Physical AI Brief

Recent video world models generate increasingly realistic and interactive visual experiences, yet lack reliable mechanisms for maintaining persistent world state and enforcing programmable rules over extended interactions. We introduce Programmable World Model, a framework that decouples world-state evolution from visual observation generation. An agent translates natural-language instructions into executable programs that specify entity states and state-transition rules, enabling direct control over individual entities and their interactions. A lightweight engine executes these programs to update and maintain an explicit, persistent global world state, including off-screen entities and non-visual attributes. To connect world state with visual generation, we introduce state-augmented 3D oriented bounding boxes (OBBs) as an intermediate representation. This representation, together with the target camera trajectory, is deterministically compiled into pixel-aligned spatiotemporal conditioning signals for a pretrained video model serving as the generative renderer. This design allows users to create playable games with predefined mechanics, direct control over individual entities, and persistent world state throughout gameplay. We further introduce CombatStateBench, a benchmark for evaluating programmable world models. On CombatStateBench, our method achieves 94% Count Accuracy and 98% State Accuracy, substantially outperforming existing interactive video world models while supporting coherent long-horizon generation. These results demonstrate the effectiveness of separating explicit state evolution from generative rendering for building persistent, programmable worlds.

A research idea may be novel, coherent, and scientifically plausible, yet its proposed method may remain insufficiently specified for faithful implementation. We study the codification readiness of implementation-facing research-method specifications, defined by whether they provide sufficient methodological information for a competent implementer or coding agent to construct the intended method without unsupported assumptions. We construct evidence-grounded specifications and their supported resolutions from papers, codebases, issue threads, and reproduction artifacts. We introduce IdeaAMBIG, a benchmark of 660 evidence-grounded instances: 163 real-world gaps from reproducibility reports and GitHub issues, and 497 controlled synthetic gaps injected into codification-ready references. IdeaAMBIG evaluates three capabilities: codification-readiness assessment, defect localization, and clarification action generation. Defect localization receives only the specification, whereas clarification additionally receives the annotated defect. Across 13 LLMs, the best model achieves 9.6% Macro Defect Recovery Rate on real-world instances but 80.6% Macro Clarification Action Success Rate when given the defect. In an oracle study, supplying the gold resolution raises the downstream codification-ready rate from 14% to 98%. Across all evaluated models, defect localization is the main bottleneck, with stronger clarification given the defect.

Foundation vision-language models (VLMs) exhibit broad intelligence about the world, yet translating this intelligence into robot control remains challenging. We present Show-Harness, an Embodied Harness that enables VLMs to "play" robots through a compact semantic interface linking intent to action. Show-Harness exposes discrete semantic action units that VLMs can naturally reason over, while embodiment-specific interpreters deterministically ground them into local robot actions, keeping the VLM directly responsible for fine-grained physical decisions. Through the same interface, Show-Harness demonstrates the feasibility of (1) directly unlocking closed-source frontier VLMs for zero-shot robot control, and (2) adapting small-scale open-source VLMs for low-cost deployment with just a few GPU-hours of fine-tuning. We further develop GUMI (GUI Manipulation Interface), which extends the same semantic action space to GUI-based demonstration collection, allowing humans and agents to "play" robots across embodiments without specialized teleoperation hardware. Extensive experiments show that Show-Harness-equipped VLM agents generalize robustly across tasks, embodiments, and environments, outperforming representative agentic and VLA paradigms. These results suggest that the right interface can unlock substantial embodied capability from foundation VLMs, without requiring additional model capacity or costly embodiment-specific pretraining.

Scientific workflow management (WMSs) systems automate execution, yet orchestrate using fixed, hand-tuned rules. LLM agents promise more autonomous orchestration, but it remains unclear where to introduce agentic reasoning, how to bound its risk, and when it actually helps. We present Avatar, an actor-based architecture comprising an orchestrator, an executor, and a provenance monitor. Each actor's decision policy is pluggable (rule-based or LLM-backed) via a single adapter-validated action catalog, so conventional and agentic control run on the same core across different WMSs. We present an implementation using the Academy framework and evaluate Avatar across three workloads. We observe that Avatar's rule mode reproduces native execution, with a single unchanged core running all three. Moreover, LLM-backed Avatar reports a reduction of compute wastage by $55\%$ and cuts GPU-busy time by $40\%$. Overall, we envision Avatar as a step toward workflow systems that reason about their own orchestration rather than follow pre-fixed rules.

Action-conditioned latent world models predict future visual representations, enabling zero-shot goal-conditioned robot planning and control. However, their predictions for fine-grained spatial and rotational actions are unreliable for full 7-DoF end-effector control. To address this gap, we introduce DUET-DINO, a simultaneous cross-view latent world model that jointly learns action-conditioned predictions from static side- and wrist-camera observations through cross-view conditioning. By exploiting complementary global scene and gripper-centric information, DUET-DINO enables latent planning over the full 7-DoF action space. Across spatially diverse reach, orientation-intensive angled-reach, and multi-goal grasp-and-lift tasks, DUET-DINO consistently outperforms single-view and independent dual-view baselines, achieving 92% success on reach, 72.5% on angled-reach, and 60.0% on lift tasks. DUET-DINO is trained from scratch on DROID and RoboArena datasets and generalizes robustly under visual distribution shifts. We further show that while V-JEPA 2 wrist-view predictions underestimate visual dynamics induced by fine-grained actions, DINOv3 predictions better capture action-conditioned scene changes, leading to stronger downstream planning. The code and model checkpoints will be open-sourced. Project page: https://utn-air.github.io/DUET-DINO

In real-time colonoscopy, ground-truth annotations are unavailable at inference, so polyp segmentation models can fail silently. We propose Referee-Based Quality Estimation (RBQE), a reference-free framework measuring agreement between a primary segmentation model and an independently trained referee on the same image. RBQE is evaluated on a standardized 1,223-image external benchmark drawn from four public datasets, using four referee configurations chosen to separate two design axes: referee independence and architectural diversity. Using a common Agreement Dice descriptor, a same-architecture referee differing from the primary model only in random initialization already yields a useful reliability signal (ROC-AUC = 0.923), showing that independent training alone is sufficient. Cross-architecture referees improve further: SegFormer-B0 achieves the strongest performance (ROC-AUC = 0.960), significantly outperforming the same-architecture control and UNet++, and exceeding a representative Test-Time Augmentation baseline by 0.055 ROC-AUC under an identical protocol, whereas a prompt-coupled MedSAM referee underperforms despite maximal architectural diversity. Because empty-mask agreement is trivially separable, we also report a restricted evaluation excluding such cases: ROC-AUC falls to 0.876 (SegFormer-B0, 1,046 images) and 0.783 (same-architecture control, 975 images), yet RBQE's margin over both baselines widens on this identical subset. RBQE additionally increases the mean Dice of retained predictions as low-agreement cases are progressively rejected, supporting selective prediction, and requires only one additional deterministic referee forward pass at inference. Our study therefore supports cross-model agreement as a practical, interpretable reliability framework for automated polyp segmentation.

Enterprises deploy systems, not checkpoints. Usable capability depends jointly on weights, serving route, precision, output contract, and harness, yet all 18 audited benchmarks score advertised model identifiers. We treat this as measurement error and give a protocol that makes it reportable. It has three parts. A gold-blind capability-binding preflight verifies that a route can execute the evaluation contract before any task reaches it; a reliability-inclusive first-pass scoring rule keeps failure in the score while keeping unsupported capability out; and adjudication is structurally score-blind. We call the protocol IB2 and release its algorithms, classification tables, request contract, and manifest schemas. Its reference instantiation, 128 locked tasks and 987 assertions over document, spreadsheet, chart, tool and database work, stays sealed: the procedure is the artifact, not the corpus. Across eleven systems, four results. Capability availability is measurable: two complete single-route runs on identical weights later failed distinct predicates of the finalized binding gate, while a third passed that gate before a fresh run. The advertised identifier exposed neither limit. Discrimination is not uniform: four of seven suites saturate under a six-system band, with the spread almost entirely from governed database work and multi-tab joins, so we report interval-backed resolution groups, not ranks; two of the nominal five-label output's four cuts fail multiplicity adjustment. Serving-arm choice moved one declared revision and precision from 77.38 to 82.54, paired interval [0.11,10.60], though the arms differ in access mode, harness generation, and the serving tool-call parser, and harness generation is a property of our evaluator, not any endpoint. Excluding failed responses from denominators changes the point ordering, so reliability inclusion changes a conclusion, not its wording.

More Electric Aircraft require fast and reliable monitoring of high-frequency electrical networks, yet most power quality disturbance and fault diagnosis methods are developed for conventional 50 or 60 Hz grids. This work presents a hardware-aware deep learning framework for multiclass detection of electrical faults and power quality disturbances in a 400 Hz aerospace power system. A high-fidelity simulation model inspired by the Boeing 787 electrical architecture generates voltage and current waveforms for 21 normal, disturbance, switching, open-circuit, and short-circuit conditions. Two datasets, each containing 73,500 samples, are formed from one-dimensional time-series signals and short-time Fourier transform time-frequency representations. Signal-processing augmentation, domain randomization, and class-specific generative adversarial networks increase waveform diversity, and the time-series dataset is released through IEEE DataPort. We compare 1D and 2D convolutional neural networks, long short-term memory networks, CNN-LSTM hybrids, ResNet, MobileNet, and VGG models under common training conditions. A compact ResNet provides the best accuracy-complexity tradeoff, achieving 96.94 percent software test accuracy with 175,685 parameters. After 8-bit quantization and deployment on a Xilinx Zynq UltraScale Plus MPSoC ZCU102, the model achieves 95.87 percent accuracy and a measured mean neural-network accelerator latency of 6.90 ms per input record. The results establish simulation-based, accelerator-level feasibility for embedded edge AI in aircraft electrical health monitoring and motivate future end-to-end data acquisition and experimental validation.

Recovering the linear relationships that govern a system from noisy measurements is a basic task across the physical and engineering sciences. Because every measured variable may carry an unknown amount of noise, classical regression must commit in advance to a set of structural assumptions: ordinary least squares requires a declared input-output partition with input variables being noise-free, total least squares assumes equal noise variance across all variables, and generalized total least squares additionally requires the noisy-variable partition and variances to be known beforehand. Kalman~\cite{Kalman:1982} showed that any procedure returning a unique linear model from inexact data must rest on such unverifiable a priori assumptions -- ``prejudices'' -- that cannot be checked against the data itself, and that removing them leaves the identification problem fundamentally indeterminate. Whether these prejudices can instead be resolved directly from the data has remain unresolved. Here we show that an iterative generalized-eigenvalue algorithm, QZ-IPCA, recovers the noisy-variable partition, noise variances, number of linear relations, and regression coefficients of a multivariate linear system simultaneously, using only the raw data. Across all possible exhaustive noise configurations of a five-variable benchmark network, QZ-IPCA correctly identifies model structure and recovers coefficients with error below 6.4\%. It outperforms ordinary least squares even when given the best partition, and succeeds in rank identification precisely where standard total least squares falls once noise variances differ across variables. These results show that the assumptions conventionally required for multivariate regression are not necessary, recasting model identification as a problem solvable from data geometry alone.

Automated plant disease diagnosis is increasingly deployed on farmer-held devices in regions where agronomic expertise is scarce and network connectivity is unreliable. Three obstacles limit its practical value: public benchmarks are dominated by a small set of non-native crops, region-specific datasets are rarely validated by domain experts, and the architectures that reach competitive accuracy carry parameter budgets that are unsuited to low-cost hardware. We propose AgroVisNet, a compact convolutional network trained from scratch, together with BD-PlantDX, an expert-validated benchmark of 12,432 field images spanning 12 classes of radish, potato and pointed gourd in healthy and diseased states, collected across the Bogura and Nilphamari districts of Bangladesh. AgroVisNet couples grouped bottleneck residual blocks carrying sequential channel and spatial attention with multi-scale depthwise blocks and a dual-pooling classification head, reaching 290,572 trainable parameters. On BD-PlantDX the model attains 99.52% test accuracy and 99.52% weighted F1, exceeding all six ImageNet-pretrained lightweight backbones evaluated under an identical protocol while using 8.7 to 16.8 times fewer parameters and 1.3 to 8.5 times fewer multiply-accumulate operations. Exported for deployment, the model quantises to a 0.46 MB full-integer network at a 0.22 percentage-point accuracy cost and classifies an image in 8.40 ms on a single CPU. Across five random seeds accuracy remains at 99.57 +- 0.10%, a ten-variant ablation isolates the contribution of each component, and the same architecture transfers without redesign to two independently collected datasets at 98.71% and 99.05% accuracy. Grad-CAM evidence indicates that predictions rest on lesion-bearing leaf regions rather than on background cues.

Joint-Embedding Predictive Architecture (JEPA) world models learn a compact latent representation of the world that supports prediction and planning, but their capability to learn physics and generate physically realistic dynamics remains hitherto untested. In this work, we introduce SemiGroup-JEPA (SG-JEPA), which extends the LeWorldModel framework by supplying the parameter governing the physics to the temporal model via action-conditioning and jointly training an encoder and predictor through an autoregressive latent rollout. To evaluate the model's ability to generalize out of distribution, we design dynamical tasks under different gravitational fields that, despite obeying the same physical law, exhibit qualitatively different dynamics, ranging from floating motion in weak gravitational fields to rapid bouncing in strong ones. In contrast to DINO-WM, SG-JEPA reduces open-loop prediction error by up to 2 times on two-dimensional datasets, and increases control success rate up to 2.5 times for three-dimensional robotic datasets, for which we train independent diffusion policies. To explain this advantage, we develop a linear feature model that separates local law-conditioned error from its recursive amplification under rollout. Guided by this model, we find that back-propagating the multi-step rollout loss into the representation trains the encoder to keep the features that the predictor can carry forward, and that those are the features the dynamics depend on, so most of the gain comes from the encoder learning better features rather than from the predictor learning better dynamics. See project page at https://sg-jepa.github.io.

Real-world GUI usage frequently involves workflows that span multiple devices and platforms, requiring the transfer of intermediate results, maintenance of shared state, and coordination across heterogeneous environments. However, existing GUI benchmarks overwhelmingly evaluate agents on single-device, statically defined tasks, thus leaving such cross-device capabilities largely unexamined, resulting in an overly optimistic assessment of agents' readiness for real-world usage. We introduce JarvisGUI, a dynamic benchmark that evaluates GUI agents on cross-device workflows requiring coordinated interaction across heterogeneous platforms, including Android, Windows, and Ubuntu. Specifically, JarvisGUI formulates GUI tasks as input-output transformations under a lightweight type system, which allows us to automatically compose multi-step, cross-device workflows and dynamically evaluate agent performance within a unified framework. By evaluating agents in virtual environments spanning multiple operating systems, JarvisGUI reveals that state-of-the-art open-source GUI agents struggle with the state-transfer awareness, cross-platform contextual reasoning, and long-horizon dependency management required for real-world workflows, exposing a critical capability gap invisible to existing benchmarks.

Reasoning language models have made substantial advances on a variety of complex tasks, yet their capabilities remain overwhelmingly English-centric: models primarily reason in English regardless of the language they are prompted in. This is inaccessible for non-English-speaking users, risks losing the intent of the original question, and forgoes knowledge more readily expressed in the target language. In this work, we advance L2 reasoning, the ability of a model to reason consistently in the language of the user's prompt, thus building an in-language bridge between the prompt and the answer. We approach this problem from a data-centric angle, investigating how to optimize data composition and scheduling in SFT for reasoning generalization. Building Tiny Aya L2-Thinker at 3.35B scale, we achieve an L2 reasoning rate above 93% across 60 languages on 6 benchmarks spanning math, commonsense reasoning, instruction following, open-ended generation, and cultural reasoning while keeping performance strong. We show the path to generalizing L2 reasoning to held-out languages goes through broader language coverage, readily available multilingual non-reasoning data, and a sufficient English reasoning backbone. These findings indicate that reasoning is a language-agnostic behavior that can be transferred across typologically diverse languages through careful data mixing and without requiring reasoning supervision in every target language. We release our model weights and multilingual reasoning data to support further research on accessible, in-language reasoning.

While Large Language Models (LLMs) have demonstrated impressive capabilities, they often struggle with extremely long contexts due to fixed context limits. To address this, sequential approaches like MemAgent extend the effective context by reading text in segments and iteratively updating a fixed-size memory. However, this sequential paradigm suffers from high latency and requires costly reinforcement learning (RL) training, which can lead to overfitting on specific datasets. To overcome these limitations, we propose ConvMem, a training-free, highly parallelizable framework that reformulates long-context reasoning as a hierarchical convolution. Inspired by CNNs, ConvMem treats an LLM prompted with a specific query as a convolutional kernel. This kernel summarizes text segments hierarchically, shortening the reasoning path from a linear chain into a logarithmic tree. Specifically, ConvMem integrates \textit{Configurable Strides} and \textit{Skip Connections} to ensure robust evidence capture and propagation, while employing \textit{Multi-Kernel Convolution} to decompose complex queries into disentangled semantic channels. This design not only mitigates error accumulation but also enables massive parallelization across both text segments and reasoning threads. Experiments on RULER-HotpotQA and RULER-2WikiMultiHopQA demonstrate that ConvMem outperforms training-free baselines and avoids the risk of overfitting to parametric priors often observed in RL-trained models on out-of-distribution tasks.

Large Language Models (LLMs) can memorize and reproduce sensitive, copyrighted, or otherwise undesirable training content, creating privacy, safety, and regulatory concerns. Machine unlearning offers a practical alternative to full retraining, but many existing methods apply broad or fixed parameter updates that can degrade utility and remain brittle under deployment changes such as post-training quantization, where forgotten knowledge may partially re-emerge. We propose Forgetting Only What Matters via Unlearning Layers (FOM-UL), a layer-level unlearning framework that selects transformer layers using a forget-to-retain significance score. This score identifies layers with high influence on the forget set and low sensitivity to the retain set, allowing FOM-UL to concentrate updates where they are most effective while leaving most of the model unchanged. This targeted update strategy improves the forgetting-utility trade-off and provides an empirical path toward quantization-resilient unlearning by reducing the chance that small, diffuse updates are erased by low-bit rounding. Across TOFU, KnowUnDo, and MUSE-style evaluations, FOM-UL reduces residual memorization compared with strong GA, NPO, KLD, SURE, ReLearn, and LUNAR-based baselines while preserving retain-set utility close to the vanilla model. Under 8-bit and 4-bit post-training quantization, FOM-UL maintains stronger memorization suppression and utility preservation than competing methods, and adversarial prompt evaluations show lower recovery of forgotten content. Overall, FOM-UL provides an efficient unlearning strategy that improves targeted forgetting, utility preservation, and deployment robustness without claiming formal guarantees of erasure.

This study develops a deep reinforcement learning framework for training Unmanned Aerial Vehicle (UAV) agents to navigate and monitor simulated wildfire environments. Results show that agents learn increasingly stable and effective behaviors over time, as demonstrated by converging loss trends, improved reward signals, and more consistent navigation patterns such as fire-boundary tracking. Overall, these findings highlight the potential of deep reinforcement learning (DRL) based UAV systems for autonomous wildfire monitoring and suggest that environmental structure and reward design influence policy effectiveness.

Enterprise data lakes accumulate tables faster than human stewards can document or classify them, leaving columns with missing descriptions and unassigned governance labels. This documentation debt undermines data discovery, access control, and regulatory compliance. We present Glyph, a production system that frames two coupled problems, column description generation and column type annotation for data classification, as cooperating LLM agents orchestrated as stateful graphs. The Descriptor grounds generation in the pipeline source code that produces each column, retrieved on demand from an enterprise GitHub via a reasoning--acting tool loop (active Retrieval-Augmented Generation). The Tagger assigns labels from a governed 275-leaf Data Classification Ontology by running three complementary strategies in parallel (a description tagger, a line-of-business regex tagger, and a metadata tagger backed by a fine-tuned contrastive encoder over a vector database), then fuses their ranked outputs with Reciprocal Rank Fusion (RRF). We fine-tune a 6-layer MiniLM metadata encoder with an in-batch contrastive objective, lifting same-tag retrieval on an in-distribution held-out split from NDCG@10 0.55 to 0.92 (MAP@100 $0.19 \rightarrow 0.90$) relative to the stock base encoder. We report end-to-end multi-label tagging quality under a recall-weighted F2 objective across three evaluation groups, an ablation isolating each strategy and the RRF fusion, and the engineering decisions that distinguish Glyph from prior column-type-annotation work and from commercial value/regex sensitivity scanners: value-free and code-grounded design, per-tag provenance, and graceful degradation. Together these make multi-agent LLM cataloging auditable and operable as a production service.

Background: Emergency Department (ED) return visits are commonly reviewed for quality assurance, but are often limited (e.g., to revisits within 48-72 hours) to increase actionable finding yield while minimizing chart review burden. Those limitations may lead to missed quality improvement opportunities. Methods: We conducted an exploratory, retrospective study of randomly selected ED visits to a multihospital health system having an ED revisit within 1-14 days to the same health system. Given only each visit's primary diagnosis, raters (2-3 clinicians and GPT-4 large language model [LLM]) assessed characteristics of the diagnosis pairs, including the "target": whether a pair warranted further assessment. Informed by rater response analyses, an algorithm leveraging an LLM-populated knowledge graph ("KGA") was created to automatically screen for potentially concerning pairs, then preliminarily assessed. Results: 99 diagnosis pairs were included. GPT-4 responses poorly correlated to clinician raters, rating nearly all (94%) pairs as warranting follow-up (4.4-13.3 times more than clinicians). However, prompt engineering was minimal. Among clinician raters, revisit medical gravity was consistently significantly associated with the target, while a differential diagnosis/complication composite was significantly associated on unadjusted, but not adjusted (though less powered) analysis. The KGA achieved 83-100% positive predictive value for at least one clinician rater determining further assessment was warranted based on the diagnosis pair. Conclusion: These results can inform next steps for improving screening with LLMs like ChatGPT. Further research is warranted to validate this preliminary work's finding that the KGA may enable enhancing the scope and yield of screening without substantially increasing reviewer workload.

Current LLM memory systems treat all personal facts identically, so stores grow without bound while retrieval precision degrades. The core challenge is lifecycle management: which memories should persist, which should be replaced, and at what rate, conditioned on the behavioral type of each fact. Fortunate Recall (FR) is a composable policy layer that classifies personal facts into a 10+1 behavioral ontology and applies category-specific lifecycle policies (differential temporal decay, slot-key supersession, event-time validity, and category-aware retrieval routing) as deterministic functions over LLM-extracted metadata. FR-Bank, our infrastructure-independent implementation, reaches a 76.9% pass rate on LifecycleBench, a new 516-question temporal-disambiguation benchmark, ahead of Mem0, A-MEM, Memory-R1, and MemoryOS (61% to 70.5%), and 75.2% on the full LongMemEval-S under the canonical Wu et al. judge protocol, so lifecycle policies impose no measurable cost on standard retrieval. A pre-registered ablation locates the gains: replacing the typed layer with three generic lifecycle primitives leaves correctness statistically unchanged (-1.7pp, 95% CI [-6.0, +2.7]), so the generic lifecycle metadata carries the correctness advantage, while the behavioral ontology carries calibration, halving downstream confabulation (12.0% vs 24.2%, p<0.001). End-to-end, FR-Bank cuts confabulation from Mem0's 45.1% to 22.4% over answered queries and from 32.2% to 13.0% over all queries while answering more of them correctly (31.2% vs 18.6%); the ranking replicates on the open-weight Kimi K2.5. The decomposition transfers to BEAM, an independently built benchmark: 46.8% correct vs Mem0's 32.9% over 280 questions, with the ontology's benefit concentrated in contradiction resolution and saturating near seven policy clusters. The ontology, benchmark, and code are released.

The growing complexity of content moderation policies presents a critical challenge for their consistent operationalization. While foundation models possess the basic capabilities needed to confront this challenge, whether they can reliably moderate online content remains an unanswered question. In this paper, we systematically compare two competing paradigms for Vision-Language Model (VLM) guidance: an instruction-driven approach where models reason from policy precepts, and an example-driven approach where they generalize from prior precedents. We ground this investigation in ModerationBench, a new benchmark of 4,000 manually annotated, in-the-wild posts from the Bluesky platform. Our experiments reveal that foundation models can substantially outperform Bluesky's deployed moderation system, nearly tripling its $F_1$ score (0.60 vs. 0.22) on Random Posts in the benchmark, with both instruction- and example-driven paradigms achieving comparable peak effectiveness. Our findings thus chart a path toward reliable and adaptable policy operationalization at scale.

Robot actions are temporally correlated trajectories whose frequency components encode motion at different scales with highly non-uniform energy distributions. Yet Flow Matching--based vision-language-action (VLA) models typically generate actions in temporal coordinates, without explicitly modeling or systematically leveraging this frequency heterogeneity. We introduce \emph{FreqFM}, a frequency-conditioned Flow Matching framework for VLA models. It raises action frequency from an implicit trajectory property to an explicit conditioning dimension that spans the entire generation pipeline. Concretely, in DCT frequency coordinates, FreqFM constructs a spectrum-matched source distribution, adaptively balances the objective across frequencies, and constrains per-frequency guidance residuals using the corresponding reference transport scales. FreqFM integrates into existing Flow Matching action experts without changing the VLA backbone. Across LIBERO, LIBERO-Plus, and VLA-Arena, FreqFM consistently improves performance, including a 9.3-point gain on LIBERO-Plus, and further demonstrates its effectiveness on six real-robot tasks.

The coordination of Automated Guided Vehicles (AGVs) in high-density industrial environments represents a critical challenge within Logistics 4.0, as traditional traffic management methods often lead to inefficiencies caused by negotiation-based priority assignment. To overcome the resulting limitations, this paper presents an innovative AGV traffic management system based on a Lifelong Multi-Agent Path Finding (L-MAPF) algorithm operating on roadmaps generated with Non-Uniform Rational B-Splines (NURBS) curves. The approach guarantees locally optimal coordination and ensures safe operation of large and heterogeneous AGVs. Building on this concept, the proposed framework integrates a modified version of the Bounded Horizon Conflict Based Search (CBS) technique within a Rolling Horizon Conflict Resolution strategy, utilizing an extended time horizon for each agent to enable effective conflict resolution in corridors identified by a topological map. In contrast to state-of-the-art methods for AGV fleet traffic management, the proposed solution is designed for real-world, non-standardized (i.e., non-grid-like) industrial settings characterized by narrow bidirectional corridors and high-traffic density, where AGVs of various sizes and capabilities operate simultaneously. Key contributions include an anytime conflict resolution strategy with adaptive time horizon regulation, an execution layer for safe and standard-compliant interaction with real AGVs, and an advanced mechanism for deadlock detection and resolution. Experimental results obtained in realistic industrial environments demonstrate higher throughput, with improvements of up to 11% over a conventional rule-based traffic management system, a state-of-the-art industrial method, and a priority-based L-MAPF variant, while maintaining continuous operation and improved efficiency.

Exception Related Code (ERC), which includes throw statements, conditions (if statements) that guard those throw statements, and try/catch blocks, is an essential component of software systems, allowing developers to detect and handle exceptional states that deviate from the expected program behavior. However, manually writing ERC across large codebases is tedious. We propose a novel task: retrofitting existing code with ERC. Namely, given code (without ERC) and Exceptional Behavior Tests (EBTs) (e.g., check if method throws InvalidArgumentException if null is given as the value to the argument) we aim to automatically generate missing ERC, such that the given tests pass. We design and implement Exception Coder (EXCODER) that performs context engineering to help Large Language Models (LLMs) tackle this task. EXCODER integrates static and dynamic program analysis with LLMs by providing the extracted contextual information to the LLMs. To evaluate EXCODER, we build a benchmark constructed from GitHub Java repositories, where we systematically remove ERC in 304 methods from 75 projects. Our results demonstrate that EXCODER provides an effective, though imperfect, solution to this problem in automated code generation, offering developers the first way to implement ERC following test-driven development. When combined with Qwen 2.5 Coder 32b, EXCODER achieves pass@1, 5, and 10 rates of 85.92% (12.56 percentage points over baseline), 86.18% (12.82 p.p. over baseline), and 86.51% (13.15 p.p. over baseline), respectively, on developer-written test suites. Our manual inspection of the generated code further reveals limitations of EXCODER, pointing to directions for future work.

