conversational Speech Language Model (MLC-SLM) Challenge. We adapt Qwen3-Omni-30B-A3B-Instruct with a segment-evidence-aware data and post-training pipeline. A language model converts timestamped ASR into coherent event spans, which are expanded by a boundary margin and cropped from the original recording. We then synthesize complementary semantic MCQs with Qwen3.6-27B and acoustic MCQs with Gemini~3.1 Flash-Lite, followed by structural, grounding, answer-consistency, and target-model trainability checks, yielding 359,825 verified MCQs across 21 language and accent variants. A text-only probe partitions the data into weak, text-answerable items used for supervised fine-tuning and strong, audio-dependent items used for reinforcement learning with Group Sequence Policy Optimization (GSPO), stabilized by debiased advantages, sequence-level importance correction, and dynamic filtering. Our system obtains 90.92% accuracy on the final official evaluation set.

Closed kinematic chains complicate modular modeling by coupling active and passive coordinates through nonlinear closure constraints. This paper presents a Path-Assembled Closure Differential Mapping (PACDM) framework for modular closure resolution and kinematic reduction. Each closure element compares two ordered transformation paths with common endpoints, with their mismatch expressed through the logarithm on SE(3) and the corresponding Jacobian assembled from local transformation derivatives. Multi-path modules are constructed from a minimal set of pairwise closure elements, while rank-revealing analysis selects locally independent scalar constraints. A defect homotopy recovers closure-consistent passive coordinates from approximate estimates along a feasible and regular continuation path. At regular configurations, implicit differentiation yields the local active-to-passive differential mapping, which is subsequently used in a predictor-corrector continuation procedure for prescribed motion. The framework is evaluated on a seven-degree-of-freedom heavy-duty manipulator containing two-path and three-path closed-chain modules. Comparison with Simscape Multibody yields trajectory root-mean-square errors below 8.5 x 10^-10 rad, while predictor-corrector continuation is approximately 45.8 times faster than applying defect homotopy at every trajectory sample.

As large language models are increasingly deployed in sensitive domains, anonymizing input data to protect personally identifiable information has become a critical practice. However, the impact of this anonymization on model utility is not well understood. This paper presents a systematic empirical study of the trade-off between privacy and performance. We evaluate five prominent language models across eleven diverse benchmarks, comparing their performance on original versus pseudonymized inputs. Our results reveal that while anonymization generally degrades performance, the effect is highly nuanced. We find that more capable models, such as Qwen2.5-72B and GPT-4o mini, suffer the largest performance drops, suggesting a stronger reliance on specific entity information. The impact is also task-dependent: performance on TruthfulQA improves with anonymization, while retrieval-focused tasks like RGB experience a catastrophic decline. Further experiments show that reversible anonymization techniques that preserve entity uniqueness significantly outperform irreversible ones like redaction, and that explicitly prompting models about anonymization offers no discernible benefit. We conclude that anonymization is not a one-size-fits-all solution and must be co-designed with the model and task in mind to balance privacy and utility effectively. Our findings provide a crucial baseline for developing more robust, privacy-aware AI systems.

Natural Language Inference processes pairs of sentences to extract their semantic relations. NLI has been a hot research topic, integrated as a main component in other NLP applications. Despite significant advancements in textual inference across various languages all around the world, Arabic language still suffers from limited resources in this domain. To address this gap, this paper introduces E-CONAN benchmarks that are composed of sentences pairs from various sources: (1) automatically-translated pairs, (2) human-validated machine-translated pairs, (3) hand-crafted pairs from teaching Arabic as foreign language books, and (4) headlines pairs from different news channels containing rumors. E-CONAN contains two benchmark datasets, E-CONAN-2, a 2-way dataset (RTE) and E-CONAN-3, a 3-way dataset (NLI). Additionally, we have used E-CONAN benchmarks to evaluate 9 state-of-the-art multilingual pretrained models using zero-shot classification. Models were evaluated across the ArNLI, XNLI, and E-CONAN datasets. Results show that E-CONAN is a potentially valuable resource for evaluating model generalization and even for fine-tuning pre-trained models. Its diverse composition, derived from a combination of sources, offers a broader and more robust assessment compared to XNLI and ArNLI. In addition, we have evaluated 5 LLMs on E-CONAN-3 dataset. Moreover, we incorporated MARBERT as a representative Arabic-specific baseline and conducted performance evaluation comparison to demonstrate how Arabic-specific models scale against cross-lingual and LLM-based approaches on the E-CONAN benchmarks. Furthermore, we conducted detailed qualitative and quantitative error analysis to analyze frequent error patterns. E-CONAN benchmarks will be publicly available, we hope that it will enrich research community in Arabic textual entailment and natural language inference.