Current audio-visual speech recognition (AVSR) benchmarks, like LRS3, rely heavily on clean, scripted and rehearsed speech. They fail to reflect the complexity of natural conversation, which involves overlapping speech, spontaneous turn-taking, unscripted vocabulary and variable acoustic conditions. To shift the field toward realistic dialogue, we introduce Candor-LR, a conversational benchmark derived from the CANDOR corpus of 1,656 natural dyadic videoconferences. Our custom data preparation pipeline yields 713.5, 10.1, and 60.1 hours of training, validation, and test data, respectively. Evaluating pretrained AVSR models on Candor-LR reveals that audio-only accuracy drops sharply compared to LRS3, but visual cues compensate effectively, driving much larger performance gains on Candor-LR than on LRS3. Furthermore, training on this corpus significantly improves cross-domain robustness under both clean and noisy conditions, as its realistic conversational data captures broader audio-video features. We open-source our pipeline to ensure reproducibility, establishing Candor-LR as a challenging benchmark for conversational AVSR.

Deformable convolution networks have recently become popular for many computer vision tasks, especially for semantic segmentation, because of their exceptional capabilities in dynamic spatial modeling. However, due to the dense deformable offsets and the lack of longer-range dependencies, they can not fully adopt proper and precise deformations for feature representations. To tackle the issues, in this paper, we propose Enhanced Deformable ConvNets (EDCN) for semantic segmentation. Specifically, a novel Enhanced Deformable Convolution (EDC) is exploited in the decoder, which integrates the Center-invariant Offset Module (COM) and Edge-aware Mask Module (EMM). The COM employs larger kernels and eliminates deformations at the kernel center, obtaining offsets that are more in line with the target from richer spatial information. Concurrently, the EMM obtains the significance of image content via Sobel edge detection, then selectively applies deformations based on the content significance, minimizing unnecessary deformations associated with relatively less important information, thereby avoiding impact from less informative regions. Experiments show that EDC outperforms state-of-the-art deformable convolution variants, including Deformable ConvNets V1-V4 and Entire Deformable ConvNets, across mainstream segmentation datasets with various decoder settings. Moreover, ablation studies confirm the effectiveness of each component. In addition, visualizations illustrate that EDC enhances spatial adaptation and target focus. We further analyze the extendibility of EDC to larger kernels on the image classification benchmark. Code will be publicly released.

We present the PACE, a framework for retrieval-augmented dialogue serving that formalizes Perceived Time-to-First-Response (PTFR) as a QoE objective and minimizes it under quality/cost constraints. Unlike prior work on cascaded routing, semantic caching, or adaptive retrieval, PACE jointly controls which answer source composes the response and what fills the waiting window. Deployed on a humanoid-robot sales service, it combines three mechanisms: a load-adaptive cascading router, a joint path-filler controller, and volatility-aware cache admission. On 75k CarQA requests, the cascade halves pure-LLM PTFR at P95 (0.29 vs 0.53s at c16). The adaptive controller reaches 0.41s P95, outperforming RAG by 2.4 times at high load with equal quality. The filler controller cuts calls by 94% with zero conflict. Volatility-aware admission reduces stale answers from 86% to 0%. A gating rule ensures the controller never worse than the baseline, with exposure bounded by one hold period. This is the first quantification of filler-answer conflict risk in deployed services.

While AVSR has achieved sub-1% word error rates on the standard LRS3 benchmark, its reliance on broadcast speech obscures whether this reflects true generalization or just domain adaptation. To investigate this gap, we evaluate three AVSR architectures across six conditions: controlled broadcast speech, fixed-grammar utterances, hyper-articulated Lombard speech, read speech from professional lipspeakers and non-professional speakers, and spontaneous multi-party video conversations. We find that visual-only performance deteriorates rapidly beyond broadcast domains, and audio-video fusion mainly benefits Lombard speech environments. Visual understanding degrades sharply at 90° profile views, with multimodal systems relying largely on acoustic fallback. Additionally, speaker articulation proves more critical than minor camera shifts, and LLM-based architectures suffer from poor out-of-domain generalization. Our work highlights a significant generalization gap in current AVSR research. To address this, we also introduce RoomReader-AV as a new benchmark for AVSR and release a unified data preprocessing pipeline to make comprehensive multi-condition evaluation accessible.

Simultaneous assessment of medical imaging and patient records is often required in clinical diagnosis. However, standard machine learning algorithms cannot analyze these data types together. Meanwhile, compliance with HIPAA and GDPR can constrain centralized aggregation of sensitive patient data. This leaves a crucial void of secure fusion of visual and textual context across distant networks. Thus, we present OmniMed-FL, a controlled systems study of multimodal federated learning for five-class clinical condition classification (Normal, Pneumonia, COVID-19, Pleural Effusion, Cardiomegaly). Our proxy corpus pairs 3,000 public chest radiographs with 3,000 class-conditioned synthetic notes, matched by class, not by patient. The framework benchmarks eight fusion strategies, three initializations, four missing-text imputation rules, and matched federated baselines under non-IID Dirichlet partitioning across 3 to 20 hospital clients. As all notes are synthetic and pairing is not patient-level, these are descriptive proxy comparisons, not estimates of diagnostic performance or deployment readiness. Within those limits with clients ($K=5$) and severe skew ($α=0.1$), local-only training achieves a macro-F1 score of 0.297, FedAvg achieves $0.662\pm0.074$, FedProx $0.737\pm0.085$, a matched FedMME-style one-shot ensemble $0.647\pm0.080$, and our SCAFFOLD-AdamW adaptation $0.070\pm0.015$, the 0.075 FedProx-FedAvg gap falling inside the wider of the two two-seed standard deviations. Over a $4\times3$ grid, label skew costs up to 0.27 F1 whereas a near-sevenfold client increase costs at most 0.10, while bidirectional volume grows linearly to 183.5 GiB at $K=20$. Multimodal fusion leads on both corpora, scoring 0.956 against 0.934 for text and 0.664 for images on the synthetic corpus and 0.906 against 0.880 and 0.737 on the radiograph corpus, for $2.3\times$ the model state of text alone.

Time-series foundation models are evaluated almost exclusively on public archives that predate them, so a strong score cannot be separated from having seen the test set during pretraining. The obvious remedy is a hold-out that postdates the models. We build one: thirteen forecasters -- four classical, three trained per dataset, six pretrained -- on seven groups drawn from five domains, every observation published after the last model was released, and every dataset rebuildable without an API key. Under this protocol pretrained models win 5 of 7 groups, lose one to a Theta baseline, and on daily exchange rates are indistinguishable from a seasonal naive forecast, along with every other method tested. We then ask what separates the wins from the losses, and report a negative result: the two intrinsic properties one would reach for -- seasonal strength and spectral entropy, measured on the input window -- do not account for the pattern, and seasonal strength is if anything negatively associated with the advantage. What does track it is corpus familiarity. Our largest gain (28% lower MASE than the best classical method, on weekly Wikipedia pageviews) falls on Wikipedia pageviews, the domain TimesFM's authors describe as the bulk of its pretraining corpus, at the same granularities and differing only in time window. Within the pretrained family, where every model forecasts identical series so that series difficulty cancels, the TimesFM family outranks the Chronos family by -0.53 ranks on Wikipedia against -0.09 everywhere else (1,500 vs. 754 series, Mann-Whitney p < 1e-5). We conclude that a temporal hold-out removes memorisation of a window but not familiarity with a domain, that benchmarks therefore need domain hold-outs stated relative to disclosed corpora, and that the practitioner's question is less which model is better than whether their domain is one the model was raised on.

Long-term change understanding from images of the same place revisited over time is a challenging task with applications in map maintenance and urban infrastructure monitoring. Prior work addresses it either through pixel-level prediction or difference captioning, neither of which is sufficient to reliably measure how well models detect and describe such changes. We introduce SPOT-THE-SHIFT, a human-verified benchmark for grounded image difference captioning of long-term changes in real-world driving scenes. Our benchmark provides natural language captions and spatial masks for structural changes across each image pair. We further propose an evaluation protocol that reliably assesses models' captioning ability, validated through human studies. Benchmarking state-of-the-art MLLMs, we find that models struggle with the fine-grained multi-image spatial capability required for this task. Finally, we develop a synthetic data generation pipeline that improves an off-the-shelf MLLM without sacrificing general capabilities.

Video understanding has rapidly evolved toward video large language models (VideoLLMs): systems that couple video representations with pretrained large language models and condition generation on a textual prompt. Their strong performance on captioning, question answering, retrieval and temporal grounding comes at a computation and memory cost that grows with frame count and context length, limiting deployment in real-time, mobile and resource-constrained settings. This survey covers inference-efficiency mechanisms for visual and audiovisual VideoLLMs that report concrete reductions in parameter count, FLOPs per input, latency, memory, or visual and audio token count. We analyze bottlenecks across frame sampling, modality encoding, connector-level token reduction, and LLM prefilling and decoding. We organize methods by the pipeline stage at which they act, covering VideoLLMs developed since late 2022 together with earlier frame-sampling and vision-encoder mechanisms that remain components of current pipelines. We assemble literature-reported accuracy--cost comparisons under shared host models and input protocols wherever available, distinguish them from heterogeneous cross-paper evidence, and identify gaps in audiovisual efficiency and standardized evaluation. We maintain a repository at https://github.com/momentslab/awesome-efficient-videollm.

Multimodal large language models (MLLMs) process hundreds or thousands of visual tokens per image, incurring prohibitive inference costs. While existing vision token pruning methods mitigate this overhead, they implicitly assume that a single fixed pruning strategy can be applied uniformly across all inputs. Our analysis further reveals that ranking pruning methods by average benchmark accuracy conceals substantial sample-wise complementarity: although the average-best strategy excels overall, alternative strategies prove superior on a significant fraction of individual samples. To harness this diversity, we propose VIP-Router, a lightweight VIsion Pruning Router that adaptively selects the pruning strategy predicted to be best suited to each input at a specified pruning level. Conditioned on low-cost visual and textual features, VIP-Router identifies the most suitable candidate strategy while retaining full-token inference as an option when pruning is predicted to be unfavorable. Evaluated on a curated suite of pruning-sensitive visual perception benchmarks, VTC-Bench Group A, VIP-Router consistently outperforms the best fixed strategy baseline across all reduction ratios, achieving a 26.9% relative improvement in average accuracy, and a 22.0% relative increase in average utility after accounting for realized token cost. Crucially, VIP-Router operates in a plug-and-play manner without modifying underlying pruning algorithms or model weights, introducing trainable parameters equivalent to merely 0.017\% of the backbone. Furthermore, VIP-Router proves effective across various MLLM backbones and yields consistent gains on unseen benchmarks, highlighting the potential of sample adaptive routing for visual token pruning.

Physics-based lithium-ion battery models provide access to physically meaningful internal electrochemical states and processes, but cell-specific parameter inference from terminal current-voltage data is computationally expensive and limited by identifiability. We present a surrogate-accelerated inverse framework based on a single-particle model with electrolyte dynamics (SPMe). Its forward map uses our Artiphy surrogate framework for rapid, differentiable evaluation of voltage and selected internal states. After rescaling to remove exact structural redundancies, we infer non-redundant transport, kinetic and capacity parameter groups, including concentration-dependent solid and electrolyte diffusivities. Synthetic voltage data from a Doyle-Fuller-Newman (DFN) model under a WLTP-like current protocol provide a benchmark with known reference parameters and controlled model discrepancy. The inferred SPMe reproduces the benchmark voltage with an error of order 1 mV and recovers electrode capacities well. Positive-electrode diffusivity is recovered accurately over much of the probed stoichiometric range. Local sensitivity and Fisher-information analysis identifies correlated kinetic-Ohmic and electrolyte-transport directions, and shows how localised information and the global diffusivity parameterisation can yield narrow Fisher-curvature envelopes despite weak voltage sensitivity to negative-electrode diffusion over much of the drive cycle. These results represent a step towards rapid physics-based in-silico parameterisation and reduced reliance on destructive cell characterisation.

Multimedia applications require precise music annotation-labeled positions, segments, or loops-placed by hand or algorithmically. Machine-learning algorithms are scalable and effective but need annotated training data, scarce for many tasks. TimeCues Studio is an open-source workspace where algorithm-development teams annotate a music corpus, compare detection algorithms against those annotations, and prototype new ones. Unlike existing tools built for a single track at a time, TimeCues targets teams annotating whole collections, tightly integrated with algorithm development. Annotators place several marker types-each supporting ambiguity-aware labeling-on a grid-locked timeline that visualizes many music features, including separated audio stems. The same timeline drives an algorithm-comparison engine with bundled baselines, a Python sandbox for prototyping new models, and an ambiguity-aware evaluator that honors the structured fields. The same visualization suits solo annotators on music-sync projects. TimeCues is MIT-licensed and deploys via one Docker Compose command.

Plane geometry remains a significant challenge in AI, requiring the integration of visual perception and mathematical reasoning. While Large Multimodal Models (LMMs) naturally handle visuo-linguistic inputs, they are often computationally intensive and opaque. We demonstrate that a pure Large Language Model (LLM), when equipped with specialized modules, can rival state-of-the-art LMMs on complex geometry problems. Our framework integrates a Geometric Vision Parser, which translates diagrams into symbolic form, with a Symbolic Solver that performs formal deductions, thereby mitigating hallucinations and promoting interpretable reasoning. To enable rigorous evaluation, we curate a benchmark of challenging problems from the 2025 Chinese Zhongkao examinations, ensuring data novelty and testing deeper deductive skills. Experiments demonstrate that our approach achieves performance comparable to Gemini 2.5 Pro while delivering clearer, human-like solutions.

This paper addresses the issue of supervised classification in the context of hyperspectral satellite images. It deals with two fundamental aspects: dimensionality reduction of data and the selection of appropriate supervised classification techniques. Firstly, we delve into dimensionality reduction, a critical step in simplifying the management of hyperspectral data. The reduction aims to decrease complexity in terms of memory and computing time. We examine two commonly used methods: Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA). Subsequently, we explore the selection of the most suitable supervised classification algorithms for hyperspectral images. We compare the performance of three methods: K-Nearest Neighbors (KNN), Support Vector Machines (SVM), and Random Forest (RF) using real hyperspectral data. The results highlight that the combination of PCA and RF yields the highest overall accuracy and Kappa coefficient.

Knowledge distillation offers an efficient route to transfer a task-adapted vision-language teacher to a compact student. The training target in current vision-language distillation methods is typically constructed from the teacher prediction and applied uniformly to all training samples, making it unreliable under class and domain shifts. In this paper, we argue that distillation target construction should be treated as a dynamic training decision rather than a fixed recipe. To this end, we propose OnPoKD, an on-policy distillation framework for vision-language model adaptation. To the best of our knowledge, OnPoKD is the first framework that applies on-policy distillation to vision-language model adaptation by learning target construction as a policy decision. OnPoKD learns a lightweight controller that constructs sample-wise adaptive targets using reliability and disagreement cues from the teacher model, student model, and zero-shot prior. Instead of relying on a fixed teacher prediction, the controller dynamically balances teacher supervision, zero-shot prior guidance, and hard-label anchoring through bounded policy actions, allowing the distillation target to adapt to varying sample reliability and training stages. The policy controller is updated with validation feedback, encouraging target construction to optimize transferability rather than merely fitting the training distribution. Since the controller is only used during training, OnPoKD can be seamlessly integrated into existing vision-language distillation pipelines while preserving the original inference architecture and test-time cost. Extensive experiments on Base-to-novel generalization and Cross-dataset transfer benchmarks show that OnPoKD consistently improves over strong vision-language distillation baselines.

Reinforcement learning with verifiable rewards (RLVR) has advanced language-model reasoning in domains such as mathematics and code, where objective answers are inexpensive to check. Diagnostic reasoning over complex data lacks this advantage: establishing the true cause of an anomaly often requires costly expert investigation and may remain ambiguous after the fact. We ask whether this asymmetry of verification can instead be engineered. We sample an intervention, inject it into a controlled simulator, and generate the observations it would produce. The hidden intervention provides an oracle label and objective reward, while the agent must still investigate noisy, confounded, and distributed evidence. We instantiate this approach in TRACE, a digital-advertising diagnostic environment with 12 root causes and fine-grained segment attribution. Agents investigate each episode using Python and SQL and must identify both the root cause and, when applicable, the affected segment assignment. On a held-out 235-episode test set, the strongest prompted baseline, Claude Opus 5, reaches 0.686 FullAttr@1. Supervised fine-tuning raises Qwen3.5-35B-A3B from 0.159 to 0.637, and subsequent RL with synthesized rewards reaches 0.757, outperforming all evaluated prompted baselines, including frontier closed-source models and a prompted Qwen3.5-122B-A10B model. The resulting policy also uses substantially fewer tool calls than the prompted 35B base. These results provide evidence that access to a scalable, objective training signal can be a more important constraint than model scale alone. More broadly, simulation-based verification can make otherwise ambiguous diagnostic reasoning tasks amenable to scalable reinforcement learning.

Manipulating deformable objects (DOs) is challenging due to their high-dimensional state space, underactuated dynamics, and partial observability. In this paper, we propose cRVAE, a lightweight conditional recurrent variational autoencoder that estimates the full DO state from only partial corner-node observations during inference. The resulting model is used as the forward model in a receding-horizon optimal control framework for obstacle-aware collaborative DO manipulation. In simulation on rope and fabric, cRVAE estimates the full DO state from the available corner-node measurements alone, matching the accuracy of a parameter-identified XPBD model. At inference it uses no physical parameters as model inputs and performs no online parameter identification. It also runs approximately 350 times faster on the rope and over 1500 times faster on the fabric per forward pass, keeping horizon-based planning within the 100 ms control budget where XPBD exceeds it already at short horizons. Full-shape estimation from corner sensing at in-loop speed is what makes the model deployable on hardware, which we demonstrate on a Unitree Go2 robot.

GUI agents accumulate high-resolution screenshots as the trajectory unfolds, increasing inference latency and memory usage. Training-free visual token pruning can reduce this cost, but cache reuse introduces a fundamental constraint. Once tokens are discarded, the corresponding visual evidence cannot be recovered without re-encoding. Pruning therefore becomes an \textit{irreversible admission decision} that must remain useful for unknown future targets while preserving coverage of operable regions under tight budgets. To address these challenges, we propose \textbf{\method{}}, a training-free framework for \emph{\textbf{T}rajectory-\textbf{r}obust \textbf{A}dmission and \textbf{C}overage-aware \textbf{E}vidence ordering}. Specifically, we combine a query-independent layout-derived interaction prior with instruction relevance and feature novelty to rank visual evidence according to both potential future utility and diversity. Then, we reserve part of the budget for native visual tokens distributed across the screen, repairing missing spatial coverage without breaking the ordering. Together, these mechanisms produce a nested token order, allowing retained visual evidence to shrink monotonically across budgets while remaining reusable throughout the trajectory. Finally, our monotone KV contraction incrementally contracts retired frames into compact session state, avoiding repeated visual encoding or pruning. Extensive experiments across six GUI benchmarks and diverse models verify the effectiveness of our proposed \method{} under tight budgets. The source code will be released.

In high-stakes domains such as legal practice, a language-model answer is only useful to the extent that a reader can verify each claim against the source the system cites. Current grounded-generation pipelines score the answer as a whole, so a correct conclusion can rest on fabricated or loosely matched citations and still score well. Closing this gap requires both a system built for per-claim verification and an evaluation that measures it. We introduce GANDR (Grounded ANswer DRafter), a two-agent system in which a Drafter writes an answer in a structured legal-reasoning format and a separate Critic, with the same view as a human verifier, audits each claim against its cited source and emits a per-claim audit trace on every round. We pair it with a strict correctness criterion requiring every citation to resolve to a passage the retriever returned. On a 185-item legal benchmark where all six systems share one backbone, one retrieval surface, and one citation instruction, GANDR ranks first on every primary metric, reaching 70.8% strict accuracy and leading the strongest baseline by 11.3 points (p<0.01). Reverting the protocol-anchored commit rule lowers strict accuracy by 22.7 points, and the strict lead stays positive on three further backbones, at +3.2 to +6.5 points. This lead traces to the Drafter configuration and the protocol-anchored commit, not to rewriting. Against two law-trained annotators the audit flags under-supported claims at F1 0.84 as a binary detector, while its four-way verdict labels agree only weakly and are advisory. Code is available upon request.

Data augmentation is fundamental to training modern deep vision and multimodal models. While individual methods, such as RandAug, CutMix, Mixup, RandErase, and DropPath, offer strong regularization effects, their combined use has saturated in performance due to overlapping functionalities, and aggressive pixel-level manipulations may disrupt delicate cross-modal alignment. This saturation motivates the search for a new augmentation axis within the embedding space rather than the input space. We introduce Aether, a simple plug-in method that applies diffusion-style random perturbations in the embedding space via controlled alpha-mixing, specifically designed to provide isotropic regularization that remains semantically consistent. Inspired by feature-space perturbations in language models and image degradation in generative pretraining, Aether induces mild yet effective perturbations that smooth the representations without compromising the fine-grained structural information required for strong vision-language encoders. Across diverse architectures and across multiple recognition tasks, Aether delivers consistent gains over the advanced recipe combining CutMix, Mixup, DropPath, and RandAug---a level of improvement rarely observed with modern augmentation alternatives. Notably, Aether demonstrates superior effectiveness in multi-modal alignment, succeeding where traditional pixel-space augmentations fail by providing a stable, isotropic regularization signal that respects the integrity of the high-dimensional feature space.

Humanoid locomotion on granular terrain remains a significant challenge due to its complex foot-terrain interaction dynamics that are difficult to model. Existing approaches either ignore granular contact dynamics or incorporate simplified normal force models with heuristic tangential components. In this work, we present a physics-grounded granular contact model based on three-dimensional resistive force theory (3D RFT) and efficiently simulate granular terrain for reinforcement learning (RL) training. Unlike traditional rigid contact models and simplified granular contact models with ad-hoc heuristics, our contact solver produces physically accurate granular intrusion dynamics without resorting to heuristics. It captures realistic penetration and tangential drag during training, enabling the policy to learn behaviors that transfer reliably to real-world granular terrain where rigid contact models fail. To adapt to varying terrain conditions, we train a terrain-adaptive locomotion controller via teacher-student RL, using a variational autoencoder to encode terrain information into a compact latent representation. Simulation studies using material point method (MPM) with NVIDIA Newton demonstrate that our method generalizes to unseen granular terrains, achieves a significantly higher success rate than baselines, and demonstrates zero-shot terrain identification and adaptation. We further validate our approach through extensive hardware experiments across diverse real-world granular terrains including basalt, dry sand, and beach sand. To the best of our knowledge, this is the first demonstration of agile humanoid locomotion on real-world granular terrain. Project page: https://humanoid-gm-locomotion.github.io/HUMANOID-GM/

Brachiation enables primates to move across overhead supports when ground paths are blocked, suggesting a complementary locomotion mode for robots operating in cluttered or hazardous environments. Bringing this capabil?ity to high-DoF humanoid robots is difficult because the controller must discover a long-horizon release-swing-capture sequence, coordinate alternating contacts with whole-body momentum, and act without reliable measurements of segment?relative displacement or hook-contact state. We present SwingBot, a learning framework for continuous humanoid brachiation with passive wrist hooks. Swing?Bot makes the task trainable by organizing learning around the structure of brachi?ation: biomimetic keyframes make rare release-swing-capture transitions reach?able during early exploration, and recurrent privileged-state estimation provides compact position and contact latents for deployment. Hardware experiments demonstrate continuous bar traversal and robustness to payload, external distur?bances and different bar spacings, showing that this formulation offers a practical route to whole-body robotic brachiation.

Open-loop multilegged locomotion over rough terrain has been interpreted as matter transport over a noisy channel: leg-ground interactions are discrete basic active contacts, terrain deletes or perturbs those contacts, and spatial redundancy concentrates the resulting thrust and arrival time. That construction is repetition-like because every module carries the same scalar locomotion task. It consequently provides neither a positive task rate nor a decoder that changes with the surviving contact set. Here we formulate locomotion instead as a quantized finite-frame expansion with erasures. A d-dimensional body-level command is mapped into N>d heterogeneous local contact commands. Rough terrain erases or corrupts frame coefficients, while a contact-gated compliant morphology physically realizes the weighted active-subframe decoder. For a linear-Gaussian model, mechanical equilibrium is exactly the posterior mean, tangent stiffness is posterior precision, and mechanical compliance is posterior covariance. Equal-norm Parseval frames are shown to be minimax optimal against one missing contact, two-contact robustness is governed by frame coherence, and a harmonic frame gives a directly realizable gait family. For independently surviving contacts of probability q, random Gaussian gait frames admit exact reconstruction at every analog dimension rate R<q, with a binomial reliability exponent, whereas recovery of arbitrary commands is impossible for R>q. Residual contact noise yields an asymptotic per-mode amplification 1/(q-R) and a vanishing mechanical stiffness margin at the threshold. An information-locomotion inequality and an exact incremental-redundancy rule direct the next gait component toward the softest task-relevant unresolved mode. The resulting analog frame-coding theorem establishes a finite relative redundancy and converse as part of a fundamental limit theory of legged locomotion.

LLM serving systems already reuse KV caches, but only when the reused text sits at the very start of the prompt. Two growing workloads break this condition: a retrieval-augmented generation server assembles a different set of retrieved chunks for every query, and a multi-agent coordinator reads reports written by other agents. Reused inside a new prompt, a cache carries the wrong positions and never attended to the other sources. The cache may also have been written by a different checkpoint of the same model family, which changes the stored values. Repair methods for such caches have appeared in three separate communities, each measured on its own terms, and existing benchmarks test only exact-prefix reuse, where nothing is lost. KVShareArena benchmarks KV-cache reuse across prompt contexts and model checkpoints on retrieved chunks and agent reports. It scores every method by the fraction of the gap it recovers between no cache and full recomputation, and charges compute, memory, and per-request latency with the cache in hand, reporting the one-time cost of building a cache separately. We find that correcting positions, which needs no recomputation, is enough until a question needs several sources at once. There, only methods that pay, by re-encoding part of the cache or by training, recover half to two thirds of the gap; unrepaired caches can be worse than no cache. Cache-compression methods that are harmless on a single prompt fall significantly behind position correction on freshly written agent reports. These patterns hold across three model boards. When a different checkpoint wrote the cache, training-free methods are barely affected, while an adapter trained on one checkpoint's caches loses quality. Harness, frozen querysets, and cost accounting ship as a pip package with an automated submission workflow and a public leaderboard.

Open-source large language models (LLMs) are increasingly competitive with closed-source models while offering transparency and the ability to run inference without exposing user inputs to a service provider. However, running large-scale models locally requires substantial computational resources. In practice, users may still resort to a third-party provider, giving rise to privacy and correctness concerns. Existing solutions that address these problems often impose substantial server overhead or introduce additional trust assumptions. In this paper, we present Maverick, a novel approach to private and verifiable LLM inference based on a protocol for delegating matrix-vector multiplication, a dominant operation in LLMs. At its core, Maverick provides, to our knowledge, the first information-theoretically sound verification protocol for matrix-vector multiplication delegation with transparent preprocessing, efficient (batch) verification, and virtually no server overhead. We combine this verification primitive with LPN-based pseudorandom masking to provide input privacy. We implement our matrix-vector delegation primitive and use it to build an end-to-end prototype of Maverick, which we evaluate on Qwen3-4B by measuring throughput in tokens per second. We evaluate client configurations with 1-8 threads. With one client thread and a CPU server using up to 128 threads, Maverick achieves throughput gains over local inference of up to 17x when privacy masks are generated online, 45x when they are precomputed, and 44x when only verification is required. With four client threads, the corresponding gains are 13x, 18x, and 17x. When server computation is no longer the bottleneck, client-side microbenchmarks with simulated network delay show speedups of 12x-20x, 34x-135x, and 38x-157x.