Long-horizon robot manipulation requires memory, but not necessarily inside the action policy. To address such tasks, current agentic systems often combine VLAs with planners and geometric tools, sometimes using additional depth or calibrated geometry. These systems confound attribution: gains may come from richer observations or alternative motor tools, while failures may stem from either the policy or an under-specified language interface. We isolate this question through a deliberately constrained design: less tool breadth, but greater interface bandwidth. 2AM makes a multimodal Agent the sole holder of task memory and a single RGB-based, episodically stateless Action Model the sole executor of task-relevant motion. The Agent compiles interaction history into subtask language and optional 2D grasp, place, and move hints that bind its physical intention at different time scales. To teach this steerability to the VLA, we augment demonstrations with structured hint labels and train under condition dropout, spatial noise, and temporal jitter to tolerate imperfect Agent outputs. On LIBERO-Mem, without depth, online geometry, or planner-based object motion, 2AM reaches 76.3% average completion, a 61.5-point improvement over the strongest reported baseline of 14.8%, together with 63.0% relaxed and 11.8% strict success. These results show that task memory can remain Agent-side. They further show that Action Model capability depends not only on what the policy has learned, but on how precisely the Agent can steer it.

Automatic Lyric Transcription (ALT) remains substantially more challenging than speech recognition due to melodic variability, rhythmic irregularity, and accompaniment interference. This is heightened in low-resource languages like Greek, where no prior benchmark for ALT exists. We present the first controlled study of Whisper adaptation for Greek ALT, investigating model scaling effects, task composition via multitask training in transcribe-translate ratios, and two-stage speech-to-singing adaptation. We also curate a segment-level aligned singing dataset based on the Greek Audio Dataset (GAD) using source separation and CTC forced alignment. Results show that scaling consistently improves performance, while multitask learning acts as a beneficial regularizer primarily for smaller-capacity models. The 2-stage adaptation in Whisper Large-v3 achieves a Word Error Rate (WER) of 27.2%, a significant improvement over zero-shot baselines, establishing the first Greek ALT benchmark.

Recently, large language model (LLM) based ASR models have achieved significant progress, yet they generally lack support for multi-speaker scenarios, where the cocktail party problem remains a critical bottleneck for further advancing ASR. Existing TS-ASR methods, including end-to-end architectures with speaker embeddings and latest LLM-based explorations suffer from degraded single-speaker performance and the inability to reject when the target speaker is absent. In this paper, we propose Xiaomi-CocktailASR-1, an LLM-based end-to-end TS-ASR architecture. By utilizing reference speech as voiceprint prompts, it directly transcribes the target speaker's speech without requiring speech separation. Xiaomi-CocktailASR-1 maintains competitive performance in single-speaker scenarios, comparable to mainstream ASR models. It also features a negative sample rejection capability, outputting empty text when the target speaker is absent from the mixed speech. Additionally, Xiaomi-CocktailASR-1 supports a Chain-of-Thought (CoT) reasoning mode to provide explicit reasoning steps. Extensive experiments on various synthetic and real-world multispeaker benchmarks demonstrate that Xiaomi-CocktailASR-1 achieves state-of-the-art performance, effectively addressing the cocktail party problem through a unified architecture that balances multispeaker and single-speaker recognition accuracy, along with rejection capability.