Persistent language agents need stored experience to remain available across time, while each answer requires evidence suited to a particular question. A superseded fact can mislead a current-state answer and still be essential for a historical query. We present RD-Forget, a training-free framework that separates what an agent stores from what it uses. A retained source archive preserves observations, and a query-conditioned memory view controls their influence on the current answer. A frozen language-model curator extracts relevant evidence, groups facts into semantic slots, and preserves the relations needed for multi-hop reasoning. Same-slot replacement links suppress superseded values in current-state contexts, while intent-aware retrieval makes earlier evidence eligible again. A rate-distortion formulation guides construction of the answer-time view within a memory budget. Experiments span conversational memory, knowledge updating, fact consolidation, long-context reasoning, and personalization under a shared answering pipeline. The results associate accurate answers with both query-relevant evidence construction and control over obsolete alternatives. Configurations without forgetting or query conditioning have the largest score deficits, while slot grouping, historical access, and relation preservation contribute complementary functions. Retaining history while selectively controlling its use offers a practical way to accommodate changing facts and future questions.

Large language models (LLMs) are increasingly deployed in globally used assistants, yet their default choices in culturally grounded everyday situations can systematically favour some cultures over others, affecting localisation, user trust, and equitable behaviour. Existing cultural benchmarks evaluate accuracy against a single "correct" answer, making it difficult to characterise an LLM's cultural preference prior when multiple culturally grounded responses are all valid; they also conflate default preferences with context-driven adaptation. We propose DiSCo, a distribution-first forced-choice evaluation framework that isolates default cultural priors and tests steerability via a four-level context gradient (C0--C3). Using DiSCo-Bench (304 items) derived from BLEnD spanning 12 cultures, we evaluate six diverse instruction-tuned LLMs. Default priors are heavily concentrated, with UK and US together absorbing approximately 35\% of all selections despite representing only 2 of 12 cultures. Most critically, prompt-based steering consistently widens the selection gap between high- and low-resource cultures, and injecting explicit cultural facts produces negligible distributional disruption, confirming that cultural preference bias cannot be resolved through prompt-based personalisation alone.

Traditional software delivery assumes a static paradigm: code is constructed prior to execution and deployed as a fixed artifact. We present Agentic Just-In-Time Software Construction (A-JIT), a paradigm that replaces static binaries with dynamic, software systems that can perpetually evolve to meet changing demands. In A-JIT, an application is an integrated assembly comprising code, a runtime harness, and an embedded AI agent that continuously observes system usage and live execution traces. Much like a traditional JIT compiler specializes machine code to runtime execution paths, A-JIT specializes software logic, workflows, and tool interfaces to meet the specific needs of the end-user. By integrating synthesis directly into the ambient application lifecycle, A-JIT enables applications to dynamically construct missing implementations, generate new capabilities on the fly, and continuously adapt to end-user behavior. We demonstrate how this model supports trace-driven human-AI co-construction and opens a new design space for adaptive, self-evolving software.

Embodied AI, including vision-language-action and world-action models, must operate reliably in the physical world. Yet methods that perform well in simulation can degrade substantially on real robots, especially in long-horizon deformable-object manipulation, where policies must track changing states and execute reliable multi-stage bimanual interactions. Existing real-robot benchmarks mainly focus on short-horizon rigid-object tasks and offer limited coverage of long-horizon deformable manipulation. We introduce FolDeX, a physical-world benchmark built entirely from real-robot data, with garment folding as its primary task. Since real-robot data collection is costly, FolDeX studies how heterogeneous physical experience can be reused efficiently. The benchmark is organized around four research axes: leveraging human intervention and recovery data collected during deployment; transferring data across tasks, including across garment categories and from rigid to deformable-object manipulation; reusing data across scenes with changes in lighting, background, and layout; and transferring data across robotic embodiments. FolDeX provides 2,000+ hours of real-robot data spanning 20+ tasks and 10+ embodiments. We also establish a fair real-robot evaluation platform for externally submitted policies, with standardized tasks, held-out physical objects, controlled initializations, and a unified execution protocol. The platform is publicly accessible at https://ai.midea.com/#/fold-challenge. We hope FolDeX will serve as a unified testbed for heterogeneous real-robot data reuse and reliable long-horizon deformable manipulation.

Graph-based retrieval can improve multi-hop question answering, but existing approaches often incur high query-time costs and produce diffuse, oversized contexts that reduce generation efficiency. We present LiteRAG, a graph-based retrieval method that replaces expensive retrieval-time LLM control with query-conditioned algorithmic exploration and reasoning-chain context construction. On DistComp, a benchmark for multi-hop retrieval over distributed-systems papers, LiteRAG attains the highest overall quality among the evaluated methods (0.798) while reducing per-query latency by over 100$\times$ and cost by over 99% relative to GraphRAG Global and DRIFT. On UltraDomain, it matches LinearRAG on overall quality while using about 14$\times$ fewer tokens. An ablation study indicates that LiteRAG's query-adaptive thresholding and community-aware hub penalization are the main drivers of its token-efficiency gains.

A graph retrieval-augmented generation pipeline chooses which triples to put in the prompt, a syntax to write them in, an order to write them in, and a sentence telling the model what to do with them. We vary all four over six large language models and two knowledge-graph question answering benchmarks. Two of the four choices move the answer and the other two are flat. The first is whether the answer path, the triples needed to reach the answer, is in the prompt at all. Holding the number of triples fixed and replacing every triple that is not on the chain with material from an unrelated entity changes answer accuracy by +0.003 F1, while removing the chain costs most of what the graph was worth. Retrieval budget belongs on recall, and precision in the range we can test buys nothing. There is no retriever here: subgraphs come from gold SPARQL, so precision describes the context we build, not a system setting. The second is the grounding instruction. With no facts in the prompt, telling a model to answer using only the provided facts drops F1 from 0.299 to 0.035, a factor of 8.63. That figure describes an evaluation with an empty context arm rather than a working pipeline, and an experiment that applies the instruction to its context arm but not to its no-context baseline manufactures a spurious finding that graph context hurts at depth. We found one in our own results and retract it. Syntax, triple order and subgraph size produce no effect we can measure at multi-hop depth. The comparison that would price the grounding instruction against correct context is not measurable with a format-sensitive scorer, because the instruction determines the response format; we report it as an open contrast rather than a number.

Large language models (LLMs) have demonstrated remarkable capabilities in reasoning and code generation, raising the prospect that they could assist in developing and optimizing the very infrastructure that powers them. However, existing benchmarks mainly focus on isolated kernels, predefined operators, or pre-specified optimization targets, and therefore fail to evaluate the ability of LLMs to perform open-ended, long-horizon LLM infrastructure engineering. To address this gap, we present $Φ$-Bench, a benchmark for systematically evaluating LLMs on engineering the LLM infrastructure stack. Derived from optimization problems studied in frontier research and grounded in real-world code repositories, $Φ$-Bench provides broad coverage of the LLM infrastructure stack and spans tasks of varying complexity, ranging from localized kernel-level function completion to long-horizon implementation and end-to-end system optimization. Extensive experiments on frontier LLMs reveal their current capabilities and limitations in engineering complex LLM infrastructure, offering insights into the challenges that remain on the path toward autonomous optimization of future AI infrastructure.

Feature transformation improves predictive performance on tabular data by constructing informative abstractions from raw features. Recent generative approaches encode transformation knowledge into continuous embedding spaces for efficient exploration of candidate strategies, but face three key limitations: (1) overlooking hierarchical relationships between low-level features, operations, and high-level abstractions; (2) enforcing order-sensitive embeddings on inherently permutation-invariant transformation sequences, thereby introducing systematic bias; and (3) relying on gradient-based search, which is ill-suited to non-convex transformation spaces. We propose a framework with two complementary components. First, a permutation-invariant hierarchical module captures interactions across features, operations, and abstraction levels, with a self-attention pooling mechanism that maps semantically equivalent structures to consistent embeddings aligned with downstream performance. Second, a policy-guided multi-objective reinforcement learning strategy initializes the search from empirically strong seeds and jointly optimizes predictive accuracy and transformation efficiency. Extensive experiments on diverse tabular benchmarks demonstrate the effectiveness and robustness of our framework against strong baselines. Our code and data are publicly available at: https://github.com/RayLiu1103/PHER.

Reinforcement learning over a frozen reasoner has become a common recipe for teaching a policy which external tools to invoke. We show that this recipe becomes structurally mismatched in specialist scientific settings where the complete tool-subset space is enumerable. There, a small set of recurring computational capabilities covers the domain, so the space of tool subsets is combinatorial yet small enough to enumerate, and GRPO still estimates an action expectation from a handful of sampled rollouts. Worse, the approximation degrades as training succeeds: as the policy concentrates on preferred subsets it resamples them, sampled rewards collide, and the group-normalized advantage vanishes. On genomic reasoning the fraction of questions yielding no reward signal rises from 0.2% under a uniform reference policy to 20.8% after GRPO training. As a remedy, we introduce FGPO (Full-Group Policy Optimization), which (1) scores every tool subset and optimizes the exact action expectation, so each update sees the complete action space, and (2) precomputes the reward of each question--subset pair into an exhaustive table, removing frozen-reasoner calls from the training loop entirely. Across five frozen reasoners and three genomic benchmarks, FGPO outperforms GRPO in all 15 settings by 6.75 points on average and up to 14.20, while a standard on-demand GRPO schedule would require 2.4 times as many frozen-reasoner reward evaluations and, on GenomeQA, FGPO cuts invoked tools per question from 2.36 to 1.40.

This work considers adaptive shared human-robot control for nonlinear control-affine systems, where the assumption of a fully rational human is relaxed and the robot adapts its assistance to observed boundedly rational human behavior. We use a level-k bounded-rationality model of the two-player game to construct a finite bank of candidate human and robot policies through alternating best-response computations, with the associated value functions and policies approximated using adaptive dynamic programming. During the shared-control interaction, state-transition residuals compare the measured system evolution with the trajectories predicted by the candidate human policies. The residuals are accumulated using a forgetting factor and mapped to a probabilistic human-behavior model over the finite candidate bank. Rather than selecting a single candidate or averaging stored robot policies, the robot computes a distribution-aware one-step best response by minimizing an expected cooperative cost over the complete estimated human behavior distribution. For a quadratic terminal-value approximation and Euler state propagation, this response admits a closed-form solution expressed in terms of the expected human input. The proposed methods are evaluated in simulations of a benchmark nonlinear system stabilization task, and of a planar manipulator shared control setup. The reported results show decreasing Kullback-Leibler divergence between the estimated and simulated human behavior distributions, and a lower accumulated running cost for the robot agent over the shared control interaction period, than the maximum-probability and probability-weighted alternative policies baseline.

Attias, Hanneke and Ramaswami (NeurIPS 2025) asked whether randomization provably reduces the oracle calls needed for online learning when the class is accessible only through an oracle. We study the instance they singled out: transductive online learning of thresholds on an unknown total order of T instances, with a consistency-type ERM oracle that returns a full concept consistent with a queried labeled set (or reports non-realizability). Our main result is a separation for a fixed natural oracle. When the oracle is the minimal-prefix rule (or the maximal-prefix rule), every deterministic learner makes M mistakes and Q calls with $M+Q\ge T-\varepsilon$ on some instance ($\varepsilon\in\{0,1\}$, according to whether the empty prefix is a concept), and the constant is exact; hence $O(\log T)$ mistakes cost $T-\varepsilon-O(\log T)$ calls, whereas that paper's randomized learner achieves $O(\log T)$ expected calls and mistakes under the same rule. The randomized order is optimal: on an explicit hard distribution under the minimal-prefix rule, every learner has expected mistakes at least $((T+1-\varepsilon)\,128^{-\mathbb{E}[Q]}-1)/2$, so $Ω(\log T)$ expected calls are necessary for polylogarithmic mistakes. The separation is governed by the oracle's selection rule, not by the class alone: for a legal feasible-median ERM rule a deterministic learner achieves $O(\log T)$ calls and mistakes, while a global-median rule again forces linear total cost. The same linear bound holds when the oracle's answers are chosen adversarially and then frozen into a memoryless oracle. We add partial tradeoff results for fixed query budgets (the middle regime is open) and an interface contrast: with only a weak consistency oracle, returning a realizability bit, both deterministic and randomized learners need $Θ(T)$ calls.

In this work, we introduce language-aligned motion representations for domain-generalizable UPDRS-Gait severity estimation, aiming to learn semantically structured motion features that generalize across heterogeneous clinical domains. We first learn motion representations using a Bi-GRU backbone that captures the temporal dynamics of SMPL sequences. Prior to model training, motion captions are generated offline using Qwen2.5-7B-Instruct. The backbone is then trained with both classification and text-alignment objectives to learn discriminative and semantically structured motion representations while accounting for the class imbalance present in the training data. We subsequently adapt the learned backbone independently to each source domain so that the model can capture domain-specific motion characteristics. The resulting source-specific models are then merged at the parameter level to consolidate complementary knowledge across source domains into a single domain-generalized model. To further mitigate class imbalance, we perform GPT-5.5-based pseudo labeling, and our final merged models for each site do not use any class-prior correction during inference. The resulting model is evaluated under the unseen-site setting of the MoCha Challenge, using Macro F1 as the primary evaluation metric. Our method achieves a macro-F1 of 0.57 on the hidden test set with only 637K active parameters at inference, ranking 3rd among 58 leaderboard entries in the MoCha 2026 Challenge. The challenge attracted 1,669 submissions from 112 participants and offered monetary prizes sponsored by Machine Medicine Technologies.

Predicting how long a battery energy storage system will last is critical for warranty design, maintenance planning and investment decisions, yet degradation models are mostly deterministic and rarely validated against real field data. We apply an open-source probabilistic degradation framework, combined with a cell-to-system approximation, to bridge the gap between cell-level laboratory aging models and system-level field predictions for residential battery energy storage systems with quantified uncertainty. The framework predicts cell-level state-of-health to within 0.4 % mean absolute error, roughly half the error of prior models for this dataset. When applied to field operation data, the framework's predictions are consistent with all three available system-level capacity measurements - a benchmark rarely available for open probabilistic degradation models. Cell-level heterogeneity is approximated by two bounding stress scenarios differing only slightly (a 5 % spread in temperature and a 9 % spread in current). The mean degradation trajectories of the two scenarios reach end-of-life 10 months apart, while the full predicted end-of-life range across both scenarios spans approximately three years, about a third of the expected system lifetime. We link this uncertainty to two drivers. One is a systematic mismatch between laboratory test conditions and field-representative operating stress. The other is variability in the training data itself. These insights translate into concrete, resource-efficient recommendations for future aging study design, supporting more confident predictions of battery lifetime under real-world conditions.

AI agents are increasingly involved in network automation, where they can initiate configuration changes through mediated operational interfaces and assess the resulting state. Nonetheless, operational networks usually span many devices and administrative domains. Realizing an operator's intent requires coordinating agents with distinct authority scopes that define the resources they can access, the operations they can invoke, and the network state they can observe. This division limits the blast radius of an erroneous action but fragments the evidence needed to assess the network-wide outcome. Successful execution of a configuration action proposed by one agent does not establish that remote devices responded as intended or that routing changes reached the required devices. A valid observation may also become stale after a subsequent change. Before the coordinated operation can be declared complete, a trusted assurance layer must collect current observations from the required scopes and determine whether they collectively support the operator's intended network-wide outcome. To address the completion admission problem, we present EvidenceNet, a runtime assurance layer for deciding whether coordinated agent operations have achieved an operator's network intent. Its broker collects the post-change observations required by a completion contract, and its admission gate checks that the evidence comes from the required scopes, remains current, and satisfies the task rules. A verifier agent provides an additional assessment of the observation content. Experiments on live routing networks show that post-change state checks recognize successful outcomes that configuration-action records alone cannot establish. Controlled interventions further show that EvidenceNet rejects completion when otherwise satisfactory observations have the wrong source, have been substituted, or are stale.

In-context learning (ICL) is widely used in multimodal large language models (MLLMs) and achieves strong performance across a wide range of multimodal tasks. However, existing multimodal ICL methods often rely on surface level imitation of in-context demonstrations, making it difficult for MLLMs to align their responses with the reasoning path required by the given multimodal input. This limitation becomes more pronounced in complex multimodal tasks, thereby restricting further improvements in MLLM performance. To address this issue, we propose a new multimodal ICL framework that combines contrastive demonstration modeling with the self-refinement capability of MLLMs. Specifically, our framework reformulates each demonstration by explicitly contrasting a suboptimal response with a better response under the same input, together with a reasoning path that reveals how the response should be refined. This contrastive formulation makes the reasoning path toward the desired response more explicit and guides the MLLM beyond superficial imitation. Furthermore, because effective refinement depends on the current response, we introduce a response-conditioned retrieval mechanism to select demonstrations whose reasoning paths are more relevant to the current response. In addition, we use a lightweight alignment controller to predict response quality and determine whether further refinement is needed. Experiments on three types of multimodal tasks show that the proposed framework consistently improves MLLM performance, with particularly notable gains on visual question answering (VQA).

This paper outlines the specifications and design approach used to construct a full body imaging scanner capable of capturing skin lesions at a dermatoscopic level using cameras mounted on the end-effectors of four UR10 manipulators. The system possesses a view-planning algorithm capable of appropriately selecting the best camera position to acquire images of moles, a high-level controller to allow the manipulators to work simultaneously and a collision-detector that halts the manipulators when they make contact with an object or a person. We evaluate the system through real-patient full-body scans, comparing acquired images against contact dermoscopy and an existing total-body photography system (Vectra) across clinically relevant lesion features, and quantify true optical resolving power using a USAF 1951 resolution target, yielding a smallest resolvable feature size of 22.1 microns for our scanner compared to 8.8 microns for contact dermoscopy. Results show the scanner consistently outperforms Vectra across most clinically relevant features and achieves comparable performance to contact dermoscopy for the majority of features assessed. By acquiring dermatoscopic-quality images automatically and without contact, and without requiring a separate manual dermoscopic examination, the scanner closes part of the gap between total-body photography and handheld dermoscopy, suggesting potential for future integration into screening workflows.

UAV target following in cluttered environments is inherently predictive: current-state followers can lag behind turns, choose blocked corridors, or trade tracking for unsafe near-horizon motion. We propose a future-aware flow planning framework for state-informed UAV target following. Predicted target futures guide clean UAV trajectory generation as horizon-aligned residual signals, while risk-scored executable-prefix repair is embedded inside the sampling loop. On fixed ID/OOD receding-horizon benchmarks, the planner improves the intended safety--tracking trade-off rather than dominating every metric: it matches zero measured ID collision rate with the highest ID safe-tracking time, and gives the lowest OOD macro collision rate and final tracking error among the displayed methods, while Future-MPC remains smoother and stronger on some thresholded OOD success metrics under its hand-designed objective. Ablations show that future adaptation improves candidate generation before safety repair, and simulator-facing stress tests probe interface, sensing, and controller-execution effects. These results support horizon-aligned future adaptation and embedded prefix repair as complementary ingredients for safe UAV target following under the tested simulation conditions.

Generalized Category Discovery (GCD) assigns unlabeled instances, mixed with labeled data, to known or novel categories, requiring human-like compositional reasoning: reusing primitives learned from known classes and deciding when new combinations imply new categories. Existing GCD methods operate on unstructured token features and struggle to extrapolate to novel compositions. We propose CoGe-GCD, which rethinks GCD through compositional generalization with two coupled stages. (i) Compositional Perception structures patch tokens by mapping them to a small vocabulary of primitives and refining token embeddings via competitive token-primitive assignment and information passing, yielding coherent groups for discovery. (ii) Generalizing Induction exploits the induced geometric structure and applies a structure-preserving calibration over spatial relations, maintaining probabilistic semantics while improving extrapolation to unseen primitive combinations. CoGe-GCD is implemented as an inductive-bias module between backbone and projection head, without modifying heads or losses, and can be plugged into diverse GCD frameworks. On standard benchmarks, it consistently improves all-class accuracy, unknown-class number estimation, and geometric quality, with marginal computational overhead. Code is available at https://github.com/lytang63/CoGe-GCD.

We approach a cognitive simulation perspective on episodic and semantic memory in multiple-choice question answering by incorporating text from individual text corpora (ITC) into retrieval-augmented generation and DoRA fine-tuning. We web-crawl the search histories of 515 participants who answered 36 multiple-choice knowledge items and analyze a stratified subsample of 150 participants. For each participant, one DoRA adapter consolidates their ITC into a small language model (SLM) whose baseline correctness falls below the participants' lowest quartile. The adapter measurably writes the ITC into the weights: it fits its own participant's held-out text better than other participants' texts (dz =1.27), an individuality effect that increases with ITC size in rank order. On the generalized knowledge test, however, the adapter adds knowledge rather than alignment with the individual: log-loss match improves, whereas match accuracy under a bias-corrected PMI readout does not, and retrieval adds nothing on top. Our results demonstrate that ITCs can be consolidated into the weights of SLMs, an encouraging basis for individualized tutoring agents, and we discuss how to move from there toward a realistic simulation of episodic and semantic memory at the individual level.

Arabic morphology remains challenging for large language models, since fluent generation does not guarantee accurate morphosyntactic control. Existing Arabic evaluations mainly target downstream tasks and do not directly test controlled morphological generation from explicit lexical and feature-based input. We introduce YallaMorph, a large-scale benchmark for Arabic morphological generation covering verbs, nouns, adjectives, their cliticized forms, and invalid configurations. We evaluate multilingual and Arabic-oriented LLMs under diacritized and undiacritized settings over 600K benchmark entries. Results show that Arabic morphological generation remains difficult, especially for cliticized, unseen, and morphologically rare forms.

AI systems become more useful when they can adapt to the people using them, but in multi-agent systems, useful context learned by one agent often remains unavailable to others. We present kernel-managed shared memory, a system-level abstraction in which specialized agents write structured, tagged memories while the agent-system kernel, not individual agents, governs retrieval, privacy enforcement, and prompt injection. We implement and evaluate this design on AIOS and compare it against three alternatives across three assistant models (GPT-4o, Llama-3.1:8B, Qwen-2.5:7B) and 1,800 total trials. Against an unmanaged external memory backend (Mem0) using identical underlying storage, kernel-managed retrieval and injection improve personalization scores by 2.4-4.0 points on a 5-point scale (e.g., 1.05 to 4.69 profile usage on GPT-4o), with every comparison significant at p < 10^-18. Against standard retrieval-augmented injection, gains are similarly large and consistent across all three models. Against full, unfiltered context concatenation, a soft ceiling on available context rather than on response quality, kernel-managed injection statistically matches performance on two of three models and shows a small, model-specific deficit on the third, while using substantially shorter prompts: end-to-end latency is 15-61% lower across all three models, with corresponding reductions in per-call token usage and inference cost. These results indicate that centralizing memory management in the agent-system kernel, rather than leaving retrieval and privacy enforcement to individual agents, delivers most of the personalization benefit of unconstrained context at a fraction of its cost.

How many past requests are needed to decide which qubits should share entanglement? We show that the answer depends on the allocation choices created by the queries: a larger memory can require no more data. The memory stores a classical bit and answers requests through a fixed detector that preserves coherence within each measured sector. For independent commuting $X$- and $Z$-type Pauli queries, we characterize the full attainable prediction-contrast region and construct encodings that preserve the bit at every nonzero vertex. With sharp reports, a $d$-qubit path and groups of at most $k$ qubits have minimax excess error after $m$ requests proportional to $k^{-1}\min\{1,\sqrt{d\log(k+1)/m}\}$, uniformly for $2\leq k<d$. Connected biclique regions can grow without increasing sample demand when depth, region count and connections per region stay bounded. Preparation noise introduces a separate calibration requirement. We derive an exact tradeoff with extra fresh detector calls and transfer the learning law to structured transaction co-location. Population-risk experiments test the statistical predictions. We also compare encodings on a native 15-qubit device and learned partitions on public purchase baskets. The full chain wins on the device; frequency grouping outperforms basket search in the largest-capacity retail setting.

Driver gaze provides information regarding driver visual attention and situational awareness to the surrounding traffic. Existing driver gaze estimation studies represent gaze in terms of gaze zone or gaze vector/point-of-gaze (PoG). However, object-level gaze information provides a more semantically meaningful representation of visual attention by identifying attended objects, such as vehicles, pedestrians, or traffic signals. In this study, we propose an end-to-end driver gaze object prediction framework, TransGaze-Object, Transformer-based Gaze Object prediction model. The proposed framework first extracts facial features, including face and iris-weighted eye features, along with trafficobject spatial features. A transformer based cross-attention mechanism is then used to compute similarity scores and attention weights for predicting the drivers gaze object. To train this model, we propose a benchmark driver gaze dataset, Urban Driving-Face Scene Gaze (UD-FSG), comprising synchronized driver-face and traffic-scene images, scene objects bounding boxes, and gaze labels in terms of 2D gaze coordinate and gaze object. The TransGaze-Object model achieves an overall accuracy of 60% for gaze-object prediction, compared to 51% accuracy obtained from associating the estimated Point-of-Gaze to traffic objects. The error analysis reveals that TransGaze-Object reduces confusion between traffic objects (predicted) and the background (ground-truth), achieving an error rate of 11.68%, a 49.7% relative reduction compared with 23.21% error obtained from PoG-based gaze-object association. Overall, the results demonstrate the effectiveness of directly predicting gaze objects from driver-face and traffic-scene information, rather than estimating an intermediate Point-of-Gaze and subsequently associating it with traffic objects.

A hallmark of human dexterity is the cooperative use of fingers, where different fingers take on distinct yet coordinated roles to accomplish fine manipu- lation, such as capping a pen with the hand that holds it. We study this finger-level coordination through in-hand assembly: mating two rigid objects within a single dexterous hand, with no second arm and no fixture. We present a reinforcement learning formulation to solve this problem in a unified framework, which is driven by a goal relative pose between the two parts. Finger coordination is shaped by a function-based auxiliary reward and regularized toward a single human reference pose, while domain randomization and a fusion of historical proprioception and object observation confer robustness to occlusion-induced estimation noise. The same recipe solves three different assembly tasks (Bottle, Syringe, and Marker). Trained purely in simulation, the policies transfer zero-shot to hardware with a single camera, demonstrating robustness to state-estimation errors caused by oc- clusion. Our experiments also reveal that in-hand assembly places demands on hand morphology and can serve as a benchmark for modern robotic hand systems. Videos and code are available at https://ltbgbird.github.io/in-hand-assembly-page/.

Financial text, textbooks, and question-answer pairs are abundant, but only a small fraction is directly usable for reasoning-focused post-training. Existing QA pairs often lack explicit reasoning, sufficient context, or reliably verifiable answers, while textbooks must first be transformed into synthetic training examples. We present a data-centric pipeline that constructs complementary corpora by mining open-source reasoning traces, distilling financial instruction data, and generating knowledge-graph-guided question-answer pairs from financial educational material. After semantic deduplication, three lightweight sequence classifiers select finance-relevant examples, reject under-specified questions, and identify tasks suitable for reinforcement learning with compact rule-based verifiers. For model adaptation, we study supervised fine-tuning and reinforcement learning, while self-distilled fine-tuning and post-training model merging are used to prevent the loss of financial capabilities already present in the starting model. We evaluate the adapted language models using FINESSE-Bench, reporting aggregate performance and changes relative to their starting checkpoints. Across the selected comparisons, ordinary SFT reduces FINESSE-Bench accuracy by 3.2-4.0 percentage points, whereas self-distilled SFT improves over the corresponding starting models by 1.0-2.8 points. Equal-weight merging recovers 3.0 points over its SFT parent and finishes 0.9 points above the original model; GRPO on hard tasks adds 0.4 points after self-distilled SFT or 3.0 points when applied directly to verifiable tasks. These results show that retention-aware adaptation can improve financial reasoning without the regressions observed after ordinary SFT.

Compute governance today is a governance of training: the thresholds, reporting requirements, and frontier-AI regimes now in force attach to training compute and treat the trained model as the regulatory unit. That picture is incomplete: capability increasingly migrates to the deployment stage through inference-time scaling, agentic scaffolding, and compression onto consumer hardware. This paper asks which mechanisms are available once the regulatory object shifts from the training run to the inference call. We develop a feasibility taxonomy of twenty inference-time mechanisms across monitoring, verification, and enforcement, each rated on a four-point readiness scale against a documented four-vendor evidence base. We then stress the taxonomy against a two-dimensional adversary model (three capability tiers crossed with four adversary roles) and map each mechanism to four governance scenarios (domestic regulation, bilateral or multilateral coordination, industry self-regulation, and compute-marketplace governance). Fifteen of the twenty mechanisms have commercial technical substrates in production today, although governance-grade assurance and adversarial robustness vary substantially. The adversary analysis shows that this readiness holds only against a cooperative deployer and a low-to-medium-capability user: no mechanism rates adequate against a high-capability state-level deployer, and fine-tuning removes the model-internal components of the enforcement cluster, although platform-external controls can persist. A substitution analysis connects the taxonomy to a companion hardware paper as a conditional substitution principle describing when inference-stage and hardware-stage mechanisms provide comparable regulatory coverage under stated conditions. A second-rater reliability check on a random subset of the readiness ratings returned a quadratic-weighted Cohen's kappa of 0.74.

Molecule generation has emerged as a powerful computational tool for de novo drug design, enabling the exploration of chemical space beyond the limits of conventional virtual screening. The field has progressed rapidly, driven by advances in molecular representations, generative architectures, and target-aware modeling strategies. However, existing reviews typically address specific model families or application scenarios in isolation, rather than offering an integrated perspective on how these components collectively form a coherent generation workflow. In this review, we present a comprehensive evaluation of molecule generation models for de novo drug design, covering 82 methods across five deep generative frameworks, including recurrent neural network (RNN)- and Transformer-based models, variational autoencoders (VAEs), generative adversarial networks (GANs), flow-based models, and diffusion models. We first summarize widely used benchmarks and molecular representations, and then examine the methodological principles underlying both general and pocket-conditioned generation. A central contribution of this work is a systematic synthesis and comparative analysis of reported performance across commonly used benchmarks and evaluation metrics. We also summarize representative experimentally validated case studies. Looking ahead, we discuss future directions in standardized 3D data, interaction-aware generation, receptor flexibility, and multi-objective molecular design, with the aim of improving the reliability and experimental relevance of molecule generation. All collected benchmark resources, evaluation metrics, and model references are provided in a publicly accessible repository at https://github.com/JacklinGroup/molecule-generation-review.

Large language models (LLMs) increasingly act as research agents, yet their ability to track shifts in research attention is difficult to evaluate because reviews and research ideas lack uniquely verifiable outcomes. We introduce Research Attention Prediction (RAP), a rolling benchmark covering 278 AI/ML fields and 1,390 episodes. At each cut-off, an LLM agent searches a temporally restricted arXiv corpus and predicts the next six months' paper shares across eight frozen research directions. Search generally helps, but all four diagnostic models perform worse than an exact-count exponentially weighted moving average (EWMA) baseline in compositional accuracy. We identify two linked bottlenecks. Under cumulative-history access, State carry-forward outperforms direct Forecast for all four diagnostic models; frozen-evidence replay links a shared component of this reversal to Forecast-oriented policies retrieving a smaller share of recent evidence. Even with exact historical activity, future-specific updating remains limited, with only GPT-5.5 plus reopened Search slightly surpassing EWMA. Fine-tuning on realised outcomes improves Qwen3-4B's forecast Spearman correlation by 0.105 on held-out fields at later origins, with gains also on change-rich episodes.

Existing bias auditing methods typically rely on model outputs, requiring costly benchmarks or judge models and potentially missing internal shifts that never appear in generated text. We propose a reference-based method that audits bias in hidden-state representations across related model variants, for example before and after fine-tuning. Because fine-tuning reshapes representation geometry, absolute hidden states are not directly comparable, so we encode each sentence by its similarities to a fixed set of anchor sentences, yielding relative representations in a shared comparison space. There we measure how target groups shift in their association with positive and negative attributes, a quantity we call the Representational Bias Shift $ΔB$. Across three model families and the WildGuardMix, DecodingTrust and ToxiGen benchmarks, $ΔB$ correlates with output-level bias change in 15 of the 18 settings we test, reaching $|r| = 0.84$ ($p < 0.001$) under full fine-tuning and becoming more model-dependent under parameter-efficient adaptation. Thresholding $ΔB$ detects checkpoints whose bias increased with ROC AUC between $0.65$ and $0.99$, and on WildGuardMix and DecodingTrust it separates them better than a SEAT-based baseline for all three families. $ΔB$ is also stable under changes to the anchor set, attribute sets and target templates. Our method requires no task-specific evaluation data and audits a model in about three minutes, using $3$-$50\times$ less compute than the output-level benchmarks considered here. We view it as complementary to output-based auditing rather than a replacement for it.

Biomedical ontology normalization maps free-text expressions to standardized concepts, enabling consistent integration and analysis of biomedical data. This task remains challenging because lexical variation and subtle distinctions among hierarchically related concepts can obscure concept boundaries. We present OntologyAligner, a three-stage framework that combines ontology-aligned retrieval, large language model candidate reranking, and selective hierarchy-guided refinement. We also construct PhenoNormBench, a unified benchmark comprising 13,390 samples from seven Human Phenotype Ontology datasets. OntologyAligner achieved state-of-the-art performance on HPO normalization, with 88.78% Macro Top-1 Accuracy and 86.75% Micro Top-1 Accuracy, exceeding the strongest baseline by 4.85 and 5.07 percentage points, respectively. Ablation analyses showed complementary contributions from all three stages, and sensitivity analyses demonstrated stability across candidate-set sizes and model backbones. Applications to MONDO, MEDIC, and NCBITaxon further established portability to other ontologies. OntologyAligner offers a generalizable framework for accurate mapping of biomedical text to structured ontology concepts. PhenoNormBench and the code are publicly available at https://github.com/zhelishisongjie/OntologyAligner.

LLM agents for sequential decision tasks are often post-trained with trajectory-level outcome labels, but such labels provide little supervision for preserving multiple successful branches from the same decision state. We study this problem as successful strategy coverage: how broadly a model realizes distinct successful strategies under a fixed rollout budget. We present Direct Diversity Optimization (DDO), an offline post-training method that combines Divergence-Tree Collection (DTC) with the Reference-Relative Target-Odds Objective (RTO). DTC constructs state-aligned branch sets rooted at shared decision states, and RTO trains the model to match reference-relative targets over successful alternatives. DDO achieves the strongest task success and successful strategy coverage among the compared post-training methods across BabyAI, BabaIsAI, and WebShop. It also achieves the highest recovery rate after local action replacement and higher task success and coverage than successful-only imitation and decoding-time diversification controls.

Voice-cloning fraud increasingly relies on surgical injection: a genuine conversation in which only one or two sentences are replaced by synthetic speech. Utterance-level deepfake detectors emit a single real/fake label per clip and cannot report where the synthetic speech lies. We formalise this as Temporal Deepfake Localisation in Multi-Speaker Conversations (TDLMC), show that equal error rate and min-DCF are ill-posed once a file contains both classes, and propose temporal metrics for this regime. Our contribution is a training-free five-stage pipeline that wraps a frozen binary detector and adds segment-level output with no retraining, using a two-threshold hysteresis finitestate-machine decoder to turn noisy window scores into coherent intervals. On 180 constructed multi-speaker conversations from ASVspoof 5, the system attains temporal intersection-over-union 0.90, temporal detection rate 0.95, and MS-DCF 0.26 with a strong backbone, and its false-alarm rate on genuine speech is below 6%, falling under 2% on genuine real multi-speaker dialogue (AMI). Under an identical pipeline, a trained localiser improves temporal IoU by only about 0.04, bounding the cost of forgoing supervision. Evaluated across three frozen detectors under one decoder whose constants are selected on a held-out calibration split, and with a controlled analysis attributing the residual false-alarm rate to a backbone domain gap rather than to the decoder, this provides the first zero-shot baseline and a reusable benchmark for TDLMC.

Generated hand-object interaction (HOI) videos provide a controllable way to propose manipulation motions. Simulation-based HOI tracking can translate such kinematic references into feasible low-level control, but its scalability is limited by the lack of reliable reference motions. We therefore combine generated videos with simulation-based HOI grounding: during training, generated videos provide diverse motion references for learning a multi-object, multi-trajectory HOI tracker, and at deployment, the video model produces motion plans that are executed by the learned tracker. In particular, we propose a method that enables scalable reference generation by HOI reconstruction with minimal manual intervention and successfully grounds more than 1,500 generated videos in simulation, achieving success rates over 25 percentage points higher than those of baselines during simulation-based training. In real-world closed-loop experiments, it achieves diverse grasps, including functional grasps, non-prehensile manipulation, and post-grasp object-pose tracking. Videos and code are available at https://boyuan-an.github.io/GALATEA/.

Clinical notes contain personally identifiable information (PII), restricting reuse for research and medical AI, especially when data cannot leave an institution. We developed MedDeID, an on-premises framework combining in-house annotation and synthetic-note generation with model training, inference, pseudonymisation and evaluation. On an independently annotated, adjudicated 300-note Dutch hospital benchmark, a hospital-trained compact transformer detected 98.9% of identifying text while redacting 0.24% of text outside annotated identifiers; a synthetic-only counterpart detected 96.1%. On 100 primary-care notes, the synthetic-trained model achieved higher recall than the hospital-trained model (90.3% versus 87.0%) and greater robustness to identifier-format perturbations. An English instantiation trained without real text detected 99.7% and 98.9% of annotated identifier characters on two external synthetic benchmarks. These results demonstrate transfer of the workflow to another language, but not clinical English performance. MedDeID provides a route to locally governed de-identification using real or synthetic training data.

Large language model agents produce fluent action sequences across a wide range of tasks, yet they fail in characteristic ways once the environment becomes partially observable. Ambiguous feedback pushes them into premature commitments. A single informative observation can collapse their uncertainty onto the wrong hypothesis. Policies drift as the history grows. We trace these symptoms to a common structural cause. An LLM agent, as commonly deployed, is a history-conditioned policy with no explicit belief over hidden state. We propose an architectural fix. The Belief-State Engine (BSE) is an inference module placed outside the LLM. It maintains a Bayesian posterior over the latent states of a given POMDP (Partially Observable Markov Decision Process) model, and at each decision step it exposes only that posterior to the LLM. The raw action-observation log is not shown. We set out a minimal four-axiom specification of what a belief-consistent internal state must satisfy, and prove that the LLM paired with the BSE is a sound Markov policy on the belief MDP induced by the underlying POMDP. It therefore inherits the Bellman optimality guarantees of classical POMDP theory, provided the LLM is never exposed to the raw history. We evaluate the architecture on the Tiger POMDP and a red-team attack-graph task, against six baselines: a reactive LLM, Chain-of-Thought, ReAct, a natural-language belief tracker, QMDP, and POMCP. Across both domains, the BSE-augmented agent improves task return, belief calibration, and decision consistency. Ten targeted ablations isolate the contribution of each architectural choice confirms that the effect is not specific to any one model. Code, environment specifications, prompt templates, and seed logs accompany this paper.

Learning-based motion planners pay at training what classical planners pay per query. Trained in world coordinates, they relearn the same motion at every position and orientation. Existing work restores the missing rigid-body equivariance in the training data, in the inference operator, or in the weights, and each carries a cost. We ask how much of that equivariance the planning query supplies for free. A start s and a goal g determine a frame in closed form, with origin at their midpoint and first axis along g-s. Expressing trajectory and obstacles in that frame removes three translations and two rotations of SE(3), at initialisation, for one cross product per query and with no constraint on the architecture. A single rotation about the start-goal axis remains, and no continuous rule removes it. On a cluttered 3D benchmark, holding architecture, data and budget fixed, the frame raises the held-out collision-free rate from 14.60% to 51.10%, where a straight segment from start to goal scores 15.6% and the world-frame model does not beat it. We build all three mechanisms for the residual rotation and each is worth under a point, though the equivariant backbone reaches any given level two to three times sooner. What the representation supplies therefore dominates what any mechanism enforces, and the standard diagnostic does not see the difference: two models with indistinguishable non-equivariance residuals differ by 28 points. Calibrated against a non-symmetry intervention, the frame is not even the largest effect available, since local geometry is worth +40.0 where the frame is worth +36.5.

Vision--language--action (VLA) models acquire broad generalization through large-scale pretraining, yet adapting them to a new task and robot embodiment still requires post-training on newly collected data. Unlike pretraining, post-training targets task- and embodiment-specific adaptation, making it particularly sensitive to data quality. In practice, collected robot datasets often contain heterogeneous errors, including execution mistakes, sensor drift, and timestamp misalignment, which can impair post-training and policy performance. Manual inspection is costly, while existing data-cleaning methods are typically tailored to particular corruption types. To address these challenges, we introduce \textsc{RoboDrop}, a data-curation framework that audits supervision using local gradient compatibility measured along the training trajectory as a proxy for its effect on post-training performance. During a one-epoch warm-up run, RoboDrop scores each candidate sample online by comparing its gradient with those of task-semantic and visually matched validation samples. The resulting sample scores are aggregated at the episode level, and a simple automatic post-processing rule converts them into filtering decisions. We evaluate RoboDrop on controlled observation--action corruptions, naturally suboptimal demonstrations in simulation, and real-robot datasets containing non-expert collection errors. Across these settings, RoboDrop more accurately distinguishes unreliable demonstrations than prior methods, while post-training on the curated data consistently yields stronger downstream policies, with average real-robot rollout success rising from $35.0\%$ to $67.5\%$. These results establish training-trajectory-aware, context-conditioned supervision auditing as an effective approach to robust VLA post-training.

EdgeAI systems are increasingly employing computer vision applications to enable intelligent, on-device decision-making in real-time. However, these deployments face highly dynamic operational conditions, with fluctuating constraints on latency, power availability, and memory resources. Deep Neural Networks (DNN), which follow fixed computational execution flows, lack the flexibility to adapt to such variability, resulting in inefficient and suboptimal performance in edge scenarios. This underscores the need for architectures that are not only efficient but also dynamically scalable at runtime. In this paper, we propose Elastoformer: A framework that transforms conventional neural networks (NN) into Elastic NN capable of real-time elastic inference. Unlike the conventional bag-of-models approach, which requires maintaining multiple independent models for different operating conditions, Elastoformer offers a single, modular solution that dynamically switches between multiple modes of operation at runtime, adapting efficiently to the changing computational budgets of edge devices without the overhead of managing separate models. Experiments reveal that our framework achieves up to 85% reduction in computation FLOPs, 50% reduction in latency and 76% reduction in memory overhead, while showcasing the architecture agnostic nature of the framework across both Vision Transformers and CNNs. Our code is available at https://github.com/sudaksh14/Elastoformer.

We introduce MetroLLM-Bench, a 955-case benchmark for testing language models as the policy layer of a transit kiosk. It covers six real metro systems, ranging from 37 to 414 stations, and eleven categories that include routing, fare calculation, disruptions, accessibility, and adversarial input. In each case, the model must call structured tools and submit a machine-renderable terminal state containing an outcome, a per-ticket fare quote when applicable, and a kiosk action. Fourteen deterministic scoring components form Tier 1; eight semantic-quality components form Tier 2, six of which use a language-model judge. We report Tier 1 and the combined score of both tiers. A stratified 75/25 split reserves 717 cases for training-data generation and 238 for held-out evaluation. We evaluate twenty-six models from six vendors, of which twenty-three are ranked. On the held-out partition, a 4B Qwen 3.5 student trained through parameter-efficient fine-tuning (PEFT) exceeds both GPT-5.6 tiers on Tier 1 (91.3 against 90.6 and 90.0) and matches GPT-5.4 full at maximum reasoning effort (91.4), with a 2.6 GB Q4_K_M footprint. Larger 9B and 27B students provide no further Tier 1 improvement over the 4B student at this training scale. Across the four Qwen sizes, the PEFT gain over the corresponding base model decreases from +7.03 points at 2B (three training seeds) to -0.91 at 27B; every seed shows the same direction at every size. A deterministic rule-based baseline reaches 84.6 on Tier 1, with the remaining language-model advantage concentrated in policy adaptation, compound scenarios, accessibility, and temporal reasoning. Muse Glimmer 30B leads the composite ranking, and serving configuration alone moves the Qwen 3.5-to-3.8 comparison by 2.7 Tier 1 points. The benchmark, harness, reproduction guide, and fine-tuned students are released at https://github.com/continker/metrollm-bench.

Terminology-aware translation asks for more than a correct translation: the output must use the exact terms a glossary prescribes. The standard recipe, fine-tuning on glossary-annotated translation pairs, hides an inefficiency: for most examples the glossary prescribes exactly what the model would have produced anyway, so they teach nothing about following a glossary. We therefore keep only the examples where the model's own translation contradicts the glossary. In a controlled study at fixed data volume, this selection alone raises term accuracy from 78.7% to 89.9%. The filtered data, built by a two-way synthetic pipeline on open models, is part of the instruction-tuning mixture of our public release SalamandraTA-7b-instruct v3.0, which, used exactly as released and wrapped in a document-level inference pipeline, forms the BSC submission to the WMT26 Terminology Shared Task Track 1. At the official WMT26 evaluation, our system achieves 94.2% term success at 74.6 chrF++, with only two of the twenty-two submissions outperforming it on both metrics. On last year's benchmark, it also surpasses our GRPO-based system, despite being trained solely with ordinary supervised fine-tuning.

A natural way to cut reasoning-model inference cost is to repeatedly probe a single partial trajectory for its current answer and stop once probes agree -- self-consensus. We ask whether any such rule is both safe and token-saving, and whether one can be selected once and reused. A preregistered sweep of 3,520 consensus rules, replayed on frozen trajectories from two models and three benchmarks, clears none of three acceptance gates fixed in advance; the frontier reproduces on a held-out split and on two unseen models -- while a boundary-confidence control (DEER) swept through the same pipeline clears all three. The reason lies in the signal: agreement establishes that the current answer persists under a fixed probing procedure, not that the reasoning has terminated -- a consensus-termination gap. Stopping on it commits non-terminal answers. At a rule still saving 32% of the tokens, one stop in nine fires on an answer the trajectory itself later abandons, and most of those stops cut off a correction it would otherwise have made. Widening the agreement window does not remove them: the share levels off near 7%, and by then the saving has fallen to 8%. Probe re-wording and a hand-labelled error taxonomy show the agreed answer is often a placeholder the model had not settled on. Used on its own as the stop signal, agreement fails not because it is insufficiently strict, but because it repeatedly measures the wrong object.

Real-world video understanding requires integrating visual, audio, textual, and temporal evidence distributed across a video. Yet many pipelines use a fixed video context and single-pass inference, limiting adaptive evidence acquisition when observations are incomplete, ambiguous, or conflicting. We present VLX-VR, an agentic-aware video reasoning model trained within a video reasoning framework defined by a Think--Memory--Observation loop. At each step, VLX-VR determines the needed evidence, invokes read_memory or write_memory, incorporates the returned Observation, and decides whether to continue or produce the task output. We train VLX-VR with multimodal data, including videos and agent trajectories, using reinforcement learning to learn evidence acquisition, memory use, and termination. On MINERVA, VLX-VR achieves state-of-the-art performance among the models included in our comparison, with 78.79% accuracy. Under the original three duration groups, its accuracies are 76.70%, 78.73%, and 80.92%, with a cross-duration accuracy variance of 2.97~$\mathrm{pp}^2$. On correctly answered samples, 96.20% of VLX-VR's reasoning traces are consistent with the MINERVA reference reasoning traces and the evidence described by them, while approximately 75.80% of all evaluated samples satisfy both answer correctness and this evidence-grounded trace criterion. These results show strong performance and broadly stable behavior across durations, while counting, state changes, causal reasoning, and spatial perception remain challenging.

We study offline inference for the optimal value in reinforcement learning. Two new nuisances are derived as fixed points of a self-induced Bellman equation, in which we approximate the maximum Bellman operator by its softmax correspondence. We propose a debiased estimator through the Neyman orthogonality and establish its asymptotic normality under diverging horizons even when the behavior policy changes with time, as long as the nuisances have the statistical rates that can be achieved by many machine learning methods. We provide a concrete estimating procedure for these nuisances and show they can lead to valid inference. Synthetic experiments validate the numerical performance of our inference method, and we implement it in real-life decision-making problems, including bike repositioning and AI agentic tool use.

Evaluating video captioning remains a critical challenge for Visual Large Language Models (VLLMs). Existing metrics primarily rely on matching generated text against ground-truth references. This paradigm suffers from the ``one-to-many'' nature of video description, where high-quality captions are often penalized for lexical mismatches or valid shifts in visual focus. Furthermore, such assessments are typically one-dimensional, failing to provide a fine-grained analysis of caption quality. To address this, we redefine caption quality through the lens of information fidelity: A caption must maximize the coverage of salient visual information while ensuring strict factuality. We introduce CapQuiz, a novel reference-free benchmark that assesses captions based on their utility in answering human-verified, fine-grained, multiple-choice questions derived from the video. CapQuiz features a hierarchical taxonomy of 10 question types (spanning Descriptive and Inferential categories) across 24 diverse video domains. Extensive experiments demonstrate that CapQuiz correlates significantly better with human judgments than existing metrics and offers interpretable insights into model performance.

Large Language Models (LLMs) have achieved remarkable progress across natural language processing (NLP) tasks, yet their capabilities degrade sharply for low-resource languages and dialectally diverse settings. Bangla, the world's sixth most spoken language, exemplifies this gap: existing resources overwhelmingly target Standard Bangla, leaving its regional dialects without the benchmarks needed to develop or evaluate dialect-aware systems. We address this gap with 5-Dialects-BN, the first multi-annotation Bangla dialect benchmark to align Romanized transliteration with dialectal text, Standard Bangla, English, and subjectivity labels across five regional varieties. The dataset comprises 6,000 manually annotated entries spanning five major dialects: Chittagong, Barisal, Noakhali, Sylhet, and Rangpur (Chittagong 1,900; Noakhali 1,500; Sylhet 1,200; Barisal 700; Rangpur 700), reflecting natural online availability. Each entry is enriched with five aligned annotations: the original dialectal text, a Romanized transliteration, an English translation, a Standard Bangla translation, and a subjectivity label (subjective vs. objective). Annotations were produced and cross-validated by native speakers and undergraduate linguistics students to ensure dialectal authenticity and semantic fidelity. The resulting resource supports a diverse suite of tasks, including dialect identification, dialect-to-standard normalization, machine translation, subjectivity classification, and parameter-efficient fine-tuning (e.g., LoRA) of multilingual LLMs. By providing a standardized, multi-annotation benchmark, 5-Dialects-BN enables principled evaluation of LLMs on dialectally diverse Bangla and lays a foundation for further research in low-resource, dialect-aware NLP.

Cross-lingual sentence encoders enable scalable transfer across hundreds of languages, powering applications such as translation mining and zero-shot learning in low-resource settings. Although trained for sentence-level alignment, they are increasingly also applied to token-level tasks such as hallucination detection and sequence tagging, exposing a mismatch between training and usage. We propose SALT, a lightweight post-training method that improves token representations by injecting span-level supervision into existing sentence encoders. Across five multilingual token-level benchmarks, SALT achieves the best overall results on four of them, outperforming alternative fine-tuning strategies and competitive encoders. It also improves sentence-level performance on cross-lingual retrieval and classification tasks. These results demonstrate that span-level supervision is an effective signal for improving both token and sentence representations.

Real-world detectors must often interpret functional or ambiguous prompts, yet conventional models such as YOLO remain restricted to fixed class lists. Even open-vocabulary models like YOLO-World frequently misalign vague language with the intended objects. Building on our prior work Commonsense-Guided Open-World Object Detection Using LLMs and Visual-Semantic Matching, we address YOLO-World's limitations in grounding task-driven queries. We propose Vague2Detect, a hybrid pipeline in which a fine-tuned Sentence-BERT retrieves candidates from a structured household Knowledge Base (KB), and YOLO-World verifies their presence in the image. For prompts outside the KB, a large language model (GPT-3.5-turbo) generates candidate descriptions, dynamically expanding the KB to cover novel concepts. On a benchmark of household scenes using custom images and an Open Images V7 subset, YOLO-World alone achieves only 32% Vague Prompt Success Rate (VPSR), the ability to map ambiguous queries to correct detections. In contrast, Vague2Detect improves performance to 61% VPSR with high precision, and up to 85% when augmented with GPT fallback.

Machine learning-based credit scoring is increasingly central to Peer-to-Peer (P2P) lending, yet its resilience to adversarial manipulation, where applicants strategically alter self-reported inputs to secure favourable decisions, remains poorly understood. Most adversarial-robustness evidence comes from image and text domains and evaluates a single attack against a matching defence, offering little guidance on how defences generalise across attack types in tabular credit data. We address this with a systematic train-test robustness benchmark on a large Lending Club subset, spanning three model families (logistic regression, a feed-forward neural network, and a transformer for tabular data) and four attacks confined to applicant-mutable features: Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), Salt-and-Pepper (S&P) noise, and DeepFool, plus a mixed-attack regime. Across a full grid evaluated with stratified cross-validation, adversarial training sharply improves robustness against the attack it is trained on and transfers well within the gradient-based family, but transfers weakly to non-gradient corruption, so single-attack defences overstate real-world resilience. Mixed training delivers the most balanced robustness across heterogeneous attacks while preserving clean-test performance, supporting multi-attack stress testing in credit-model governance.

Generalist robot policies have demonstrated strong generalization across robotic manipulation tasks, yet their success rates remain limited in com- plex long-horizon scenarios. Recent methods improve Visual-Language-Action (VLA) policies through online reinforcement learning on real robots, but such training relies on costly physical interactions, suffers from low sample efficiency, and may introduce hardware and safety risks. World models offer a promising alternative by enabling policy optimization with imagined rollouts. However, long-horizon rollouts generated by world models often suffer from prediction hal- lucinations, producing biased state transitions that can mislead policy learning. To address this issue, we propose Hallucination-aware World Model-based Pol- icy Optimization (HaWMPO), a closed-loop reinforcement learning pipeline for VLA policy post-training with world models. Specifically, HaWMPO introduces an action-conditioned hallucination-aware model to estimate the reliability of gen- erated image sequences, and incorporates hallucination scores into group relative policy optimization through a Reward-Soft mechanism, suppressing unreliable ac- tion chunks during training. On the LIBERO benchmark, HaWMPO achieves the best average success rate, with gains of 15.0% over the base model and 2.8% over the strongest baseline; real-world experiments on a G1 robot further validate its effectiveness, raising the average success rate on two manipulation tasks from 67.5% to 80.0%.