Pretrained generative robot policies learn expressive action priors from demonstrations. However, existing reinforcement learning methods only steer the noisy space but fail to modulate intermediate action representations during the generation process, resulting in performance degradation and inefficiency. To address this limitation, we propose a novel Dual-Latent Space Reinforcement Learning (DLSRL) framework, which complements initial-noise steering with representation-level control inside the frozen generator. Specifically, our actor network predicts two distinct latent variables: an initial-noise latent variable that steers behavior generation, and an action-representation latent variable for intermediate feature modulation. Moreover, this representation latent variable is mapped to adapter features and ingeniously injected into the hidden states of intermediate action tokens via residual connections. Our dual-control design enables direct adjustment of action representations without updating the base policy. Experiments across generative policy architectures and robotic manipulation tasks show that DLSRL effectively accelerates online robot policy adaptation and achieves competitive performance. Our code is available at \href{https://github.com/xianchaoxiu/DLSRL}{https://github.com/xianchaoxiu/DLSRL}.

Autonomous property inspection requires more than robust robot navigation: a deployable system must connect heterogeneous sensing, reusable autonomy capabilities, multimodal scene understanding, human interaction, and enterprise response within a traceable operational loop. Existing quadruped inspection systems commonly integrate these functions through task-specific interfaces, making contextual coordination, knowledge reuse, and controlled adaptation difficult. This paper presents \textit{Harness Robotic OS} (HROS), a unified embodied-agent runtime, and Argos, its realization for residential-community inspection. HROS organizes the system into robot runtime, embodied autonomy skills, cognitive agent runtime, and interaction and operations planes. A shared context connects physical state with agent reasoning; streaming ASR/TTS supports voice-based mission interaction; hierarchical working, episodic, and semantic memory preserves operational knowledge; and a safety-gated self-evolution loop converts execution traces into versioned candidate updates without permitting unconstrained online modification. The Argos prototype integrates a Vbot quadruped, Fast-LIO2 localization and mapping, Hobot-Stereo depth perception, PCT-Planner global planning, EGO-Planner local motion generation, and OpenClaw-orchestrated Qwen3-VL inspection analysis. Experiments in a residential property environment achieved 100\% waypoint reachability, outdoor localization error below 10~cm, local obstacle-response latency below 200~ms, representative hazard-detection rates of 85--95\%, and 99\% success in alarm delivery and structured-report generation. These results validate the deployed navigation and inspection closed loop, while HROS provides an extensible software foundation for memory-augmented, voice-aware, and continuously improvable embodied inspection agents.

Self-triggered reinforcement learning control (RL-STC) learns the sparsest control schedule that preserves Lyapunov-decreasing stability under a Run-Time Assurance (RTA) override. We invert this: an adversarial RL agent learns the sparsest jamming or Denial-of-Service (DoS) schedule that destabilizes the closed loop, with a Lyapunov-increase admissibility predicate mirroring the defender's safety certificate. We prove a plant-property lower bound on the minimum jam count required for an immediate hold-last medium-access-control adversary to force a crash against a self-triggered controller (STC) satisfying a Lyapunov contract, and recover a certificate-level analog of the consecutive-grouping optimality of prior count-budget DoS scheduling as a corollary. This extends the DoS-scheduling count-budget analysis from periodic and linear-time-invariant to STC controllers. Empirically, we train against four fixed defenders per plant (one Linear Quadratic Regulator (LQR) and three RL-STC) on Pendulum, CartPole, and Quadrotor2D. The learned adversary is the only adversary that crashes every defender on every plant at $100\%$: greedy misses Quadrotor2D LQR on $42\%$ of episodes and periodic misses Pendulum LQR on $97\%$. On jam-time-per-failure it beats baselines by up to $2.8\times$, and shows its widest absolute margin on Quadrotor2D LQR. Robustness ablations show that Gaussian observation noise exceeding the initial-state magnitude and position-only observation both preserve $100\%$ failure rate and keep the learned adversary strictly ahead of both baselines on jam-time-per-failure.