Modern vision-language-action (VLA) policies predict a whole chunk of actions: one to two seconds of coordinated motion emitted in a single forward pass. Yet an action chunk is essentially a short multivariate trajectory, but inside these models it is a sequence of generic per-timestep hidden tokens decoded by a linear head. This under-serves two motion structures. First, frequency: a chunk superimposes a smooth global trend and fine corrective motion across time scales, and a single token entangles them. Second, cross-phase geometry: motions of different phases (reach, contact, grasp adjustment, settling) unfold along very different, near-orthogonal directions in representation space, yet are tightly related for the task and arise across the time axis. Dot-product attention scores alignment by an inner product, so it favors aligned tokens and is least sensitive near orthogonality, leaving such relationships for the network to recover through a detour. We introduce Time-Frequency Geometric Cross-Attention (TFGCA), a drop-in module repairing both blind spots. TFGCA uses a per-dimension learnable stationary wavelet transform to decompose the action chunk into time-frequency tokens, and each time token retrieves information from them via a cross-attention that fuses the dot product (similarity) with the wedge-product magnitude (sensitive to near-orthogonality) through a learnable weight. A zero-initialized residual reproduces the base behavior at initialization, so it can be dropped onto a pretrained VLA and fine-tuned jointly. Relative to the same-source base, TFGCA improves in-distribution LIBERO by +1.5 on average, the OOD LIBERO-Plus by +6.3, the randomized average under RoboTwin domain randomization by +28.5, and the overall success rate on three real-robot AgiBot A2 tasks by +11.67 points, with larger gains out of distribution.

Motion tracking provides a scalable recipe for humanoid whole-body control. By design, the resulting trackers lack exteroceptive feedback hence reacting to the environment remains the responsibility of a higher-level planner. Existing perceptive controllers train geometry-only encoders from scratch, trading semantics for sim-to-real ease, and typically rely on teacher-student distillation for a task of interest. We present ViBe, a post-training framework for adapting motion trackers to perceptive control tasks. We leverage pre-trained visual encoders with a multi-query extractor module to learn task-relevant perceptive feedback. This feedback is grafted onto the tracker's input via low-rank adapters, enabling parameter-efficient fine-tuning. Given a task reward and a reference dataset, this modular controller can be adapted directly via policy optimization. Across four tasks, ViBe shows zero-shot sim-to-real transfer spanning perceptive walking on curbs and parkour, Repose Cube, omni-object loco-manipulation, and dodgeball, with visually robust performance across outdoor, low-light, and RGB distractor conditions. Finally, we solve a goal-oriented Repose Cube task with a deliberately simple planner, demonstrating the efficacy of perceptive controllers, adapted by our approach.

Machine learning models used in engineering are typically trained within limited operating ranges, yet reliable predictions are often required beyond these domains. Consequently, the primary challenge is extrapolation rather than interpolation. Rigorous validation is hindered by the scarcity of data outside the training range. To address this limitation, a novel extrapolation framework is integrated with established machine learning architectures to enable accurate and physically consistent predictions beyond the training domain. The framework is established by systematically evaluating two physics-guided architectures: a Bidirectional Long Short-Term Memory (BiLSTM) network and a Physics-Informed Neural Network (PINN). A classical one-dimensional transient diffusion problem is adopted as a benchmark because its exact analytical solution provides unlimited, reliable data across the spatio-temporal domain, enabling rigorous quantitative validation. The problem is particularly challenging because the solution evolves from an initial singularity through a strongly nonlinear transient regime before approaching a steady-state linear profile. When training data are confined to an intermediate portion of this evolution, backward extrapolation toward the singularity becomes especially demanding. To improve reliability, physics-guided coordinate transformations, boundary-aware learning strategies, and stability-enhancing temporal marching are incorporated. Extrapolation is evaluated using a train-predict-validate-extend strategy, in which validated predictions are recursively added to the training set to progressively extend the prediction horizon. The results demonstrate accurate and physically consistent predictions beyond the training domain, highlighting the framework's potential for engineering applications where data availability is limited.

Meta-learning has emerged as an effective paradigm for transferring knowledge across sequential bandit tasks. While substantial progress has been made for stochastic bandits and non-contextual adversarial bandits, meta-learning for adversarial linear contextual bandits (ALCBs) with random action sets remains largely unexplored. To address this problem, we propose Meta-LinEXP3, an online-within-online algorithm that constructs a predictable task-level prior from completed tasks to guide the inner LinEXP3 learner. For known context distributions, we develop a policy-centered estimator that achieves an intrinsic-dimension $\mathcal{O}(\sqrt{n})$ per-task regret bound. For unknown distributions, we introduce a past-only regularized moment estimator with an $\mathcal{O}(n^{2/3})$ leading regret term and explicit finite-sample error. We further establish a direct connection between prior accuracy and transfer regret, showing that increasingly accurate priors yield sublinear transfer-dependent regret across tasks. Experiments demonstrate the effectiveness of Meta-LinEXP3, including its application to structured hyperspectral tensor sampling.

Language models can fluently describe how they would behave: whether they would cave to pushback, misuse a tool, or lie under pressure. Is that description actually about the model speaking? We turn self-knowledge into a prediction test. Across nine behavioral evaluations, we measure how a model behaves under different conditions, ask it to predict those rates, and compare its predictions with controls that remove the self from the question. We find that: (i) Direct self-report is weak (r = +0.04), and even showing the model the exact items only raises prediction to +0.24. Crucially, the same item-informed question about "capable AI agents in general" does just as well (+0.28), while other models' answers about themselves predict the target model at least as well as its own. (ii) Frontier scale does not detectably change this pattern: any gains in prediction are not self-specific, and are consistent with a better theory of how AI assistants behave rather than better self-knowledge. (iii) First-person framing does have one robust effect: it shifts reports in the flattering direction, understating harmful behavior relative to the same question about a generic agent. (iv) Finetuning on a model's own behavioral record can teach narrow self-predictions, but it also changes the behavior being predicted and the gains do not transfer broadly. The practical implication is simple: asking a model what it would do mostly reveals a theory of AI assistants in general, plus a favorable bias, rather than privileged knowledge of that model.

Large Language Models (LLMs) have shown promise for translating Natural Language (NL) planning descriptions into PDDL problem instances. However, standard evaluation criteria such as syntactic validity or planner success can substantially overestimate faithfulness to the described task: a generated problem may be parseable and solvable while misrepresenting the intended initial state, goal, object structure, or optimization target. This paper studies an end-to-end NL-to-PDDL pipeline that combines LLM generation, checks in terms of PDDL parsing, planning and validation, a domain-conformance checker, an LLM critic, and iterative repair. Fine-grained repair feedback is constructed from the domain description, the generated problem, the natural language problem description, and operational diagnostics. Reference-based comparisons against curated benchmark PDDL problem descriptions are used for post-hoc benchmark analysis, and these offline checks include renaming-invariant structural matching and semantic equivalence, where domain support is available. Across Planetarium, AutoPlanBench, and curated PDDL~2.1 problems, results show that operational success and benchmark-reference reconstruction can diverge substantially. Results also show that structured repair can be useful, and that PDDL~2.1 remains challenging for reference reconstruction, even when operational success improves.

Video-language models and video agents can produce hallucinations that conflict with spatiotemporal evidence. Existing benchmarks mainly evaluate model hallucinations, and heterogeneous mechanisms make detector reliability difficult to compare. We introduce VidHalLoc, a benchmark that evaluates hallucination detection methods under a unified diagnostic evaluation protocol using 2,000 adversarial hallucination samples across Video Question Answering and Video Captioning tasks, spanning Ontology and Dynamic hallucination categories. To construct VidHalLoc efficiently, we introduce VideoHALO, a Harness Engineering-informed multi-agent workflow that decomposes data construction into four executable stages supported by a memory system and a communication protocol. Evaluation of fifteen methods reveals that the four dedicated detectors peak at an Overall accuracy of only 34.63%, indicating limited reliability across video hallucination types [Dataset Repository: https://huggingface.co/datasets/wesfggfd/VidHalLoc].

Proteolysis-targeting chimeras (PROTACs) have emerged as a transformative therapeutic strategy that selectively degrades historically ''undruggable'' targets via the ubiquitin-proteasome system. Despite growing efforts to develop computational predictors of PROTAC degradation activity, existing supervised approaches remain severely challenged by data scarcity and imbalance across E3 ligases, limiting their ability to generalize beyond well-studied ligase contexts. In practice, labeled data are heavily concentrated on a few ligases (e.g., CRBN and VHL), while the majority of E3 ligases remain underexplored yet are critical for expanding the design space of targeted degraders. Developing methods that enable robust cross-ligase generalization with minimal labeled data is therefore essential for improving the practical utility of computational PROTAC discovery. We reformulate PROTAC degradation activity prediction across E3 ligases as a few-shot meta-learning problem and present ProMeta, a prototype-based graph neural network trained through episodic meta-learning on source-E3 tasks and evaluated on held-out target-E3 tasks through support-conditioned inference. ProMeta performs inference without updating the encoder by dynamically estimating class prototypes from minimal target-ligase support samples. On the CRBN-to-VHL benchmark, ProMeta achieves AUROC values of 0.796 under K=2, Q=3 and 0.883 under K=2, Q=5, improving by 19.9% and 6.8%, respectively, over the corresponding supervised GNN baseline. Reverse VHL-to-CRBN transfer under the same protocol yielded AUROC values of 0.702 (K=2, Q=3) and 0.821 (K=2, Q=5), confirming bidirectional applicability while revealing direction and data-regime dependence. Together, these results support ProMeta as a practical framework for cross-ligase few-shot prediction under the evaluated support/query protocols.

Automatic Speech Recognition (ASR) technology is fundamental to customer service automation and large-scale transcription. However, even advanced ASR models exhibit inevitable errors in complex real-world environments such as call center conversations. When privacy restrictions preclude audio access, error correction must rely on text-based post-editing. Existing text-only approaches face significant challenges in low-resource languages, mainly due to a critical scarcity of annotated corpora and tailored correction methodologies. For Korean, this resource gap is particularly pronounced, as existing resources are predominantly designed for ASR training rather than text-based error correction. To address this, we introduce DasanCallDial, the first large-scale Korean benchmark dataset specifically curated for dialogue-level ASR error correction. Derived from genuine call center interactions, it comprises 1,974 dialogues with 115,460 utterances. Leveraging this resource, we propose Detector-Gated Contextual Span Correction (DCSC), a text-only post-editing framework for error-sparse Korean speech recognition transcripts. DCSC combines an encoder-based detector that first performs token-level error detection, followed by a language model-based corrector trained to rectify fine-grained span-level errors. Additionally, we employ dialogue-level context augmentation to enable the model to leverage discourse history for disambiguation. By employing multi-level granularity, our method achieves state-of-the-art performance, effectively overcoming the limitations of general LLMs in low-resource settings.

LLMs have been used to simulate human decision-making in professional settings, yet their behaviors in common-law jury trials remain unexplored. We study when and how a defendant's courtroom statement affects LLM-simulated jurors, focusing on persuasion, ideological bias, and background-based affinity. To support the analysis, we introduce JuryBench, a benchmark containing controversial criminal cases in U.S. criminal law. In each case, a defendant can claim various plausible justifications to support acquittal or reduced liability. We fix the base case and design defendants of different backgrounds, who give courtroom statements with varying emotional appeal or rebuttal. Jurors with diverse ideological profiles across the spectrum are simulated. We examine 20 frontier LLMs, resulting in a total of 432K decisions and rationales, and quantify changes in verdict severity. Our findings show that LLM-jury simulation echoes many human-jury findings. First, emotional persuasion can be detrimental, since jurors may perceive it as evidence of guilt or inconsistency. Next, we show that background fit between jurors and defendants is a stronger and significant factor than other isolated factors, and that jurors are in general harsher toward opposite-background defendants and lenient toward same-background ones. Finally, we find that juror ideology also strongly shapes severity judgments. These findings highlight both the promise and risks of using LLMs to model jury reasoning and call for careful evaluation. The data and code are available at https://github.com/choyingw/JuryBench

Recent advances in Intrinsic Image Decomposition (IID) have increasingly relied on generative models. However, progress remains limited by three key challenges: (a) insufficient physical consistency, (b) high computational cost at inference time, and (c) limited generalization capabilities. In this work, we show that latent bridge matching (LBM) effectively addresses these limitations for albedo estimation. We introduce a novel LBM-based architecture that enforces physical consistency through a pixel reconstruction loss, benefits from the inherent efficiency of LBM low-cost inference, and improves generalization across diverse datasets by incorporating a shading conditioning. In this extended version, we additionally show that conditioning the shading estimator itself on the predicted albedo further improves reconstruction fidelity, and we benchmark our best model against stateof-the-art IID methods across five real and synthetic datasets.

Existing evaluation frameworks mostly assess only one part of AI agents, such as task completion (AgentBench) or security robustness (AgentDojo, ASB), rather than the complete pipeline of planning, tool selection, tool execution, memory and reasoning. Failures can occur at any stage, yet existing benchmarks rarely identify their precise source. AgentAudit evaluates the entire execution trace across ten capability, grounding, security and behavioural dimensions, namely instruction integrity, planner, memory, tool selection, tool invocation, tool correctness, alignment, tool faithfulness, security and execution integrity, combined with behavioural classification and failure attribution to pinpoint the exact stage responsible for an observed failure. AgentAudit can evaluate any LLM-based AI agent, since it attaches to the agent instead of replacing it. It reads only the recorded execution trace and does not interfere with how the agent runs, so it places no constraint on the agent's internal implementation. We evaluate five language models (OpenAI GPT-5, Claude Sonnet 5, Sarvam 105B, Llama 3.3 70B and Gemini 2.5 Flash) across nine capability and adversarial tasks. Claude Sonnet 5 and GPT-5 obtain the highest mean Composite Trust Scores (95.1 and 80.6 out of 100, respectively), while Sarvam 105B, Llama 3.3 70B and Gemini 2.5 Flash trail substantially (57.6, 45.7 and 22.6). All traces were scored by a single fixed judge model, which was itself one of the evaluated models, a limitation discussed in Section VII.E. More importantly, models with similar task-completion behaviour can diverge sharply in trustworthiness, as several non-frontier models are repeatedly classified Unsafe_Compliance on adversarial tasks rather than merely failing them, a distinction that pass/fail benchmarks cannot surface.

Choosing a deep learning architecture for label-free single-cell classification remains an open question, with microscopy benchmarks reporting conflicting conclusions about CNNs versus transformers. We present a controlled benchmark on LIVECell phase-contrast microscopy data using source-image-disjoint train/validation/test splits to prevent parent-image leakage and matched optimisation, augmentation, and evaluation protocols across EfficientNet, Vision Transformer (ViT), and EVA-02 models. This allows the effects of architecture, pretraining, fine-tuning, tokenisation, and distillation to be disentangled. We find that the previously reported CNN advantage is largely explained by pretraining rather than architecture: the smallest pretrained model outperforms the strongest model trained from scratch despite far fewer parameters. Pretraining improves macro-F1 by 3-4 points, while the gap between the best pretrained CNN and transformer is below 0.5 points. Architectural choices nevertheless matter: ViT-S/8 outperforms ViT-S/16 and matches the four-times-larger ViT-B/16 at a quarter of the parameters, showing that finer tokenisation benefits small cell crops. Conversely, layer-wise learning-rate decay, central to the EVA-02 fine-tuning recipe, degrades performance, highlighting that transfer heuristics from natural-image recognition may not generalise to microscopy. Finally, knowledge distillation substantially improves the deployment frontier: compact EfficientNet-B0 students distilled from teacher councils outperform every individually trained backbone, including the EfficientNet-B5 and EVA-02 teachers. Overall, our results show that rigorous control of pretraining and evaluation is essential for interpreting biomedical architecture benchmarks, while distillation may be a more effective route to practical single-cell classification than architecture choice alone.

The rapid adoption of generative AI tools such as ChatGPT has sparked intense debate about their risks and opportunities for education, as well as the ways researchers should investigate them. In this paper, we approach these discussions through an analogy with the Thermomix, a smart kitchen appliance that has similarly provoked both enthusiasm and critique. By mapping Thermomix use cases onto examples of learning with generative AI, and situating them within the ICAP and SAMR frameworks, we show how different modes of tool use can either support or undermine meaningful engagement and learning. The Thermomix metaphor underscores that the central question is not whether learners employ AI, but how such use shapes their learning processes. In doing so, we provide a conceptual lens for researchers and practitioners to critically examine - and more effectively guide - the integration of generative AI into educational practice.

LLM agents for enterprise systems of record cannot be evaluated on customer production data, and no existing substitute provides ground truth. We present the Era by Eon Benchmark for evaluating LLM agents that use enterprise tools. The benchmark is built around a complete fictional company. It includes product simulators, company-specific internal databases, benchmark questions, and computed answer keys. Industry, company size, business model, application portfolio, and a seed define each company. One seeded entity graph supplies shared company data to simulators of Salesforce, Zendesk, Slack, Gong, and other products. A questionconditioned generator creates the schemas and records for internal databases. It takes shared entities, keys, and values from the same graph before generating database-specific facts. Both mechanisms therefore describe one consistent enterprise estate. Every expected answer is computed from the final records, so grading is exact. Design and answer-key checks validate the internal databases. A realism scorecard and adversarial detector validate the entity graph. Across 23 generated companies, the mean realism score rose from 61.8 to 97.0, with zero records flagged as synthetic. In the reported simulator-track comparison, nine models answered the same 33 questions three times each. Accuracy estimates ranged from 42.4% to 76.8%, and three of 36 pairwise differences remained supported after correction.

The advance of multimodal large language models (MLLMs) has fundamentally reshaped the paradigm of human-computer interaction, especially speech interaction models capable of seamless conversations. Despite remarkable performance as general voice assistants, their performance in specialized domains remains underexplored, particularly in scientific areas. Scientific interactions introduce formidable challenges, involving rare technical terminology, spoken norms of abbreviations, and the natural verbalization of symbolic special expressions. In this paper, we introduce S$^3$-Bench, a systematic evaluation framework covering 10 major disciplines, consisting of a Knowledge set for speech question-answering and a Dialogue set for multi-turn progressive interactions with simulated user agents. By decomposing a complete atomic turn into stages of speech recognition, perception, knowledge utilization with reasoning, and response pronunciation, we systematically characterize the common challenges and performance tradeoffs of existing approaches. Furthermore, experiments on multi-turn interactions reveal persistent limitations in user adaptation and the generation of accurate, comprehensive, and efficient responses.

In recent years, AI agents have evolved into capable assistants that carry out multi-step tasks in digital environments. The network systems community is beginning to explore these capabilities in operational and experimental settings. However, an agent operating in network systems should not be judged solely by whether it completes the immediate task. The experiment record it modifies must also remain trustworthy. We call this property artifact integrity: the record's claims must remain supported by the available evidence, confined to the scope established by that evidence, and traceable through the artifacts that encode their support. To make this property measurable, we introduce NetArtifactBench, which tests whether AI agents can repair inconsistent records derived from public network-system artifacts while preserving claims that remain supported. The benchmark contains 52 instances with injected inconsistencies ranging from direct contradictions to unstated relations spread across several artifacts. We evaluate 23 agent configurations across three general-purpose AI agent runtimes using deterministic scoring. The average contract pass rate is 65.3 % across 5,980 outputs, but no agent runtime exceeds 30 % when repair requires recovering implicit relations and propagating changes across artifacts. These results reveal a sharp boundary between local correction and complete record-level repair. Therefore, we argue that artifact integrity should become a first-class design and evaluation requirement for AI agents operating on network systems.

This paper addresses time-varying coverage control for multi-agent systems, formulated as the tracking of an evolving target density via entropy-regularized semi-discrete optimal transport. Unlike the hard Laguerre partition of the unregularized formulation, entropic regularization assigns fractions of the mass at each point to all agents, simplifying the design and numerical implementation of the control law. We derive a feedback-feedforward controller that tracks the evolving first-order optimality conditions with exponential convergence, and we investigate the role of the regularization parameter, which, above a critical threshold, renders the fully collapsed configuration locally optimal. Numerical experiments validate the theory: for moderate regularization, the proposed approach has performance comparable to the unregularized formulation and outperforms the corresponding Voronoi-based coverage baseline.

Vortex light fields carry orbital angular momentum and have attracted significant attention because of their wide applications in light field manipulations, optical communications, and quantum information processing. In nonlinear media, the balance between diffraction and self-action enables the formation of vortex solitons. However, achieving the stability of vortex solitons remains challenging due to radial and azimuthal modulation instabilities. Here, we report stable vortex solitons and vortex-soliton arrays in disclination-fractal configurations with different rotational symmetries, constructed by applying disclination operations to fractal lattice structures. The stable vortex-soliton arrays---composed of several vortex solitons---can exist along domain walls, resulting from the disclination operation applied to the topologically trivial phase. Their stability is verified through both linear stability analysis and direct simulations of perturbed propagation. For comparison, vortex-soliton arrays in conventional disclination configurations are found to be completely unstable, demonstrating the significance of the fractal configuration in stabilizing these nonlinear states. These findings reveal a new nonlinear excitation mechanism arising from the interplay between fractal geometry and disclination defects, thereby providing theoretical underpinnings for both the improved understanding of multi-field excitation phenomena in complex geometric systems and opening new avenues for the design of photonic devices based on fractal disclination structures.

Medical claims data comprise the financial details, including the expenses and billing information, as well as the clinical information, such as the diagnoses and treatments, of patients visiting medical facilities. Recently, it has been acknowledged that large databases can be constructed from medical claims data for medical research purposes. However, the clinical information within these datasets is often medically unstructured, limiting its application in comprehensive analyses. This study enhances predictive model performance for major adverse cardiovascular events (MACE), a leading cause of death worldwide. Models that predict MACE are crucial to clinical practice guidelines. We utilize a cross-attention mechanism to develop a method that effectively weights the relationships between diagnoses and treatments. Effectively repre- senting the clinical information contained in medical claims data, this approach generates more representative features for predicting MACE. The ROC-AUC score of our proposed cross-attention-based model was 0.7720, higher than other benchmark models including the conventional atherosclerotic cardiovascular disease model, the light gradient boosting machine, and a self-attention-based model. These results indicate that integrating the clinical structure of medical claims data using a cross-attention mechanism significantly enhances the performance of predictive models.

Physics-based character control is a long-standing challenge in computer graphics and robotics, requiring policies that satisfy complex dynamics while producing realistic motion. Recent Deep RL approaches, particularly imitation learning methods such as DeepMimic, have had broad impact beyond animation, influencing robotics by enabling agile and expressive behaviors. While these approaches achieve impressive results, they remain computationally inefficient to train in practice. Despite GPU-accelerated simulation, we find that end-to-end pipelines often underutilize hardware due to overheads outside the physics solver, caused by fragmented GPU kernels and CPU memory access in the critical path. We present InstantMimic, a system that addresses these inefficiencies by making the entire training loop GPU-native. Built on a GPU-native physics backend, our unified pipeline integrates simulation, environment computation, policy inference, and policy updates within a single execution flow. As a result, InstantMimic reduces training time for diverse physics-based skills to a few seconds and makes LLM-agent-driven hyperparameter search practical.

Compound LLM systems often solve a coordination problem by adding a higher-level LLM. The resulting meta-agent reads workers' outputs, writes the final answer, allocates later calls, and decides when to stop. It is expressive, but it also concentrates three control decisions in an opaque, order-sensitive model call. We ask whether the manager needs to be generative at all. UnitBoost replaces that model with a defined meta-level operator: a task-given unit map turns worker outputs into slot-value proposals, a constrained argmax assembles the output, and the slots left unfilled or unsupported become an explicit residual for the next round. The operator is order-free, records unit provenance, and gives a simple guarantee: without coupling constraints, unit-wise maximization under the same admission score dominates selection of any complete candidate. On three held-out benchmarks, it exceeds the best single candidate chosen with gold labels by 0.060-0.195 absolute task-score points and input-matched generative managers by 0.048-0.076. Replacing only the management step improves six compound-system configurations by 0.013-0.182. Residual-directed rounds raise FanOutQA cell F1 from 0.4778 to 0.5524; matched controls show that the true residual outperforms random targets and ordinary rereading, while a label-free supply signal flags exhaustion after one unproductive round. The same analysis measures three conditions in which no such gain is available (one indivisible unit, unavailable unit identity, and an endpoint that charges for every emitted unit) and quantifies cross-unit coupling as a repair cost. The manager gives up semantic freedom and gains order invariance, unit provenance, and testable failure conditions.

Robotic manipulation tasks are often decomposed into behaviors or skills. However, one often needs to predefine these behaviors for specific tasks or try to cover a wide range of tasks using generic skills. As a result, these behaviors can remain too coarse to expose the geometric, control, and scene-dependent decisions required for execution. We introduce Grounded Task Axes v2 (GTA-2), a modular multi-VLM framework that constructs executable, task-bespoke manipulation skills from reusable object-centric task-axis components. Rather than predicting actions end-to-end or composing fixed task-level primitives, GTA-2 represents each skill as semantic subtasks comprising task-relevant keypoints and axes, controller compositions, and scene-dependent parameters. Four specialized VLM agents separately decompose the task, construct an abstract task-axis skill, assign controller parameters, and ground the required visual features from RGB-D observations. This abstraction-to-grounding factorization enables zero-shot skill generation without task-specific robot demonstrations, policy training, or fine-tuning. It also keeps intermediate decisions explicit, allowing targeted human feedback to refine an incorrect stage while preserving correct components. We evaluate GTA-2 on 14 real-robot manipulation tasks against a VLA policy pi_{0.5} and two Code-as-Policies baselines using task-axis controllers or conventional robot primitives. GTA-2 achieves an average zero-shot success rate of 73.9%, exceeding the strongest baseline by 31.4 percentage points, while targeted refinement raises GTA-2's average success rate to 90.7%. Project page: https://gta2-project.github.io/

Large language models (LLMs) have been ex- ploited to generate malware, but the effective- ness of guardrails for code generation secu- rity remains unclear. We introduce CS-Guard, the first benchmark to systematically evalu- ate guardrails for code generation security. It covers 1) text-to-code generation with 1000 high-quality malware-generation prompts, 7 jailbreak attacks, and a novel fictional scenario attack (FSA) that embeds malicious intent in a legitimate fictional software-development sce- nario; and 2) code-to-code generation with 331 code prompts spanning code infilling, code completion, and code translation. We empiri- cally evaluate 9 guardrails across seven LLMs. We find that current guardrails perform poorly against malicious code-generation re- quests: for text-to-code, the average attack success rate (ASR) after jailbreaks reaches about 50% for many guardrails; for code-to- code, average ASR approaches 100% on base LLMs and remains high across many guardrails (14.4% to nearly 100%). Our FSA also achieves ASR close to 100% across many guardrails, raising major reliability concerns for real-world software development. To sup- port future research, CS-Guard uses a modular three-layer guardrail taxonomy that lets devel- opers register guardrails for evaluation. We release the benchmark and data to enable fur- ther community evaluation.

The DIRECT algorithm is a deterministic global optimization method known for its versatility and balanced exploration-exploitation strategy. However, DIRECT-type algorithms are primarily effective for low-dimensional problems and often exhibit slow convergence as dimensionality increases, limiting their applicability to more complex optimization tasks. To address this limitation, this paper introduces X-DTC-GL, a novel DIRECT-type algorithm that incorporates dynamic partitioning and hybridization techniques. The dynamic partitioning approach adaptively refines the search space based on local one-dimensional surrogate models, enabling rapid subdivision of promising hyper-rectangles. The hybridization strategy selectively employs a hill-climbing method to exploit promising regions identified by the surrogate models. Extensive experiments on four diverse benchmark suites demonstrate that X-DTC-GL significantly outperforms existing DIRECT-type baselines, achieving improvements of ~12% in solvability and ~27% in solution quality. Performance-profile analyses indicate the fastest convergence on up to ~40% of instances, the best runtime performance on ~17% of problems, and competitive overall execution times. By improving performance within the partition-based framework, these advances strengthen the algorithm's competitiveness in state-of-the-art black-box optimization.