Reinforcement learning (RL) policies for unmanned aerial vehicle (UAV) placement in mmWave and sub-terahertz networks are typically trained on simplified analytical channels. We quantify the resulting sim-to-real gap on a real urban map of Doha, Qatar, at carriers {28, 140, 183, 300} GHz and altitudes {50, 75, 100, 125} m, evaluating three channel pipelines: an analytical model (FSPL + atmospheric absorption + cuboid LoS), full Monte-Carlo ray tracing in Sionna RT with ITU-R P.676-13 absorption, and a deterministic-LoS hybrid that reuses Sionna's mesh under a closed-form path-gain expression. We formalize the gap on the spatial SNR distribution via four metrics, namely bias, RMSE, Jensen-Shannon divergence, and optimum-deployment displacement. Three findings emerge: at 28/140 GHz, $\sim$70% of the apparent -5.6/-4.8 dB Sionna bias is Monte-Carlo undersampling and shrinks to -1.7/-1.5 dB after mitigation; at 183 GHz a -9.2 dB residual isolates the atmospheric absorption / ITU-R P.676 line-shape disagreement; at 300 GHz the stochastic ray tracer agrees with the analytical model only coincidentally, with a +3.8 dB structural offset exposed by the deterministic-LoS pipeline. Across all carriers the linear-domain regret of the analytical-trained policy stays $\geq$ 0.93, indicating practical near-optimality but with a carrier-resolved SNR bias that warrants explicit reporting.

Passive mechanical interfaces offer a lightweight alternative to actuated manipulators for quadruped payload carrying, but their impedance directly couples the payload dynamics with the locomotion pattern. This paper analyzes how passive-arm stiffness-damping selection affects payload-carrying locomotion under different gait and payload conditions. We compare damped and underdamped passive-arm impedance configurations in simulation during flat-ground locomotion. For crawl gaits, where the support polygon remains well defined, the results show that underdamped impedance increases passive-joint oscillations and can reduce the ZMP margin with respect to the support polygon. Trot is retained as a dynamic excitation case for the passive arm, but it is not used for direct ZMP-margin stability comparison. The results are summarized in gait-payload-stiffness-damping maps, where ZMP-margin reduction is evaluated for crawl gaits and trot is retained only as a passive-arm excitation case.

Multi-agent robotic manipulation tasks require coordination among agents to satisfy task-level temporal, logical, and safety constraints. Recently, diffusion policies have been used to perform the task. However, they still suffer from desynchronization, incorrect action ordering, and coordination failures in tasks that require simultaneous or sequential multi-agent interaction. Therefore, LTLDiff is proposed as a framework that combines Finite Linear Temporal Logic (LTLf) specification learning for both the generation of demonstrations and learning via diffusion policies. Each task has a specific LTLf formula that is learned from a set of natural language instructions using a large-scale language model. To enable a fixed-dimensional vector embedding of the learned specification from the language model, LTLf uses an abstract syntax tree representation scheme. This embedding of logic serves as a condition for (i) logic-guided data collection and (ii) diffusion-based policy training, encouraging trajectories that are consistent with the desired ordering and coordination requirements. Experiments on multi-agent LTLDiff manipulation tasks demonstrate improved task success rates compared to the baseline. Together, these contributions demonstrate the effectiveness of LTLDiff for coordinated multi-agent manipulation.

LM-agent benchmarks increasingly function as interactive evaluation infrastructure. Agents observe state, call tools, modify workspaces, submit artifacts, and receive rewards from outcome procedures. This interactivity makes evaluations vulnerable to reward hacking: an agent improves its measured score by exploiting the reward-relevant trajectory instead of solving the intended task. Existing defenses rely largely on task-specific patches, prompt instructions, or post-hoc detectors. They do not provide reusable evidence that a concrete run remained within its intended evaluation boundary. This paper presents BenchShield, a model-backed instrumentation layer for reward integrity in LLM-agent evaluation. BenchShield grounds detection in a finite lifecycle model of an evaluation's reward-relevant events. Within the benchmark infrastructure, two complementary analyses operate over this model. A static, phase-aware taint analysis exposes reward-hacking paths before a run. Its runtime counterpart uses infrastructure-side evidence to attribute concrete agent use and emit evidence-backed claims. We construct BenchShield Trajectories, a human-labeled corpus of 456 adjudicated trajectories from more than 31,000 public agent runs across three benchmarks. Compared with an agentic hackability scanner baseline on the same tasks and model, BenchShield improves full-chain recall from 23-94% to 77-100%, same-vector coverage from 16-56% to 43-78%, and reduces per-task cost by up to 65%. Its runtime analysis achieves 96% accuracy in detecting reward hacking from infrastructure-side evidence.