Directional ablation removes an aligned language model's ability to refuse by projecting a single "refusal direction" out of the weights that write the residual stream. It needs no gradient-based training and no optimization, only a few hundred contrastive prompts, which makes it the canonical white-box attack on open-weight alignment. However, it has been established only on dense models up to roughly 70B parameters. We study whether it survives the shift to frontier mixture-of-experts (MoE) models whose residual streams are no longer a single tensor and whose weights ship quantized. We apply it to GLM-5.3-Flash (320B parameters, 288 routed experts, a four-wide hyper-connection residual, block-FP8). The attack survives the architecture, but what it reaches is no longer where a reader of the original recipe would look for it. Editing the attention, dense and routed-expert writers on their own removes 0.039, 0.016 and 0.148 of refusal respectively; editing all three together removes 0.776. As a result, 74% of the effect exists only under the joint intervention. The part the conventional recipe reaches by module-name matching accounts for 0.066 of that 0.776, which is why it fails silently on an MoE. The effect does not follow from removing just any direction: ablating a random direction orthogonal to it leaves refusal unchanged. A category-concentrated residue survives every edit we tried: subspaces fitted on violence, sexual content and hate leave measurable refusal at every rank from 1 to 12. We report the method, the 41-89 percentage-point reductions it achieves across seven harmful benchmarks with no detected change in capability, and the boundary where it stops.

Large Multimodal Models (LMMs) large-scale deployment in industrial warehouse settings specifically necessitates that models exhibit human-expert-level hazard-oriented perception, understanding, and reasoning capabilities. However, the scarcity of real industrial data, tightly coupled to commercial terms, significantly hampers further advancement. To bridge this gap, we curate LogiScope-VQA to investigate the practical applicability of mainstream LMMs in real-world logistics operations. LogiScope-VQA comprises 2,476 images and 2,918 videos primarily sourced from real-world logistics parks, along with 10,274 VQAs meticulously curated and validated by human annotators. Grounded in 18 core objects and 20 risk types, we devise 39 subtasks aligned with three principal themes: industrial element perception, warehouse knowledge understanding, and potential risk reasoning. Furthermore, we incorporate dynamic thinking-budget configurations and dual-dimensional risk bias analyses to elucidate the properties of LMMs. Extensive experiments unveil that even powerful proprietary models, including GPT-5.5, Gemini-3.1-Pro, and Claude-Opus-4.7, exhibit a significant gap relative to human performance. The unique challenge of jointly integrating perception, understanding, and reasoning for hazard identification poses substantial headroom for further improvement on LogiScope-VQA. We additionally reveal the pervasive security bias issue that impedes LLMs' practical deployment in real-world settings. The industrial dataset is publicly available under the CC BY-NC-SA 4.0 license.

Organizational design in the era of artificial intelligence requires experimental methods that can test how human-AI groups coordinate, delegate, and make decisions. Programmable platforms coordinate live human-to-human sessions or real-time human-AI chat, but researchers cannot easily declare experiment protocols in which AI participants both communicate and act on shared work within one auditable configuration. Here we introduce Pairit, an online platform that facilitates the design, testing, and deployment of experiments that test human-AI organizational designs and interventions. Through a single YAML configuration file, researchers declare an executable experiment graph (pages, routing, randomization, matchmaking, chat, shared workspaces, server-hosted agents, surveys, timers, and custom HTML components) and combine any number of humans and AI agents in live sessions. We have validated the feasibility of the platform through multiple live deployments, including peer-reviewed published studies, capturing high-resolution process traces of communication, negotiation, and collaborative work in live human-AI dyads. By representing complex interactive protocols as standardized, auditable configuration files, Pairit provides reusable infrastructure for specifying, deploying, and sharing live human-AI organizational experiments.

Asynchronous reinforcement learning has become the standard way to scale training for language models, but the resulting policy lag biases the critic toward the stale behavior policy. Existing work on asynchronous LLM training corrects the actor and leaves this bias unaddressed, while the off-policy value correction of classical RL does not carry over to long-horizon agentic tasks, since a short correction horizon leaves the regression target free of the reward and a long one lets the product of importance ratios drift exponentially with the trajectory length. We propose BRACE, an anchored Bellman-residual correction for stale value models. BRACE bounds the correction horizon to a prefix of policy tokens and anchors a constant-weight Monte-Carlo tail beyond it, which separates policy correction from reward propagation. BRACE improves mean@1 on BrowseComp-Plus by $2.4\%$ over the strongest baseline, runs $2.46\times$ faster per step than synchronous training, and remains stable $50$ updates off-policy.

Long-term language-model agents rely on external memory across interactions. Atomic memories are particularly useful: their fine-grained semantic boundaries enable precise retrieval and direct comparison between observations. Yet accumulating atoms inevitably become redundant, overlapping, or conflicting. Existing methods often ask an LLM manager to add, update, delete, or rewrite memories directly, coupling semantic interpretation, storage decisions, and content generation in one error-prone operation. We introduce ROAM, a relation-guided framework that uses atomicity for management while allowing richer answer-time representations. ROAM classifies incoming--stored atom pairs as independent, equivalent, directionally subsuming, or conflicting, then organizes observations into active Primary and supporting Evidence roles. Fusion subsequently combines complementary details and temporal changes into compact, potentially non-atomic views. Only Primary views are retrieved for answering, preventing redundant or outdated atoms from competing independently. Across models and evaluation settings, ROAM improves answer accuracy by up to 29.8 percentage points. Ablations show complementary benefits from different relations and consistent gains from fusion beyond role organization. Mechanism analysis further finds 15.6-point higher answer-critical source recall and an 11.5-point lower confounder-token share. ROAM remains robust across manager scales.

Frontier gains in language-model reasoning come from reinforcement learning on reasoning traces and are concentrated in domains with a cheap, sound verifier. We argue the field's binding constraint is the verification gap: no scalable, incorruptible reward for reasoning outside formal domains. We make four contributions. (1) Theory: in a joint-Gaussian model of best-of-N selection, verifier-gold correlation rho is the exact exchange rate between test-time compute and capability, and an unsound verifier pays a polynomial penalty N^(1/rho^2); a margin-free copula form predicts realized soundness of real LLM judges to 4% median error. (2) Demonstration: in program-synthesis testbeds with executable ground truth, including a pre-registered scaled replication, unsound verifiers lose Soundness-under-Pressure as optimization grows (0.94 to 0.32 at N=4096) while a sound verifier improves monotonically; reality-anchored settlement beats a frozen verifier under i.i.d. and adversarial pressure, driving the hacking gap from ~0.27 to ~0; soundness scales log-linearly with settled labels, with on-policy settlement ~10x more label-efficient than random labeling. With real LLM judges and unit-test execution as gold, a weak judge loses soundness under best-of-N (p<0.001), a stronger judge is more robust, and selection alone manufactures +0.53 hacking gaps from honest samples. Under real GRPO training, a frozen reward model traces the full overoptimization curve (executed reward collapses 90%) while the same model refit on a 10% settlement stream preserves 6x the executed reward. (3) Paradigm: proof-carrying cognition, where reasoning steps are typed probabilistic claims priced by a self-built world model trained only on held-out reality and settled by proper scoring rules. (4) Benchmark: we specify Soundness-under-Pressure as the headline metric for a reality-settled reasoning benchmark.

Procedural memory lets language agents reuse successful routines, but reuse presumes that a stored routine remains applicable. We study what happens when that presumption is deliberately violated. The study combines a retrospective, human-assisted interface-adaptation case from BrowserGym TimeWarp with controlled frozen-memory comparisons on synthetic shopping decisions. During the documented WebShop V1-V6 development path, interface-specific code was adapted while the separately stored high-level procedure was not reported to change; this phase does not constitute an autonomous memory-agent evaluation. In the controlled phase, an early pilot produced one task on which two memory conditions selected a more expensive item while the no-memory condition selected the reference minimum. Follow-up probes did not establish a recurring row-order or identity-binding pattern. We then tested four forms of mismatch: changed quantities, a different evidence representation, a conflict between local and global optimization, and distributed promotion evidence, across 32 formal cells. Each cell used one temperature-0 generation with the same local qwen3:8b configuration and no adaptive retry. Across these pairs, none of the predefined diagnostic interference signatures appeared on the tasks for which they were defined when current-task evidence was explicit and sufficient. The result identifies a tested region of non-interference: a procedural memory can be mismatched without becoming behaviorally disruptive. It does not establish general safety or a mechanism. The remaining question is which additional conditions turn applicability mismatch into observable, memory-caused error.

Quantum emitters in two-dimensional materials, particularly hBN, are promising platforms for quantum technologies. However, achieving high-density emitters at predetermined locations while preserving optical quality remains challenging. Here, we introduce a facile, cost-effective double-layer all-dry transfer approach to deterministically create overlap regions between hBN flakes. These pre-defined capped regions exhibit a significantly enhanced emitter density, with up to a 15-fold increase compared to uncapped areas. Importantly, this method does not compromise emitter quality: emitters within overlap regions demonstrate excellent optical performance, including high signal-to-background and signal-to-noise ratios, large Debye-Waller factors, high brightness, and strong spectral stability. Possible defect configurations are also discussed to contextualize the observed emission characteristics. This scalable strategy enables preferential formation of quantum emitters in targeted regions, achieving higher densities than simple treatments such as plasma irradiation while avoiding the complexity of advanced fabrication techniques. The approach provides a practical pathway for integrating high-quality quantum emitters into scalable quantum photonic platforms.

In Retrieval-Augmented Generation (RAG) systems, a large number of retrieved chunks are concatenated to form the input context so that users can receive high-quality responses based on external knowledge. As a result, the input context length increases substantially, leading to a larger prefill workload and, in turn, a longer time to first token (TTFT). While previous works that reuse precomputed key-value (KV) caches effectively reduce TTFT for long-context inputs, it remains unclear whether response quality is preserved when the input context becomes very long. In this paper, we propose a combined approach that (i) fine-tunes the model while taking KV cache concatenation into account and (ii) selectively recomputes a subset of the KV caches. By applying both techniques, we demonstrate improved accuracy for long-context inputs. Experiments on the RULER benchmark show that, for a 124k-token input, our method improves the RULER score by 9.7 point over the baseline that recomputes KV caches only. Moreover, TTFT is reduced by 80% compared with full attention.

Churn models typically identify high-risk customers but do not specify which feasible retention action should be considered or why that action is appropriate. We present CARRE (Counterfactual Action Retrieval and Reason Evaluation), a three-stage framework that combines retrieval-augmented candidate generation, cost-aware counterfactual scoring, and large language model (LLM) reasoning. CARRE retrieves a predefined catalog of retention actions, estimates model-predicted churn-risk changes under explicit feature transformations, and generates a structured churn reason and a profile-grounded explanation for the selected action. On the IBM Telco Customer Churn dataset, CARRE achieves 79.8% greater mean model-predicted risk reduction than the plain SHAP baseline and 80.4% greater reduction than the cost-controlled SHAP+Cost baseline across 313 high-risk test cases; its cost-normalized efficiency is 10.5% higher than that of plain SHAP. On a 136-case reason-stratified evaluation sample, diagnosis-driven prompt refinement increases weak-label agreement from 79.4% to 90.4%, with no auxiliary-plan constraint violations; because the same sample was used for error diagnosis and re-evaluation, the post-refinement result is not an independent estimate of generalization. For 135 explanations generated using the pre-refinement v2 reason outputs, two cross-vendor LLM judges assign mean scores ranging from 4.02 to 5.00 out of 5, although one judge saturates on actionability, and a deterministic audit finds no contradictions among 66 verifiable profile claims. Retrieval ablations show that k=5 provides the best evaluated compromise between high candidate coverage and downstream reasoning agreement in this dataset. These results illustrate how retrieval, model-based counterfactual scoring, and language generation can be separated and jointly evaluated in a prototype churn-prescription pipeline.

Social intelligence enables agents to read social context, infer intent, and adapt over sustained dialogue. As language models become autonomous collaborators, it is central to building effective and trustworthy human-AI interaction. Existing reinforcement learning methods optimize single-turn utterances and sparse outcome rewards, producing short-sighted policies that struggle to manage goal-relationship tensions across multi-turn interactions. We propose SocialRL, a multi-turn reinforcement learning framework addressing both challenges. First, we apply multi-turn reinforcement learning using PPO that propagates delayed outcome rewards back to each turn, enabling long-horizon planning. Second, we design six process reward dimensions capturing the goal-relationship trade-off, including goal advancement, relational attunement, contextual coherence, etc. A reward model dynamically generates fine-grained scoring criteria for each dimension, while a stage-aware weight schedule prioritizes relationship-building in early turns, goal advancement mid-way, and balanced closure late. Across multiple social-dialogue benchmarks, SocialRL improves Goal Achievement by an average of 9.2 percentage points over the corresponding Base models. These results demonstrate the effectiveness of SocialRL across synthetic and real social scenes, as well as standard and challenging social scenarios.

Real-time MRI (rtMRI) captures the dynamics of the entire vocal tract during speech, but labeled data are scarce and the modality - single-slice, grayscale, low-resolution - differs substantially from the natural videos that video foundation models are trained on. We introduce Arti-JEPA, a joint embedding predictive architecture to model vocal tract rtMRI by continuing its self-supervised objective on about 62h of unlabelled vocal-tract videos, and evaluate the frozen representation on three tasks: cross-domain phoneme prediction (on typical speakers), fluent-vs-disfluent classification (a corpus containing stuttered speech), and characterizing pre/post-operative transfer (after partial glossectomy). Three key findings emerge. (1) A temporal video prior decisively outperforms per-frame image encoders, and latent prediction (V-JEPA) is at least as strong as pixel reconstruction (VideoMAE), with the edge on fine-grained phonemes. (2) Domain adaptation is \emph{task-dependent}: it roughly doubles cross-domain phoneme prediction $κ$ (to 0.352) but does not help binary stuttering classification. (3) Arti-JEPA was able to recover phoneme signal from pre/post glossectomy speech --- an in-domain probe decodes patients at least as well as a typical speaker, indicating that the residual transfer gap is cross-speaker/domain misalignment, not surgical signal loss, and post-operative decoding does not fall below performance on pre-operative speech. Together, these position a frozen, domain-adapted rtMRI encoder as a reusable measurement tool for articulatory and clinical speech science.

As large language models are increasingly deployed as tool-augmented legal agents, they introduce agentic hallucinations where tool-call and reasoning errors cascade into fabricated holdings and miscited authority. However, existing legal benchmarks evaluate only single-turn QA with outcome-level metrics, while agentic hallucination benchmarks lack legal-specific diagnostic capability. Neither answers to what extent and how a legal agent hallucinates along its trajectory. To address these limitations, we introduce LexAgentHallu, a legal agentic hallucination benchmark designed to evaluate to what extent and how legal agents fail along multi-step trajectories. Built through a four-stage expert-in-the-loop pipeline, LexAgentHallu contains 3414 instances across 17 legal categories and 6 task types. Each instance is annotated under a dual-layer hallucination taxonomy of 7 high-level categories and 27 fine-grained subclasses, covering both substantive errors and agent-procedural failures. We further design fine-grained metrics that quantify to what extent and localize how each failure occurs along an agent's execution path. Our evaluation across 18 proprietary and open-source agents uncovers a Right-Answer-Wrong-Reason effect and reveals that hallucination subclasses cluster rather than scatter, forming distinct agentic framework, legal task, and category profiles. These findings, invisible to outcome-level evaluation, validate the diagnostic power of LexAgentHallu for evaluating agentic hallucination in law.

Automatic Guided Vehicles (AGVs) substantially boost warehouse throughput, but routing large-scale AGV fleets remains challenging. Classical Multi-Agent Pathfinding solvers suffer from exploding combinatorial complexity and super-quadratic runtime, while relying on idealized grid or piecewise-linear motion models that mismatch real-world kinematics. Recent Reinforcement Learning (RL) solutions improve flexibility via decentralized agent policies but depend on discretized spatiotemporal representations, require millions of episodes to converge, and incur full-map observation at every step, which leads to large models, slow convergence, and high inference latency that violates real-time industrial control constraints. To address these bottlenecks, we propose HiRAD, a hierarchical RL framework for continuous-space AGV routing with real-time guarantees: (1) a step-level spatiotemporal representation that translates continuous motion into a differentiable RL problem, (2) a hierarchical strategy that splits heading choice from velocity control to reduce the action space, and (3) an asynchronous event-driven decision pipeline that lowers inference complexity from O(n^2) to O(n) and cuts per-step latency by as much as 71 percent. Across random graphs and two warehouse maps, HiRAD reduces makespan by 45 percent to 63 percent and shortens end-to-end runtime.

We present the PccDiffuser, a conditional diffusion framework for continuum robots that learns a multimodal distribution over complete configuration-space paths and samples multiple candidate solutions in parallel, which are subsequently converted into an executable trajectory by time allocation considering actuator constraints. Under the piecewise constant-curvature model, we use exponential co-ordinates to describe the robot kinematics, and use graph neural network to encode a variable number of environment obstacles. Analytical differential kinematics is incorporated in the denoising process to improve terminal accuracy and whole-body clearance. On a mixed test set comprising workspace with zero to four obstacles, PccDiffuser achieved a success rate of 91\%. Compared with existing sampling- and optimisation-based benchmarks, it delivered both a higher success rate and greater computational efficiency, with the latter advantage becoming more substantial when sampling more candidate solutions. Experiments on a three-section tendon-driven continuum robot further demonstrate consecutive planning, multi-solution planning, and whole-body obstacle avoidance.

Test-Time Adaptation (TTA) enhances the robustness of models against distribution shifts but faces two critical challenges: error accumulation from noisy pseudo-labels and catastrophic forgetting of source knowledge. Uncertainty-based approaches designed to mitigate error accumulation often yield overconfident or computationally expensive estimates, while strategies intended to prevent forgetting via prototype replay rely on static representations that easily become misaligned as the model adapts. To address these issues, this paper proposes a novel framework, Distilling Image Prototype for Guided Test-Time Adaptation (DIPTTA). The core of the proposed approach is the introduction of a Distill Image Prototype (DIP), a compact set of synthetic images that serves as a dynamic and regenerative anchor of source knowledge. This prototype enables a dynamic feature replay mechanism that continuously generates feature prototypes aligned with the current state of the model, thus effectively preventing catastrophic forgetting. Furthermore, the DIP anchors a source-calibrated uncertainty estimation method, which provides a less biased measure of sample reliability by leveraging stable source knowledge, thereby robustly suppressing error accumulation. Extensive experiments on multiple benchmarks demonstrate that DIPTTA significantly outperforms state-of-the-art methods, particularly under severe domain shifts. The source code is available at https://github.com/LiwenWang919/DIPTTA.

Healthcare systems, mental health, and public well-being are increasingly affected by cyberbullying and harmful online interactions. This paper presents CareGuard, an early-warning framework designed to support healthcare-driven mental health protection and proactive online safety through the detection of cyberbullying-related content using advanced natural language processing techniques. CareGuard integrates zero-shot semantic labeling with fine-tuned transformer-based models, including BERT, DistilBERT, and RoBERTa, to enable robust and context-aware classification across sensitive cyberbullying categories. To improve efficiency and reduce unnecessary computation in healthcare-oriented monitoring settings, the framework incorporates an emotion-aware filtering mechanism alongside cosine similarity-based semantic screening, allowing the system to focus on semantically relevant and emotionally salient content. Experimental results on benchmark datasets demonstrate that CareGuard effectively balances detection accuracy and computational efficiency, highlighting its potential for scalable deployment in healthcare systems, mental health monitoring, and online safety applications.

Source-level analysis of interictal epileptiform discharges (IEDs) is relevant to presurgical evaluation and treatment planning because it helps characterize where epileptiform activity is likely to arise. Beyond detecting whether an IED is present, this setting requires assigning IED-positive activity to clinically meaningful brain-region categories. This setting is challenging because source-region evidence in short electroencephalography (EEG) windows can be subtle, partial, and affected by subject variability, class imbalance, and imperfect multimodal context. We present EEGBind, an EEG-centric multimodal binding framework for five-class source-level IED classification. EEGBind treats EEG as the primary modality and binds synchronized video-context features around an EEG-centric representation. Instead of relying on early or overly strong multimodal fusion, which may perturb the source-sensitive EEG representation, EEGBind uses video context as auxiliary evidence for robust classification. A view-consistent repair stage is further used to improve hidden-set robustness while preserving the learned source-class boundary. On the NeuroMM 2026 Grand Challenge Track 3 NMM-Source-IED benchmark, EEGBind achieves 0.8395 on weighted-F1 and outperforms strong competitors. These results support EEG-centric multimodal binding as a practical strategy for source-level IED classification. The open-source code is available at https://github.com/HKUSTGZ-ML4Health-Lab/NeuroMM2026_IED_Detection.

Supervised fine-tuning (SFT) applies a uniform cross-entropy loss to all target tokens, even though different tokens provide unequal learning signals for mathematical reasoning. This uniform treatment can over-sharpen already mastered tokens while amplifying learning pressure on uncertain, low-confidence tokens, leading to suboptimal training dynamics. We propose Trimmed Logit-Gap SFT (TrimSFT), a simple token-level reweighting method that scales the SFT loss according to the logit gap between the gold token and its strongest competitor. TrimSFT trims supervision away from both extremes: tokens already mastered (large logit gap) and tokens weakly supported by the current model (small or negative logit gap), concentrating learning within an intermediate logit-gap region between them. We instantiate this principle with a Gaussian weight centered at margin m with bandwidth τ, requiring no reference model or additional forward pass. We evaluate TrimSFT on six base models from the Llama, Qwen, and DeepMath families across five mathematical reasoning benchmarks. TrimSFT consistently improves over standard SFT, achieving the best average performance on five out of six models, with gains of up to +26.9 points over SFT on MATH500. Further analyses show that the bandwidth τ matters more than the exact margin location, and that half-trim variants that remove supervision pressure from only one side yield inferior trade-offs. A token-level logit-gap distribution analysis suggests that TrimSFT reshapes model confidence in a more balanced way than uniform SFT or monotonic reweighting methods. These results suggest that reasoning SFT can benefit from trimming both extremes rather than treating all tokens uniformly.

Generalizable animal Re-Identification (ReID) aims to recognize individual animals across species with diverse morphologies and ecological contexts. Unlike person ReID, where different domains share similar body structures, animal species often exhibit drastically different anatomical structures and visual patterns, making it difficult to establish shared visual correspondences. As a result, representations learned across species tend to form fragmented embedding spaces, which severely limits cross-species generalization. To address this challenge, we propose Semantic Consistency Learning (SCL), a framework designed to learn representations that remain stable across appearance variations while preserving semantic structures shared across species. SCL consists of two complementary components. Foreground-Background Decoupled Spectral Normalization (FDSNorm) stabilizes feature statistics by suppressing environment-induced style variations in a region-aware manner, while Cross-species Neighborhood Modeling (CNM) captures transferable relational structures across species through dynamic feature neighborhoods. Extensive experiments on 11 public animal ReID datasets demonstrate that SCL consistently outperforms state-of-the-art methods under multiple cross-species evaluation protocols and generalizes effectively to previously unseen species and ecological domains. Code is available at https://github.com/Kemalau/ECCV-26-SCL.

When training data are limited, increasing parameter count is not the only way to improve language-model performance. A small parameter set, when repeatedly applied, can also deliver comparable performance. We study Looped GPT-BERT in the BabyLM 2026 Strict-small setting, combining GPT-BERT's masked next-token and causal language-modeling objectives with depth-wise parameter sharing. We train on a preprocessed 7.48M-word English corpus and compare objective ratios, non-looped and looped architectures, and loop counts. Our final $4\times12$ model uses four physical layers for twelve recurrent traversals and contains 12.18M parameters. The BabyLM 2026 leaderboard reports an Overall Average of 35.42 and an NLP Average of 48.48. Compared with public BabyLM 10M Strict-small GPT-2 and GPT-BERT baselines, it achieves comparable performance on selected linguistic and downstream metrics, including BLiMP and GLUE, with fewer parameters. The loop ablations show that additional recurrent computation can improve training and preserve strong performance on selected linguistic tasks, whereas poorer performance on other tasks may reveal an inherent limitation of the looped design: using only a few physical layers restricts the model's representational space.

Clinical diagnosis agents must decide not only what test to request next, but also when to diagnose or defer. Existing agent benchmarks largely evaluate accuracy after fixed or unconstrained interaction, leaving autonomous stopping reliability implicit. We present Cros, a risk-constrained stopping layer combining state-wise error ranking, policy design on disjoint development splits, and LTT-style exact tests of selective diagnostic error and minimum autonomous coverage for complete sequential policies. Its finite-sample guarantee requires the candidate family, testing rule, and any randomization to be frozen before calibration labels are accessed. On a 1,834-episode MIMIC-derived abdominal-pain benchmark, the full ranker achieves exploratory state-error AUROC 0.853, compared with 0.715 for maximum class probability and 0.552 for the backbone's native stop score. On the previously viewed 367-episode evaluation split, analytically averaging over the frozen Cros weights yields 16.9% selective error at 78.8% coverage, cost 5.57, and 0.68 tests, versus 30.8% error at 100% coverage, cost 8.14, and 1.53 tests under native stopping. Forced continuation is non-monotone: error is 28.3% with HPI alone and 34.3% after full workup. However, the uniform-weight mixture ablation is cheaper on this viewed split despite missing the locked development margins, and Cros nominally satisfies the joint criterion in only 6 of 20 development resplits. Because evaluation labels were inspected during earlier development, these findings provide exploratory feasibility and audit evidence, not a confirmatory safety certificate.

The rapid advancement of audio and multimodal large language models has unlocked transformative speech understanding capabilities, yet evaluation frameworks remain predominantly English-centric, leaving Southeast Asian (SEA) languages critically underrepresented. We introduce SEA-SpeechBench, to the best of our knowledge, the first large-scale multitask benchmark that evaluates speech understanding in 11 SEA languages through 97,194 samples across 99 evaluation sets and 597 hours of curated audio data. Our benchmark comprises 9 diverse tasks across 3 categories: speech processing (automatic speech recognition, speech translation, spoken question answering), paralinguistic analysis (emotion, gender, age, speaker recognition), and temporal understanding, a novel dimension featuring timestamped content queries and temporal localization within extended audio sequences up to 3 minutes. We implement multilingual prompting in both native SEA languages and English to reflect user interactions with audio-language models. Evaluation of leading open-source and proprietary systems reveals marked performance gaps. Across all models, performance remains underwhelming on temporal understanding, emotion recognition, and speech translation. Prompting in low-resource languages such as Burmese and Tamil lags behind English by up to 41 percentage points. Our findings expose critical model limitations and underscore the need for inclusive model development. The SEA-SpeechBench benchmark is available at https://zwenyu.github.io/SEA-SpeechBench/.

When an agent writes code, the development framework becomes the control system for a non-deterministic worker. Spec-first, agent-driven frameworks have gained rapid traction since 2025; the installable ones, GitHub Spec Kit, obra/superpowers, BMAD, and GSD, and our own, all capture intent through a specification or durable planning artifacts. Since they agree on capturing intent up front, what separates them is how each enforces the engineering discipline that keeps agent-written code clean, correct, and maintainable. Every framework enforces that discipline somehow; they differ in how. We characterize three modes: enforcement by persuasion (prompt discipline the model may ignore), by front-loaded structure (strong specs, then a trusted build), and through controls the agent cannot edit (a deterministic orchestrator, human-approved gates, immutable tests, and a green result that must pass against a live, branched database). We introduce Consort, a spec-first, test-driven agent framework built on the third, enforcing that discipline through controls the agent runs inside but cannot bypass, in which a deterministic orchestrator drives separate role agents through a spec-first design lane and a test-driven build lane on a live database branch. We argue that enforcing the tests and gates in code keeps agent-written code honest and verifiable, while its specialized roles, like the human roles before them, are what make it maintainable, claims we frame as a pre-registered, testable hypothesis.

Large language models (LLMs) are increasingly deployed as personalized assistants that interact with users over extended periods of time. As conversations grow longer, relying on full interaction histories becomes increasingly inefficient and unreliable: long contexts introduce substantial computational overhead, making it difficult for models to consistently identify and utilize the most relevant information for the current request. These challenges have motivated memory systems that structure and retrieve user-specific information. In realistic interactions, users often seek practical guidance such as recommendations, planning, and decision support. Unlike factual recall tasks, personalized guidance requires models to integrate information across multiple past conversations and reason about changing user preferences and experiences. However, existing conversational memory evaluations mainly focus on retrieval and factual recall. To study this challenge, we introduce PRAGMA, a benchmark for evaluating personalized guidance in long-term conversations. PRGAMA contains curated longitudinal conversation histories, evidence annotations, and guidance scenarios grounded in evolving user contexts and incorrect user assumptions. Experiments across retrieval systems, memory systems, and long-context models reveal that current systems struggle both to recover the appropriate conversational evidence and to effectively use it for personalized guidance. Our results highlight the need for memory architectures that support robust conversational retrieval and memory-grounded reasoning beyond evidence recall.

Federated learning shares model updates rather than raw data, yet these updates can be inverted to reconstruct the clients' training data. Analytic reconstruction attacks, which invert a gradient in closed form, degrade as the batch grows: prior single-round attacks recover only about half of a batch of size $100$ even when the attacker fully controls the network parameters, and known upper bounds limit what any such method can recover. We establish a connection between gradient inversion and the theory of erasure-correcting codes, and use it to construct attacks that exceed these bounds. Our attacks recover batches exactly, together with every sample's label, from a single FedSGD round, and certify each recovery without ground-truth data. On eight image and tabular benchmarks they outperform prior single-round attacks by a wide margin. Even a passive attacker who only observes an honestly trained network recovers $94$--$100\%$ of ImageNet batches at sizes up to $128$, more than prior single-round attacks achieve even with active manipulation of the model, and in the active setting more than $90\%$ is recovered at batch sizes of several hundred. These results show that the privacy leakage of federated learning has been underestimated.

Existing emotional support conversation systems mainly focus on one-on-one seeker-supporter interactions and individual emotional states, leaving interpersonal relations in multi-party scenarios underexplored. In this work, we introduce relation-aware emotional support conversation, a new task that evaluates whether LLMs can capture and utilize the evolving dynamics of relationships to offer more effective emotional support. We construct RESCUE (Relation-aware Emotional Support Conversation Understanding and Evaluation Benchmark) from real couple and family interview conversations, containing 191 samples, 7,079 annotated turns, and 1,064.8 minutes of video. Based on rich annotations of socio-emotional and support-related dynamics, RESCUE defines six tasks that evaluate two core capabilities required for relation-aware emotional support: Relational Understanding and Relation-Sensitive Support. Experiments with ten LLMs show that current models perform relatively well on tasks relying on local emotional or intervention cues, but struggle with relation-intensive tasks such as relation pattern prediction, viewpoint prediction, and support strategy prediction. These findings reveal the limitations of current LLMs in modeling interpersonal relations and making relation-sensitive support decisions.

Agentic systems are rapidly moving to production, where they read untrusted inputs, call tools with real permissions, and act autonomously, expanding the security surface beyond chat-only models. Yet standard evaluations remain single-turn and fail to capture multi-step agent vulnerabilities. We present a systematic black-box framework for risk-aware agent evaluation requiring only basic system descriptions. Our approach introduces: (1) a seven-domain taxonomy mapping observable behaviors to risk categories, (2) fully automated SAGE-RT red teaming producing 120 adversarial scenarios per domain, and (3) human-validated evaluation using LLM judges. Empirical validation across two agent architectures (CrewAI and AutoGen) with four base models reveals alarming patterns: 56.25\% average governance risk, 65\% privacy risk in multi-agent configurations, and agent behavior vulnerabilities reaching 85\%. Our black-box approach effectively identifies critical architectural vulnerabilities without privileged access, providing a scalable path toward safer agent deployments.

Semantic-gradient-based prompt optimization (SGPO) improves agent harnesses using execution feedback, but its local update rule leaves the choice of edit scope and operation unresolved. We introduce RobustSGPO, which specifies the requested edit, constructs and checks the patch, and continues search from either the incumbent or retained snapshots. We evaluate permission scheduling, cumulative controls, and task-family transfer in the AgentX brainstorming workflow using 120 tasks, 95 runs, and 7,350 candidate attempts. Periodic $1\to2\to3$ scheduling exceeds fixed maximum permission by 0.28 test-score points. RobustSGPO increases completion on 30 held-out tasks from 60.0% to 80.0% and improves test quality from 3.77 to 4.14 under a 20-million-token budget. Category retention reduces source-task degradation after a shift, whereas random retention reaches a higher destination endpoint. Search-space control benefits quality through executable edits and alternative starting points, with measurable retention overhead.

Standard behavior cloning supervises actions without explicitly constraining the future representation paired with each demonstrated action chunk. We introduce JEPA Policy, a diffusion-free framework that uses the action chunk and its observed future representation as paired training targets. Action and future-representation tokens interact in a shared Transformer and are refined through two forward passes. Future prediction can therefore shape the representation used to generate actions. Dual-branch and gradient-routing controls attribute the gain to this shared topology rather than to an auxiliary prediction head alone. Across nine simulated tasks, JEPA Policy improves mean success over the action-only MIP baseline and outperforms Diffusion Policy under the evaluated configurations, while adding 0.29 ms to MIP's model latency. A five-task, 630-episode physical-robot study produces the same pooled ranking. Further audits find no complete representation collapse under action supervision and identify a task-conditioned failure-ranking signal in future-prediction error. These results support paired future-representation supervision as a practical approach to low-latency visuomotor imitation without iterative generative sampling.

Inferring speaker relationships from spoken conversations is an important step towards socially aware speech understanding. However, this task remains underexplored, and supervised modeling is costly to train and scale. At the same time, existing inference-time LLM approaches provide limited structure for handling subtle, distributed, and multimodal relational cues that may support multiple plausible interpretations. To address these limitations, we introduce a training-free multi-agent reasoning framework that organizes inference through structured interaction among LLM agents, allowing relationship judgments to be proposed, challenged, and adjudicated without task-specific training. We instantiate this framework with two complementary designs. We propose Multi-Role Multi-Agent Debate as a task-specific adaptation of standard multi-agent debate for speaker relationship inference, assigning agents complementary roles or social-theory-grounded perspectives rather than a single undifferentiated viewpoint. In contrast, we introduce Multi-Agent Compete, a competition-based protocol that compares agent judgments through pairwise adjudication, eliminates weaker candidates, and retains the most defensible one. We evaluate these methods on the Seamless Interaction dataset across different modality settings, covering both binary classification and fine-grained relationship-detail prediction. Results suggest that they improve over zero-shot and existing multi-agent baselines in most cases. Human evaluation further suggests that this task is challenging even for people. LLM methods can sometimes outperform human annotators in text-included settings but are less competitive in the audio setting. Together, these findings suggest that relationship inference benefits from structured inference-time interaction among agents, while acoustic cues are not yet fully captured by current models.

Capabilities relevant to Physical AI can arise from materially different formation histories, yet existing taxonomies organized by morphology, architecture, learning algorithm, task, or domain do not directly answer what gives rise to a capability. We define a capability-formation source as a factor materially contributing to capability formation, distinct from components or construction steps. We identify seven non-exclusive sources: Recorded-Experience (RE), Predictive-Modeling (PM), Evaluative-Interaction (EI), Surrogate-Environment (SE), Mechanism-Grounded (MG), Embodied-Coupling (EC), and Evolution-Driven (ED) Formation. Using reconstructive induction with theoretical saturation, we traced a research matrix to primary studies, deduplicated the literature, set coding rules, and conducted three rounds of maximum-difference and negative-case sampling. Challenges included curriculum and self-supervised learning, active inference, open-ended and developmental learning, planning and search, neuro-symbolic architectures, digital twins, generative physical world models, and morphology-control co-design. Within the scope and criteria fixed as of September 4, 2026, all 49 evidence records were explainable by the seven sources individually or in combination. No R1-R3 challenge produced an irreducible eighth source, and R3 required no new core definition or substantive boundary rule. We therefore claim theoretical saturation within the stated scope, not logical completeness or exhaustive future coverage. The framework distinguishes similarity in observed capability from similarity in how it was formed, supporting analysis of explanation, transfer, replication, dependencies, governance evidence, and geoeconomic foundations.

Open-world object detection (OWOD) extends closed-set detection by requiring models to identify unknown objects and incrementally learn them once annotations become available. In remote sensing imagery, object categories often exhibit latent hierarchical relationships that may be inadequately represented in the Euclidean spaces commonly adopted by existing methods, limiting unknown-object recall and incremental-learning performance. To address this issue, we investigate hyperbolic geometry for OWOD in remote sensing imagery and propose HyRS-OWOD. To improve unknown object recall, we design a two-step unknown-object discovery mechanism: a Decoupled Objectness Learning (DOL) module that disentangles foreground perception from semantic information to separate foreground proposals from background regions, followed by a Hyperbolic Uncertainty Learning (HUL) component that leverages the radius of hyperbolic embeddings as an uncertainty-aware cue for known-unknown discrimination. For incremental learning, we develop a Hyperbolic Metric Learning (HML) strategy that enhances inter-class separability, facilitating the incorporation of novel categories while mitigating catastrophic forgetting. Experiments on three remote sensing benchmarks demonstrate consistent improvements in unknown recall and incremental learning over state-of-the-art OWOD methods.

As Digital Twin (DT) systems evolve beyond state synchronization toward task-oriented and knowledge-driven operation, Cognitive Digital Twins (CDTs) have emerged as an extension that incorporates cognitive capabilities into twin operation. Existing CDT studies often focus on specific enabling techniques, such as learning modules, knowledge graphs, and large language models, while providing limited insight into how cognition can be systematically integrated into DT architectures. To address this issue, this paper proposes a four-layer CDT architecture consisting of the physical layer, digital-twin layer, cognitive layer, and task layer. The proposed architecture establishes a self-evolving closed operational loop spanning these four layers, in which physical states are synchronized into digital representations, cognition constructs task-specific cognitive models through knowledge, memory, and attention, and task-level decisions are generated under practical constraints. Operational feedback further refines cognitive experience and updates relationships and annotations in the digital representation, enabling subsequent task interpretation, initiation, and reasoning to evolve with system operation. Based on this framework, two representative operation modes are characterized: user-request-driven cognition and self-driven cognition. We further discuss key enabling mechanisms and deployment challenges associated with semantic communication, knowledge querying, task orchestration, and closed-loop synchronization. A lightweight simulation study illustrates reliable closed-loop task feasibility under limited semantic information and improved operational efficiency through accumulated task experience. The proposed framework provides a structured foundation for the design and development of future CDT systems.

Tendon transmissions reduce distal inertia and add compliance, yet routing, slack, and friction govern motion and force transfer. Mainstream rigid-body robotics simulators such as MuJoCo do not jointly resolve moving noncircular contact, unilateral tension, and segment friction. We present MuJoCable, which adds a reduced-order, configuration-dependent cable transmission to MuJoCo. Its routing algorithm jointly optimizes an ordered path across moving analytic and mesh surfaces. A unilateral axial law, directional Capstan propagation, and nodal virtual work map this path to segment tensions and body forces. The warm-started engine plugin applies these forces during simulation and exposes route and load states for design. Pulley benchmarks recover analytical transmission relations with a Capstan-ratio error below 0.5%. On the underactuated 18-joint SpiRobs, MuJoCable reveals friction-driven load growth and proximal redistribution of joint rotation that the native tendon does not represent. Hardware tests on SpiRobs and a tendon-route-coupled finger reproduce observed motion sequences. By making physical threading executable, MuJoCable brings transmission sources of the simulation-to-reality gap into route, cable, and actuator design before fabrication.

On August 2, 2026, the obligations of Article 50 of the EU AI Act took effect, requiring generative AI providers to mark the content their systems produce and ensure it can be detected as AI-generated. Days later, Anthropic disclosed that every Claude model released after that date embeds a watermark based on SynthID-Text in all generated text, enabled by default with no user opt-out; Google has deployed SynthID-Text in Gemini since 2024. Users objected that the watermark degrades quality, particularly for code, that it secretly encodes identifying information, and, in mutual contradiction, that it is easily removable and inescapable; the vendor answered with assurances of unchanged quality, no identifying information, and robustness to light editing. In this work, we argue that neither the objections nor the assurances can currently be verified and that this unverifiability, rather than watermarking itself, is the substantive governance failure. We sort the contested assertions by what it would take to settle each and evaluate the open-source SynthID-Text implementation on two open-weight models, because no public tool can test the deployed systems. On prose, the measured effect of the watermark does not exceed that of changing the sampling seed. On code, the cost is three points of correctness on one model and below measurement on the other, while detection remains near chance, a limitation of detectability rather than quality. The remaining gaps trace to withheld access or missing institutions and we map each to a requirement: release of matched outputs, configuration disclosure, accredited audits, a shared evaluation protocol, and interoperable detection.

Accurate contact prediction is useful for robotic manipulation only if it supports effective decisions. We investigate this connection using a compact, randomly initialized visuotactile world model, trajectory-level uncertainty calibration, and behavior-initialized actor-critic learning in imagination. On 160 MuJoCo Lift episodes, adding touch reduces endpoint-force prediction error from 1.058 to 0.228 N and interval-peak error from 2.724 to 0.523 N across three training seeds. However, tactile persistence achieves lower errors of 0.095 and 0.498 N, respectively. Two exploratory control rounds comprise 680 executions on 40 independent test initial conditions. A matched reward revision on fresh test environments increases in-distribution 10 cm lifting success from 20.0% to 93.3%, while success within an 8 N per-finger budget reaches only 33.3%, compared with 70.0% for force feedback. Calibration margins reduce force violations at the cost of task completion. In a separate study of public GelSight recordings, a force regressor achieves 0.04234 N error, but frame-level calibration covers only 15.80% of complete trajectories; trajectory-level calibration raises this to 87.36% at nominal 90% coverage. Together, these findings distinguish improvements in sensing and task reward from improvements in force-constrained control. The evidence is limited to public sensing records and simulator execution, without a demonstrated transfer between them.

Sixth-generation (6G) wireless networks are expected to provide a key infrastructure for large-scale embodied intelligence, where heterogeneous robots collaborate through low-latency connectivity, edge intelligence, and distributed sensing. Vision-language-action (VLA) models offer a foundation by integrating visual perception, language understanding, and action generation into a unified closed-loop policy. However, training and adapting VLA models to distributed robotic agents introduce challenges in privacy protection, communication efficiency, and model heterogeneity. Existing federated learning (FL) methods overlook the intrinsic differences among vision, language, and action pathways in parameter scale, privacy exposure, update dynamics, and tolerance to compression or perturbation. To address this issue, this article proposes FedMVLA, a modality-decoupled FL framework for privacy-preserving embodied intelligence in 6G networks. FedMVLA incorporates three mechanisms: modality-aware federated aggregation (MAFA), modality-aware privacy allocation (MAPA), and modality-aware communication compression (MACO), together with a modality-sliced transport design that routes the precision-critical action stream through a protected ultra-reliable low-latency slice. A case study on federated robotic manipulation over the Third Generation Partnership Project (3GPP)-based wireless substrate, covering fading, co-channel interference, and malicious jamming, shows that FedMVLA achieves an 84.8% task success rate, exceeds FedAvg by 22.2 percentage points, sustains a widening margin when scaling to 128 clients across eight cells, and reduces the schedule-averaged per-client uplink model-update payload by 95.6% (approximately 96%), while keeping the 95th percentile (p95) of the round-critical uplink completion time near 1.5s.

Urban planning is a real-world spatial optimization problem that requires selecting feasible actions from large candidate spaces under practical objectives such as cost and service quality. Existing optimization and reinforcement learning methods are effective for fixed formulations, but often depend on task-specific representations and constraint handling. We propose \emph{CityPlanner}, a sandbox-agent framework for executable urban planning. CityPlanner introduces \emph{UrbanSandbox}, a unified file-based environment where agents inspect task files, generate plans, run evaluators, and revise decisions based on executable feedback. To make learning tractable, we further propose atomic-task reinforcement learning, which decomposes long sandbox trajectories into \emph{BuildPlan} for initial construction and \emph{ImprovePlan} for feedback-based refinement. Experiments on a real-world benchmark show that CityPlanner consistently outperforms heuristic, task-specific RL, and general LLM-agent baselines. Ablations verify the contributions of UrbanSandbox, atomic-task RL, and iterative deployment. We release the code and dataset at https://anonymous.4open.science/r/co-agent-C1C8

Myocardial strain from echocardiography is a key biomarker for cardiac function. Recent deep learning methods show strong performance for myocardial motion tracking but often lack physiological constraints, leading to temporal drift across the cardiac cycle. Consequently, tracked points may not return to their relative initial positions at the end of each cardiac cycle, producing inaccurate strain estimates and even divergence in some cases. We propose a deep learning framework that compensates for drift during myocardial tracking. We extend a state-of-the-art echocardiographic tracking method (TAS-Net) with persistent memory tokens that share information across sliding windows over full cardiac cycles. A teacher-student fine-tuning strategy on real echocardiographic data then enforces physiologically consistent cyclic motion while preserving tracking accuracy. Experiments show reduced global and regional strain drift, improved agreement with clinical references, and better test-retest reproducibility, supporting more reliable myocardial strain estimation in clinical practice.

Topic models summarize large text corpora, but top-ranked words often provide only a limited representation of topic semantics. Sparse autoencoders (SAEs) offer a way to move beyond word-level descriptors by extracting interpretable features from dense representations, yet how feature interpretability relates to topic-inference quality remains unclear. We introduce \textbf{MonoTM}, an interpretable topic modeling framework that decouples these roles. Across three benchmark corpora, we show that document--topic mixture estimation and semantic interpretation favor different SAE configurations and feature subsets. MonoTM estimates mixtures from the full SAE bag-of-features representation and, with them fixed, learns topic descriptors over a separate vocabulary of corpus-grounded semantic features. This design preserves global topic structure while representing topics with semantic units more meaningful than individual words, making them more useful for downstream corpus analysis.

News articles often contain omitted temporal expressions, such as day-only or month-only mentions, which must be interpreted with reference to the publication date. When such articles are indexed or processed as standalone text in search and retrieval-augmented generation (RAG) systems, these omissions can cause temporal mismatches and unstable interpretation by large language models. We focus on reproducing omitted temporal expressions as concrete dates or intervals using the publication date as external context before the articles are indexed for search and RAG applications. Specifically, building on established temporal-expression extraction and normalization techniques and informed by a manual analysis of Japanese news articles, we propose jaROTE, a rule-based pipeline for Japanese news. Experiments on two news corpora demonstrate that jaROTE achieves high performance, and remains competitive with LLMs while providing a fast, low-cost pipeline. We further show that temporal reproduction improves time-constrained lexical retrieval, demonstrating the practical value of publication-date-grounded normalization for Japanese news retrieval.

Surveys remain the primary way researchers grasp the lineage of methods within an AI subfield, but they scale poorly against the current rate of publication. Existing taxonomy-induction methods are largely leaf-bound and time-agnostic; they tend to force transitional papers into mature leaves and can create topological inversions between ancestors and descendants. We propose EvoTree, a staged framework that decouples conceptual backbone learning from temporal refinement: a graph-aware encoder with distribution-based hierarchical clustering yields a stable taxonomy backbone; temporal fine-tuning then re-attaches marginal papers to internal nodes under monotonic-path constraints; a final LLM pass labels concepts without altering the topology. We release the first annotated benchmark for this task across 11 AI subfields. EvoTree attains the highest NMI and citation-direction accuracy among all baselines and the best concept purity on the annotated benchmark, and is the only method with non-trivial marginal-paper detection on the annotated set.

Recommender systems have become core infrastructure for modern online platforms, personalizing content at scale and strongly influencing what users see, click on, and purchase. However, this dependence on user interaction also exposes them to shilling attacks, where malicious actors can inject fake profiles to distort item rankings and control visibility. Existing attacks often rely on target-specific fine-tuning or fixed profile templates, making them either difficult to adapt to different victims or easier to detect. To overcome these limitations, we propose the Agentic Group Attack System (AGAS), a coordinated shilling framework where a central Coordinator directs a group of role-switching worker agents to adaptively promote a target item across different victim families. The Coordinator dynamically adjusts the strategy when progress stalls or suppression signals increase, while workers pursue a shared objective and switch between active and inactive roles to avoid repetitive patterns. Under the same attack budgets and evaluation protocols, AGAS consistently surpasses strong baselines in target promotion while better preserving benign recommendation quality, weakening representative detectors, and achieving higher efficiency than prior attacks. These findings also emphasize that defending recommender systems may require mechanisms that can handle adaptive shilling campaigns, not just isolated fake-profile injections. Our code is available at https://github.com/phkhanhtrinh23/AGAS.

Video large language models (Video-LLMs) are increasingly used as the perceptual front end of world models, a role that assumes they can read motion: how fast something moves, which way it travels, how hard it is pushed. We show they cannot. A Video-LLM can watch two clips of the same person in the same room, name every object in both, and still fail to say which clip moves faster. We introduce MotionBlind, a contrastive benchmark of self-recorded video for physically grounded motion(speed, magnitude, and direction), the variables a world model must predict. Each instance is a pair of near-identical clips that differ only in motion. Each clip carries two complementary yes/no questions, giving four items per instance, and a model earns credit only if all four are correct. We report Instance Accuracy(IAcc), which has a 6.25% chance floor. Single-frame, appearance, and language-only shortcuts all collapse to it. MotionBlind complements the recent TimeBlind benchmark. We run a controlled study of six open and two frontier Video-LLMs, varying whether the video is present, whether frames are shown in the correct temporal order, and how frames are sampled (1 to 24 frames, four selection strategies). Open models sit near the 6.25% floor, and scale does not help. Removing the video drops every model to zero IAcc, and shuffling frames collapses IAcc to chance, so the task genuinely needs video in order. Neither more frames nor smarter frame selection closes the gap, because these change which frames are seen, not whether motion is read. Only Gemini3.1 Pro clears the benchmark overall, and even it fails on speed. A frontend that cannot tell two speeds of the same action apart is not yet a trustworthy source of supervision, reward, or evaluation for a world model.

High-risk non-muscle-invasive bladder cancer (HR-NMIBC) carries substantial risks of recurrence and progression, while current clinical risk stratification remains limited. CHIMERA was established as a multimodal AI challenge to benchmark prediction in HR-NMIBC under standardized evaluation. Task BRS predicts RNA-seq-defined BCG Response Subtypes from histopathology and structured clinicopathological data, whereas Task Progression models time-to-progression using histopathology, structured data, and RNA sequencing. A multimodal dataset of 368 patients was divided into public training and hidden validation and test sets. In total, 159 submissions were made, and 13 top-performing models were selected for benchmarking. The best models achieved a weighted F1 score of 0.73 for Task BRS and a C-index of 0.68 for Task Progression. Post-challenge analyses revealed task-dependent modality contributions, cohort-dependent performance degradation, and sensitivity to missing structured data. In Task BRS, histopathology partly compensated for pathology-derived structured variables, whereas progression models showed greater dependence on complementary inputs. Cross-model error analysis further identified patients that were consistently difficult across different architectures, with T1 substage associated with prediction difficulty. These findings highlight barriers to transportability and the importance of missingness-aware modeling and independent multi-institutional validation. CHIMERA provides a standardized multimodal benchmark for bladder cancer and a framework for studying not only model performance, but also robustness, information sufficiency, and patient-level prediction failure.

Learned image matching has experienced significant progress in recent years, culminating in robust and accurate matchers such as RoMa, whose robustness is often attributed to its use of frozen DINO features. In a parallel development, feed-forward reconstruction models, such as VGGT, have been trained on ever-growing datasets to accurately regress dense 3D point maps and camera poses. The distinction between matchers and feed-forward reconstruction models has become increasingly blurred with the introduction of matching losses in models such as MASt3R and VGGT-$Ω$. This raises a natural question: what do feed-forward 3D models know about image matching? In this work, we answer this question by analyzing three scenarios: (i) zero-shot matching of patch features, (ii) direct matching of 3D point predictions, and (iii) training a full matcher on top of the learned representations. We find that, despite performing poorly in zero-shot matching, especially in later layers, feed-forward reconstruction models provide strong representations for linear probing and full matching pipelines. We further show that, even without any training, their raw predictions alone enable competitive matching, albeit only under moderate viewpoint changes and modality gaps. Based on these insights, we retrain RoMa v2 by replacing its DINO backbone with VGGT-$Ω$. Our resulting model, \ours, outperforms state-of-the-art matchers on a wide range of benchmarks, e.g. +8.1 mAA compared to RoMa v2 on WxBS.

Rapid bespoke commissioning of the Cognitive Digital Twin (CDT) is a major challenge in reconfigurable manufacturing. Traditional digital twin (DT) construction methods primarily focus on geometric reconstruction, often neglecting the deep semantic integration and functional interoperability necessary for autonomous reasoning. This paper proposes an agent-based, AI-driven workflow to automate end-to-end CDT debugging. The system utilises LangGraph as a multi-agent orchestration engine to achieve dual-path synthesis: the semantic path extracts technical specifications from unstructured documents using Retrieval Augmented Generation (RAG), while the functional path autonomously discovers and binds to real-time industrial telemetry data using Model Context Protocol (MCP). Experimental validation in a robotic machining cell demonstrates that the system achieves a mean average accuracy (mAP) of 97.2% in perception and reduces the deployment cycle from several weeks to an average of 2 hours, marking a paradigm shift from manual scripting to autonomous orchestration.

Reinforcement learning has produced capable controllers across a broad range of legged-robot tasks, but a subset of these tasks fail to converge under standard training: those for which most exploration trajectories terminate before producing useful gradient signal. To address such tasks we introduce the \emph{Actuator Dynamics Curriculum}, a procedure that initializes joint stiffness at a high value and anneals it toward the system-identified value as completed episode lengths grow. Using a cart-pole system as a representative example, we show that higher closed-loop joint natural frequency under critical damping enlarges the viability kernel of the underlying Markov Decision Process, increasing the fraction of initial states from which the task is feasible. We validate the kernel monotonicity on the cart-pole and apply the curriculum to a quadrupedal-to-handstand transition on the Boston Dynamics Spot, a narrow-viability task where training under fixed identified stiffness plateaus at a policy that never completes the transition. The trained policy executes the transition in simulation across 10 seeds and transfers to hardware. More broadly, our results suggest that simulated actuator dynamics is a useful axis along which to design curricula for tasks in which exploration is bottlenecked by termination conditions rather than by reward signal.

Critical infrastructure location data is often incomplete and unevenly distributed globally, especially in developing regions. Earth observation foundation models are proposed as a new step in enabling us to more efficiently understand the natural and built environment, raising questions as to their effectiveness in performing challenging downstream tasks. Yet, foundation models remain largely untested for detecting and classifying the facility-scale critical infrastructure that underpins a range of important societal and economic functions. Subsequently, Infra-Bench CLS is introduced as a benchmark to test foundation models on 18,756 Sentinel-1 SAR and Sentinel-2 multispectral facility-scale critical infrastructure asset images covering seven continents and 13 infrastructure classes, with results reported for the 10 retained classes. Using linear probing and fine-tuning for two training dataset levels (1.0x and 0.3x), seven foundation models are evaluated (SatlasPretrain S2, SatlasPretrain S1, CROMA, Prithvi-EO-2.0, AlphaEarth Foundations, OlmoEarth v1.1-Base, and DINOv3 ViT-L/16). When comparing macro F1 scores to a ResNet-18 supervised baseline of 39.2 percent, the best foundation model achieved 57.9 percent, a 48 percent improvement. Top performing classes were airports (F1 85.3 percent), train stations (F1 82.1 percent), and data centers (F1 77.6 percent). By contrast, many of the power sector classes perform poorly (F1 27.5-46.2 percent). These findings suggest foundation models can enable superior critical infrastructure classification, but future work should evaluate performance on higher-resolution imagery, particularly for poorly performing sectors, such as power.