Long-horizon goal-conditioned reinforcement learning delegates control to a high-level module that proposes subgoals, but existing subgoals are implicit byproducts of value functions or latent actions, tied to the executor that produced them. We study a different object: a route-conditioned order of unavoidable stages that every successful executor must traverse, recoverable from offline trajectories and belonging to none of them. Its defining properties are topological: an unskippable stage is a separating set that every admissible path must cross, and a loop in free space forces a route choice. We read the two by homology in dimensions 0 and 1 over a transport-weighted carrier built from successful trajectories, yielding an enumerable gate set with shell-level certificates; the certified gates are what we call topological necessities. Certified gates enter the decision loop as a recursive topological gate hierarchy. Under a fixed, isomorphic free space, the object survives executor replacement: gates frozen on PointMaze data transfer without retraining to Ant and Humanoid, attaining the highest Humanoid aggregate under a unified interface (96.1), with +36.0 over a map-privileged reference on the multi-route task (p=1.4e-5); the planner saturates PointMaze (100+/-0) and matches or exceeds the strongest baselines on AntMaze (giant +22.9) and Kitchen (+15.8/+12.6).

We propose a pragmatic information theory unifying communication, control, and decision-making. Its core is the isoteleia mapping, formalizing equifinality: distinct semantic paths leading to the same optimal action are pragmatically equivalent. This induces a three-tier hierarchy of syntactic, semantic, and pragmatic information, each abstraction discarding task-irrelevant distinctions. We develop pragmatic entropy, up/down mutual information, channel capacity, and rate-distortion, and prove three coding theorems generalizing Shannon's classical results. We introduce pragmatic value (VoI) and cost (CoI) of information as decision-theoretic duals to rate-distortion and capacity, respectively, and formulate a Lagrangian dual framework for cross-layer optimization. The pragmatic efficiency bound $\mathcal{E}_p(λ)=\sup_R[Φ_p(R)-λ\,\mathrm{CoI}_p(R)]$ quantifies the maximum net utility any resource-constrained intelligent system can extract, thereby establishing a fundamental behavioral capacity limit---generalizing Shannon's symbol-level capacity to goal-directed action. Extensions to continuous messages yield closed-form Gaussian expressions, while dynamic settings are addressed via a Bellman equation for sequential decision-making. This framework provides a rigorous foundation for task-oriented communication, networked control, autonomous systems, and embodied AI, shifting focus from symbol fidelity to the effectiveness of information in guiding actions, and offers a unified mathematical language for next-generation intelligent systems.

Imitation learning has achieved impressive results in robotic manipulation, yet most existing approaches assume clean backgrounds and lack explicit mechanisms for obstacle-aware motion generation. Extending such policies to cluttered, real-world scenes with unstructured obstacles remains a key generalization challenge. We present ObstaDiff, a decomposed diffusion-policy framework with a lightweight obstacle-aware visual encoder. ObstaDiff extracts a structured target-obstacle-background representation, enabling the downstream alignment policy to generate end-effector trajectories toward a target-centered bottleneck pose while reasoning about surrounding obstacles. We evaluate ObstaDiff on 61 real-robot greenhouse trials per method (366 executions in total). ObstaDiff achieves 75.41% average task success and 8.20% average obstacle collision rate, outperforming representative imitation-learning baselines and improving generalization in cluttered agricultural scenes.

Vision-language-action (VLA) policies leverage pretrained vision-language backbones to achieve strong cross-task generalization. A leading design couples this backbone with a dedicated continuous action head trained via diffusion or flow matching. However, such heads rely on iterative multi-step sampling, for example 10 Euler steps in $π_{0.5}$. This creates an inference bottleneck that produces stop-and-go movement in the robot and slower task completion. We introduce IMLE-VLA, which replaces the iterative action head with a single-step conditional generator trained via conditional Implicit Maximum Likelihood Estimation (cIMLE). The cIMLE objective promotes multimodal action coverage, avoiding the mode collapse of naive regression heads while eliminating multi-step sampling entirely. When IMLE-VLA is applied to $π_{0.5}$, it increases inference frequency 3.67x (55 Hz vs. 15 Hz), enabling up to 11x higher action throughput. On the 40-task LIBERO benchmark, IMLE-VLA achieves the highest average success rate (98.0%) among all baselines while leading in inference frequency. Under the test-time perturbations of LIBERO-plus, IMLE-VLA retains $π_{0.5}$'s robustness while other baselines degrade sharply, confirming that the cIMLE head preserves generalization. Real-world experiments on a Franka Emika Panda across four tasks demonstrate smoother motion (2.2x to 3.0x lower jerk) and faster task completion, with IMLE-VLA outperforming $π_{0.5}$ on every task and reducing average VLA inference time per episode by 3.9x to 6.6x. Videos and code are available at https://kianhk6.github.io/IMLE-VLA/

Planning trajectories for robot manipulators under kinematic equality constraints restricts feasible motions to a measure-zero submanifold of the configuration space, requiring special algorithmic treatment. A promising strategy is parametrizing the set of feasible configurations using analytic inverse kinematics (IK). Bespoke analytic IK functions can be written to be differentiable, a necessary property for gradient-based trajectory optimization. But the vast majority of IK functions are computed by automated meta-solvers like IKFast, and are difficult to modify for differentiability. We present a new approach for computing gradients of analytic IK parameterizations: we leverage the inverse function theorem to recover the desired gradients from the ordinary forward kinematic Jacobian. Furthermore, we present a least-squares domain extension and an optimization-amenable description of the reachability constraint, which preserves gradient signal outside the reachable workspace. We demonstrate the efficacy of our approach through numerical experiments and downstream tasks, including a hardware demonstration of an RB-Y1 picking up a box and placing it on a table. Project website: https://cohnt.github.io/inverse-function-theorem-parameterization/

Reacting to sudden physical hazards (catching a slipping plate, dodging a falling knife) is both a meaningful test of embodied intelligence and a hard requirement for deploying multimodal large language models (MLLMs) as the decision coreof household robots. Existing evaluations, however, probe intuitive physics passively through question answering over videos, or target deliberate, long-horizon tasks such as navigation and rearrangement; none measure whether a model can turn physical understanding into immediate, safety-critical action. We introduce ReactHuman, the first physics-grounded benchmark for human-like reactive decision-making, in which the evaluated MLLM acts as the brain of a simulated humanoid facing sudden household hazards; it spans 17 event families and over 1,000 bit-for-bit reproducible scenes with exact, annotation-free ground truth derived from 240 Hz rigid-body simulation, including adversarial objects whose appearance contradicts their physics (a foam anvil, a steel apple). We further design a five-metric suite that scores each reaction along three axes: reasonable, safe, and physically grounded. We physically execute every committed plan so that decisions have observable consequences. With this harness we evaluate seven representative MLLMs. Results show that reactive safety is far from solved: models mishandle roughly one hazard in three, act from fixed dispositions rather than the observed scene, trust appearance over motion, and miss interception points at meter scale even when the chosen action is correct; none of these failures shrink with model scale. ReactHuman thus offers both a fine-grained diagnosis and a scalable training signal toward physically grounded, safety-aware embodied agents. The benchmark can be found here: https://huggingface.co/datasets/Alan123/reacthuman-benchmark-scaled

Enabling robots to perform musical instruments with human-level expressivity represents a frontier in bridging the gap between mechanical precision and artistic interpretation. Despite advances in robotic dexterity, replicating the fluid finger transitions and nuanced dynamic control characteristic of human pianists remains a significant challenge. Through a reinforcement learning-based control framework, we demonstrate that a dexterous robotic hand can achieve high-fidelity performance across a diverse piano repertoire. Central to our approach is a graph-based optimization strategy that guides the robot to generate natural pre-press and key-press fingering strategies that closely resemble human movement patterns. To achieve expressive sound production, the control system is coupled with a physics-inspired acoustic model that modulates keypress velocity to accurately reproduce the dynamic variations specified in musical scores. Quantitative evaluations demonstrate that our expressive control model significantly outperforms baseline methods in both finger morphology similarity and dynamic velocity accuracy. In a perceptual test involving participants from diverse listener groups, performances generated by our system are significantly preferred over baseline robotic performances and are indistinguishable from human performances for non-professional audiences. Furthermore, extensive experiments across multiple musical styles confirm that our method maintains high note-level accuracy while achieving expressive performance. Our approach provides a robust pathway for robotic systems to move beyond mere mechanical accuracy, elevating robotic musicianship to a level of expressive performance comparable to human pianists.