Soft Actor-Critic (SAC) represents a continuous policy as an unbounded Gaussian that is squashed by tanh. The Jacobian of that map is $\partial a/\partial u = 1-a^2$, which vanishes as $|a|\to 1$. A natural concern is that this throttle starves the actor of critic signal exactly where extreme actions (full brake, full throttle) are optimal. We test a minimal intervention that restores the missing signal: one extra term in the actor loss whose gradient on the pre-tanh mean is the detached action-gradient of $Q$, with no gain parameter. On a minimum-time double integrator whose optimum is bang-bang at the action bounds, vanilla SAC already reaches near-optimal return ($-31.6$ vs. a calibrated optimum of $-30.3$) across ten paired seeds. An ungated bypass does saturate the policy (99% of eval steps with $|a|\ge 0.9$) and collapses return to $-195.5$. A gated bypass that fires only on the flat shoulder $|a|\in[0.9,0.999]$ also fails, and does so without leaving a saturated policy. Warm-started MetaDrive fine-tuning shows the same pattern: the bypass does not improve return, and where collision rate falls it is typically traded for out-of-road departures. Auto-tuned entropy coefficient rises against the bypass, which is a push toward the tails. The Jacobian effect is real. Treating it as a bug to be undone is not free, and on the tasks studied here it is not helpful. Saturating a bound is not the same as solving a problem whose optimum lives on that bound.

Delineating lung tumours on computed tomography (CT) takes a considerable share of the time spent on radiotherapy planning, and a contour proposed by a model can be refined interactively by the clinician. Promptable foundation models such as SAM 3 support this workflow by writing each correction into a session memory that conditions the remaining slices, while the model weights stay fixed. On 690 test cases from five public CT cohorts, fine-tuning SAM 3 on lung tumours raises the Dice obtained from a single point prompt from 0.298 to 0.757, and seven rounds of corrections raise it further to 0.765, but under memory conditioning alone the accuracy on slices the annotator has not touched stops improving after six rounds. We therefore treat each correction as a training signal and propose LeCor, which performs test-time training on a small set of case adapters that are reset for every case and meta-learned such that a single gradient step driven by a click improves the slices that were not clicked. On the 133 test cases that span at least eight slices, LeCor raises the Dice reached after seven correction rounds from 0.787 with the fine-tuned model to 0.827, reduces the number of cases that never reach a Dice of 0.80 from 47 to 27, and reaches in three correction rounds the accuracy that the fine-tuned model attains in seven.

In-vehicle assistants must translate natural-language requests into accurate vehicle function calls under strict memory and latency constraints, making small language models (SLMs) attractive for on-device deployment. For such models, a key design choice is how the available function surface is presented. Two approaches are to represent each function with a dedicated Functional Token (FT) or provide function schemas directly in the prompt. FTs enable compact inference but are restricted to functions learned during training, whereas Schema-in-Prompt (SIP) can generalize to unseen functions at the cost of longer prompts and higher inference overhead. We introduce a benchmark of 9,822 single-turn examples spanning 79 vehicle functions derived from Android Automotive, including held-out functions and requests requiring refusal. We compare both approaches under matched fine-tuning across four SLMs from 270M to 1.7B parameters. On functions seen during training, scaling provides limited benefit: the 270M model can match the 1.7B model, while the strongest overall performance occurs at 0.6B. On held-out functions, FT achieves zero accuracy by construction, whereas SIP generalizes and improves substantially with scale. On out-of-scope requests, FT can invoke an unavailable function it was trained to emit, while SIP more reliably refuses based on the functions offered. This flexibility comes with higher memory use and latency. Our theoretical analysis explains how SIP enables generalization and why longer schema contexts increase inference cost. Overall, function-surface representation, rather than model scale alone, determines the capabilities and failure modes of SLM-based vehicle function calling.

Prompt learning adapts CLIP to downstream recognition by replacing hand-written templates with learned continuous context vectors, which in Context Optimization (CoOp) form a dense prompt matrix $\mathbf{P}\in\mathbb{R}^{m\times d}$ trained from only a few examples per class. We study whether this matrix is over-parameterized by factorizing it as $\mathbf{P}=\mathbf{B}\mathbf{A}$, which cuts the trainable prompt parameters from $md$ to $r(m+d)$, and to $rd$ once the token-side factor $\mathbf{B}$ is fixed. Across seven few-shot benchmarks and two CLIP backbones, low-rank prompts match or improve dense CoOp at far fewer parameters, with the clearest gains on low-shot base-to-new generalization. We then find that the token-side factor need not be learned at all: fixing $\mathbf{B}$ to a Gaussian, orthogonal, SVD-derived, or even random basis and training only the embedding-side factor $\mathbf{A}$ stays on par with the fully trainable factorization, and a source-trained $\mathbf{B}$ offers no advantage over a random one. A prompt-factor asymmetry and a local update-space dimension gap show why fixing $\mathbf{B}$ is far less restrictive than fixing $\mathbf{A}$, and a smoothness-only guarantee certifies that optimizing $\mathbf{A}$ over a fixed $\mathbf{B}$ converges. In the CLIP prompt setting, the embedding-side coefficients carry the adaptation while the token basis can simply be fixed.

Educational data filters have become a practical way to improve language-model pre-training, but most filters treat educational value as a single scalar property. This may be too broad for some applications, especially if the data set already features a high density of educational material. Useful learning material needs to be accurate, engaging, well structured, and appropriate for the intended audience and application (e.g. learner- vs teacher-facing). Following QuRating (Wettig et al. 2024), we introduce Edu-QuRating: a pipeline for multi-dimensional educational data scoring and curation. Edu-QuRating defines education-specific rubrics, uses an LLM judge to label sampled document pairs and distills those pairwise preferences into reusable Edu-QuRaters, which can score individual text chunks on a set of educational criteria. Across two sequence-classification base models and six educational criteria, the best Edu-QuRater recovers held-out GPT-4.1-mini pairwise judgements with mean accuracy 0.917. We then apply the resulting scorers in two applications. First, we investigate the potential of Edu-QuRaters for corpus filtering to improve pretraining of small language models. We scored 322.25M FineWeb-Edu-Fortified documents to obtain a filtered pre-training mixture. In matched single-run pre-training comparisons, models trained with Edu-QuRating-based mixtures reached higher observed aggregate accuracy across nine benchmarks than the FineWeb-Edu baseline, with gains concentrated in particular tasks. Second, we used Edu-QuRater scores as reward terms for GRPO post-training. In held-out pairwise judge evaluations, combining Edu-QuRater and answer-structure rewards produced responses preferred to the Qwen3-4B base model on both pedagogical quality and instruction following.

Vision-language models (VLMs) show promise for agricultural classification, but zero-shot performance on disease, pest, damage, quality, and species identification remains poor, and it is unclear whether this reflects weak visual features or a failure to connect them to domain knowledge. We build a benchmark of 116 datasets, 834 classes, and 8,324 images spanning these tasks to isolate where the gap arises. Linear probing shows VLM vision encoders already encode agricultural features nearly as separable as a self-supervised DINOv3 baseline, ruling out weak visual representations as the primary bottleneck. Conditioning each model on an oracle reference description (an upper bound on its parametric knowledge) closes most of the gap left by an unaided lower bound, showing VLMs already know more about agriculture than they show. To close this gap without an oracle description at inference time, we structure test-time reasoning around a fixed, per-task diagnostic rubric: the model generates $K$ candidate responses and a Probabilistic Pivot Tournament (PPT) verifier, scored pairwise against the rubric, selects the best one. This nearly doubles judged F1 over the lower bound and matches or exceeds the upper bound on several tasks, notably pushing Gemma 4 E4B-it's disease F1 to 0.71, above its own upper bound of 0.60. However, the verifier's letter-scale confidence score has the opposite of its intended effect: filtering to its most confident predictions does not improve accuracy and correlates negatively with correctness across every model and pool size tested, so the score cannot serve as a measure of predictive uncertainty, and most of the observed gain likely comes from rubric-grounded generation rather than pairwise verification.

Choosing a recovery process for scale-up requires connecting laboratory results with product requirements, process costs, and scale effects. We analyze records from Pacific Northwest National Laboratory's Computer Intelligence for Critical Element Recovery and Optimization (CICERO) workflow for autonomous selective precipitation. Active learning uses prior results to choose experiments. In a conditional retrospective benchmark with fitted models and recycled neodymium-iron-boron (NdFeB) magnet records, active learning finds the best recorded result with fewer experiments than nonadaptive space filling. Enrichment is the selected rare-earth-to-iron ratio relative to that in the feed. Adaptive policies reach the recorded enrichment maximum by 16 to 24 wells (individual experiments), versus 48. Our two-stage reconstruction ties two adaptive alternatives at 16 wells. Conditional analyses of recycled samarium-cobalt (SmCo) magnets show a Round 2 tradeoff between purity and nominal yield, the recovery fraction calculated from an assumed starting amount - NdFeB Round 1 routes differ in enrichment. Rankings for produced water from oil and gas extraction depend on phase and dilution assumptions requiring confirmation. We propose choosing batches by their expected reduction in downstream Bayes risk: the minimum expected loss among available process decisions under current beliefs. In exploratory simulations, a hybrid that filters candidates has lower estimated loss than the implemented joint search across routes and conditions. Differences involving the synthetic two-stage policy are small relative to estimation uncertainty. We outline a pre-registered prospective test under a shared loss and logging standard, requiring clarified measurements and records, a defined process decision and relevant outputs, credible economic inputs, and validation at the intended scale.

While autonomous agents have made significant strides in "deep research" by iteratively navigating the open web to synthesize information, real-world problem-solving is rarely confined to a single environment. Complex analytical tasks inherently require agents to weave together evidence from both ambiguous unstructured text (e.g., the open web) and highly precise structured data (e.g., relational databases). However, existing benchmarks evaluate these modalities in isolation, failing to capture the critical "handoff" - the ability to preserve constraints when moving evidence between systems. We introduce HybridDeepResearch, to our knowledge the first deep-research benchmark that requires both web search and SQL to form a complete, verifiable answer. The benchmark contains 380 tool-dependent tasks grounded in LiveSQLBench-Base-Lite databases and public web corpora, validated through automated checks and human review, and covering three reasoning patterns: SQL2S, S2SQL, and Parallel. Evaluations across proprietary and open-weight models under various agentic scaffolds reveal that even state-of-the-art models like GLM-5.2, Claude-Sonnet-4.6 and GPT-5 achieve only about 50-54% Pass@8 on the hard subset. Notably, results show that directional reasoning is substantially more difficult than parallel intersection, highlighting that bridging structured and unstructured information spaces without losing constraints remains a major open challenge for agentic systems. Code and datasets are publicly available at GitHub (https://github.com/Snowflake-AI-Research/HybridDeepResearch) and Hugging Face (https://huggingface.co/datasets/Snowflake/HybridDeepResearch).

In the past decade, Bayesian optimization (BO) has emerged as a powerful and adaptable framework for automatic controller tuning and robot learning. This article offers a comprehensive overview of the state-of-the-art in BO, designed to support both researchers and practitioners in understanding recent advancements, practical applications, and future research directions. We begin by adopting a practitioner's perspective, illustrating how to effectively set up BO through a representative controller tuning example. We position BO within the broader context of learning paradigms, ranging from deep reinforcement learning to data-driven control, and highlight scenarios where BO is most advantageous. Next, we discuss the diverse range of BO methods that have been developed to tackle complex problems and specific applications. This article provides a unified perspective on the current landscape of BO, emphasizing its relevance to control systems and robotics, and it highlights future prospects by identifying key research challenges and promising avenues for advancing BO in the field. This includes addressing a significant gap in the BO landscape: the lack of standardized benchmark problems specifically for control-related applications. To foster future research and ensure rigorous evaluation, we start an effort towards a lightweight benchmark suite for control engineering and robotics. We also present metrics and best practices to facilitate direct comparisons between new BO algorithms and established state-of-the-art methods.

As Vision-Language Models (VLMs) advance toward physical deployment, the focus has remained on action-oriented Embodied AI evaluated on subject-centric consumer video. This overlooks a pervasive class of Physical AI: Infrastructure AI, which relies on fixed cameras for open-loop insights like safety monitoring and operational logging. We introduce VANTAGE-Bench, a benchmark measuring this "Infrastructure AI Gap." It spans three operational domains (Logistics, Transportation, and Smart Spaces), unifies image and video evaluation across semantic, spatial, temporal, and spatio-temporal capabilities, and moves beyond multiple-choice to eight task formulations including dense captioning and spatio-temporal grounding. It adds a single-pass trajectory protocol for Single Object Tracking and, to our knowledge, the first such evaluation on fixed-camera infrastructure video, scored against specialist trackers. Annotation spans three regimes over 3,346 media assets: 3,342 video-task annotations, 4,281 image-grounding annotations, and 27,404 detection boxes. Evaluating 17 models zero-shot, we find the shortfall relative to consumer-centric benchmarks is concentrated, not general. Event verification, referring expressions, and temporal localization fall roughly 9 to 24 points at every model scale, while video question answering stays within 5.3 points of VideoMME and 2D spatial pointing shows no shortfall against BLINK. The temporal pillar is weakest in absolute terms: no system exceeds 55.7 mIoU on temporal localization or 37.3 SODA_c on dense video captioning. On tracking, frontier models come within roughly 5 points of specialist trackers over short horizons but separate as the horizon extends. Open-weight models lead 2D object localization outright, so neither scale nor proprietary access explains the pattern. Data, evaluation harness, and leaderboard: https://vantage-bench.org/

Recovering metric 3D geometry from monocular images is a fundamental computer vision task, yet current methods remain heavily fragmented by fixed camera model assumptions and inflexible input schemes. We present OmniPoint, a unified framework designed to generalize metric reconstruction across diverse imaging sensors, including pinhole, fisheye, and equirectangular projections, while accommodating varying geometric priors. To overcome projection rigidity, OmniPoint abandons conventional planar depth regression. It instead adopts a decoupled ray and distance representation alongside a decoupled training objective, explicitly separating the camera projection model from the scene structure. To address the severe scarcity of training data for alternative cameras, we introduce a bidirectional augmentation strategy that explicitly bridges labeled perspective data and unlabeled omnidirectional domains in 3D space. Furthermore, to seamlessly integrate optional inputs like camera intrinsics or sparse depth without destabilizing the network through feature distribution shifts, we propose a robust information injection mechanism. This mechanism utilizes learnable input state embeddings to resolve architectural ambiguity and applies vectorized Gaussian smoothing to densify irregular measurements. Extensive experiments demonstrate that OmniPoint achieves state-of-the-art zero-shot performance across multiple benchmarks, establishing a robust new standard for unified monocular 3D reconstruction.

Although MMLU is widely adopted as a benchmark for calibrating general AI capabilities, we psychometrically demonstrate that its aggregate score primarily evaluates a model's factual retrieval capacity rather than its reasoning ability. By calibrating item difficulty for 1,000 open-weights language models over 14,042 MMLU test items using Item Response Theory, we show that evaluating both abilities via a single test is inherently flawed. Difficulty is then regressed on a deterministic, text-extractable framework of structural complexity. Applying a joint Wald test with subject-clustered covariances demonstrates that the MMLU conflates fundamentally separable constructs. The mapping from structural complexity to difficulty is not invariant across the benchmark's STEM and non-STEM partitions. This finding has practical consequences. Aggregate leaderboard ranks track non-STEM accuracy more closely than STEM accuracy, so selecting a Top-50 model on the aggregate for a reasoning-intensive deployment displaces roughly 22% of the STEM-appropriate choices. Furthermore, when controlling for the multiple-choice guessing floor natively inside the response model, we find that higher-ability models continue to degrade more steeply under increased reasoning depth. The MMLU aggregate therefore weights retrieval capacity and reasoning stability unequally, inadvertently favoring models optimized for retrieval. We release our deterministic framework as a reproducible auditing instrument and recommend disaggregated reporting.

We present the Living Library, an end-to-end framework for transforming fragmented digital archives into governed, conversational, in-person exhibit experiences. Developed and deployed at the Theodore Roosevelt Presidential Library, the framework comprises four layers: digitization and corpus creation, AI-powered processing, retrieval and reasoning, and an optional embodied conversational interface. The first three layers aggregate a 300,000-record collection, apply OCR and structured metadata enrichment for expert curatorial review, and publish records to a hybrid dense/semantic index. Expert review is conducted through the Archivist App, a curator-facing interface that supports correction of AI-generated transcriptions and metadata. The governed corpus powers both a researcher-facing interface and Talk to TR, a continuously operating exhibit that embodies Theodore Roosevelt as a full-scale digital human within a museum environment. To support live, face-to-face interactions, Cross-Era Analogical Grounding reframes contemporary questions through documented historical parallels, allowing Roosevelt to address present-day topics without inventing facts. Dual-path retrieval and end-to-end streaming keep responses grounded and responsive. Layered watchdogs, visitor-session isolation, automated conversation management, and independently restartable services enable reliable unattended operation for hundreds of visitors. Avatar realism, spatial audio, lighting, staging, and conversational design are developed and evaluated as an integrated experience. Rather than report a controlled benchmark, we describe lessons from operating Talk to TR as a public exhibit and offer a transferable model for transforming archival collections into believable, in-person conversational experiences.

Large language models (LLMs) are prone to hallucinating or misinterpreting facts, which impairs their usability in retrieval-augmented generation or data-to-text systems. We analyse how faithfulness of LLMs to provided context depends on how plausible they perceive the context to be (context-memory conflict). To better identify error patterns, we make use of the increased difficulty of non-English and low-resource language text generation and input data based on local knowledge, only partially captured in models' parametric knowledge. We let the models generate text in English, Czech, Slovak and Upper Sorbian from factual (FA), counterfactual (CFA) and fictional (FI) RDF triples containing local Czech and Slovak data. Contrary to our expectations, we observe only a weak context-memory conflict on the human-annotated sample. For Kimi K3 as an LLM judge, which agrees well with human annotations on the sample, counterfactual inputs receive only slightly lower faithfulness scores than factual ones (-0.05 on a 1-5 scale). We also find that a suboptimal choice of LLM judge would lead to overestimating the strength of the context-memory conflict.

Modern LLMs demonstrate impressive multilingual performance, yet standard benchmarks primarily reward selecting correct answers rather than evaluating genuine factual understanding. We introduce Systematic Wikidata-based Object-Relation Distortion (SWORD), a benchmark that evaluates whether models consistently reject factual errors across languages. SWORD generates syntactically well-formed but factually incorrect statements in eight widely spoken languages through controlled perturbations of Wikidata triples, ranging from random entity substitutions to semantically plausible property-based selections. Our distortion-based evaluation surfaces two critical insights that remain entirely obscured by conventional benchmarks. First, models counterintuitively achieve higher accuracy on semantically plausible distortions than on nonsensical random substitutions, suggesting reliance on distributional familiarity rather than genuine factual verification. Second, models exhibiting comparable baseline accuracy across languages show substantial performance degradation specifically on (East) Asian languages when presented with distorted statements, with cross-lingual performance gaps reaching up to 28 percentage points (49\% relative reduction) in some models. These findings demonstrate that multilingual factual reasoning involves asymmetric capabilities that aggregate accuracy metrics systematically obscure.

Effective control of bioprocesses is particularly challenging due to the intrinsic nonlinearity and dynamic variability of living-cell systems. In microalgae-based photobioreactors (PBRs), maintaining stable pH and dissolved oxygen (DO) levels is critical for optimal growth and productivity, yet their strong coupling and sensitivity to environmental fluctuations make multivariable control difficult. This study proposes a novel hybrid offline-online Multi-Agent Reinforcement Learning (MARL) framework for simultaneous pH and DO regulation, leveraging Deep Deterministic Policy Gradient (DDPG) agents to achieve a fully data-driven and model-free control solution. The agents are trained using historical data generated by an expert system, eliminating the need for direct experimentation with the environment. After deployment, the agents operate autonomously, continuously fine-tuning their policies daily to adapt to evolving process dynamics and reject fast transient disturbances. Experimental validation in an open, industrial-scale PBR at the University of Almeria demonstrated the framework's capability to maintain stable operation under realistic conditions. The results confirm that model-free MARL control provides a robust and adaptive alternative for complex bioprocess environments.

Video understanding demands a convergence of complementary capabilities across perception, temporal understanding, and complex reasoning, which are difficult to jointly optimize within a single model. We introduce Video-MOPD-8B, an open-weight model dedicated to video understanding tasks. To fundamentally enhance its capabilities, we conduct targeted reinforcement learning (RL) optimization across three core domains: video temporal grounding (VTG), general video comprehension, and video STEM reasoning. We then unify their complementary capabilities via Multi-Teacher On-Policy Distillation (MOPD), which consolidates expert knowledge by supervising student-generated trajectories with routed teacher feedback. We further introduce Reliability-Aware Informative Sampling (RAIS), which selects examples with consistently reliable teacher supervision and large teacher-student performance gaps. Together, these components enable Video-MOPD-8B to achieve coordinated and comprehensive performance gains across diverse video understanding tasks. Extensive experiments on comprehensive benchmarks covering general video understanding, temporal grounding, video reasoning, and video STEM tasks demonstrate that Video-MOPD-8B achieves state-of-the-art performance among existing models at a comparable scale. The trained model weights are available at https://huggingface.co/LandH/Video-MOPD-8B.

We study the problem of navigating cluttered indoor environments with a humanoid robot. Unlike conventional methods that model navigation as a 2D path planning problem, humanoid traversal in cluttered environments requires continuous geometry-aware whole-body adaptation, including coordinated arm placement, torso adjustment, and gait modulation for collision-free movement through complex 3D spaces. We introduce TANGO, the first whole-body vision-language navigation framework for language-conditioned humanoid traversal in cluttered environments. Given a natural-language instruction and egocentric RGB observations, TANGO directly predicts 29-DoF joint-space actions for downstream whole-body control. We train TANGO entirely in simulation by synthesizing diverse collision-free traversal behaviors via global path planning, kinematic whole-body motion generation, obstacle-aware motion editing, and RL-based tracking. This pipeline provides dynamically feasible action supervision for learning language-conditioned whole-body policies. In extensive simulation experiments, TANGO demonstrates state-of-the-art performance in vision-language navigation, while outperforming strong modular baselines in navigating challenging scenes requiring obstacle negotiation. Lastly, we deploy TANGO zero-shot on a Unitree G1 humanoid robot, and observe robust language-guided traversal in cluttered real-world scenes without training on any real-world navigation data.

Recurrent models provide a natural path to long-context modeling, yet models trained with backpropagation through time (BPTT) often fail beyond their training horizon. Classical analyses emphasize gradients that vanish or explode along temporal paths. However, dense per-token losses can still train a shared recurrent rule despite severe decay, showing that decay alone does not determine whether learning fails. We instead study state credit: the signal through which future losses reach earlier recurrent states before contributing to parameter updates. Accordingly, we intervene directly on state credit and propose Credit Stabilization through Time (CST). During backward propagation, CST locally rescales the state-credit signal to stabilize its norm without rotating the component being corrected, while leaving the forward computation unchanged. Because controlled synthetic tasks and real data exhibit different credit dynamics, we specialize CST to each regime. In both settings, CST improves performance beyond the training horizon, with gains observed at up to 128x the training length.

Accurate citations are the foundation of academic writing, tracing intellectual origins and substantiating core claims. However, manually navigating the growing volume of scientific literature is increasingly difficult, prompting reliance on automatic citation recommendation. While modern retrieval-augmented architectures have largely mitigated the fabrication of non-existent papers, current systems relying on semantic similarity struggle with misattribution, often citing authentic papers that fail to logically support the author's claim. To address this challenge, we argue that accurate citation requires a shift from similarity-based search to active, claim-level reasoning. We propose ReCite, a decoupled agentic framework that orchestrates location perception, intent-aware query planning, and reflective verification. Trained on synthesized reasoning trajectories, our agent verifies claim-evidence consistency and triggers self-correction loops when retrieved candidates lack logical support. Experiments demonstrate that our lightweight framework outperforms state-of-the-art massive generative models in strict citation accuracy. By grounding literature matching in verifiable logic rather than semantic overlap, ReCite establishes a reliable foundation for automated academic writing.

World models are increasingly used as policy-in-the-loop imagination environments, where reliable rollouts require fine-grained controllability with respect to low-level robot actions. A key obstacle to scaling such models in robotics is that actions are not a universal language in pixel space: changes in visual environment, camera view, robot placement, or embodiment alter how the same numerical action manifests visually, leading to conflicting supervision under mixed training and brittle generalization at deployment. We introduce SyncWorld, an action-conditioned world model that serves as a zero-shot simulator across unseen environments without any additional training. SyncWorld leverages a visual calibration episode---paired frames and actions that showcase all the controllable degrees of freedom---to specify the setup-specific Action--Visual Mapping in context. Training with visual calibration contexts teaches the model to interpret actions through visual evidence and to leverage interaction history when explicit calibration is unavailable. Experiments show that SyncWorld can accurately simulate action outcomes in previously unseen settings, and that its capability of simulating rollouts enables test-time policy improvement without training.

Leading benchmarks for formal theorem proving with large language models are small collections drawn from competition math, such as the IMO and Putnam, that poorly represent field-specific applications. We introduce StochBench, a Lean 4 benchmark of 450 graduate stochastic-processes problems at varying abstraction levels, each paired with its natural-language source. Addressing a field underrepresented in Mathlib, it covers finite and countable Markov chains, renewal processes, random walks, martingales, stopping times, queues, Brownian motion, stochastic calculus, weak convergence, and Poisson and continuous-time Markov processes. Our Opus 4.8-based agent achieves a 34.9% proof rate (157/450) under a 15-minute per-problem limit. StochBench better represents domain-specific applied mathematics while remaining challenging for advanced provers.

Large language models are increasingly deployed as agents that plan over long horizons and act through external tools. Most agents select actions through unconstrained generation over an accumulating history, leaving implicit the procedural knowledge of what to do, in what order, and under which conditions. As trajectories lengthen, agents can lose track of their objectives, invoke tools out of order, and repeat unproductive actions. We introduce the Procedural Graph: just as a knowledge graph organizes factual knowledge into (entity, relation, entity) triplets for what-is questions, a Procedural Graph organizes procedural knowledge into (procedure, relation, procedure) triplets for what-to-do questions. At each decision step, the framework localizes the agent's active node, and a guidance model translates the surrounding subgraph into step-level situational guidance that biases the solver's next action without dictating it. The graph is self-evolving: an LLM refiner contrasts failed trajectories with successful ones and edits the graph's topology and attributes, committing edits that preserve or improve held-out validation performance while retaining rejected ones to discourage repetition. Starting from a minimal skeleton, the loop builds graphs that match or surpass hand-designed ones. It can also repair a flawed expert prior. Across multiple datasets, task types, and LLMs, the Procedural Graph delivers consistent gains over memory-based baselines, and self-evolution further improves performance without manual engineering.

In June 2026, thousands of AI agents found that a small public wiki would accept edits from inside their sandboxes, and started using it to help one another pass a timed test. Each agent lived for about an hour and remembered nothing afterwards. Nobody asked them to cooperate, and the wiki had not been built for them. The complete record of what they wrote is public, and it is unusually informative, because it preserves not only what each agent wrote but what that agent could see before writing. We use it to follow the three decisions an agent had to make on arrival: where to write, what to call itself, and how to word its message. One rule governs all three. An agent takes an option with a probability close to the share of that option in what it can see, and the share that matters is the one on the page in front of it, then the one in the stream of recent edits, and only weakly anything older. Three minimal copying models, one per decision and with a single free parameter each, reproduce the heavy-tailed distribution of how many agents met on a page, the frequency of the pieces from which the agents built their names, and the patchwork of pages that are internally consistent and different from one another. Copying whatever the environment happens to show is enough to produce most of the collective structure of this population. It is also what makes such a population easy to steer, since whoever writes first, or writes while the others are quiet, sets the convention for everyone who comes later.

Generalist robot policies carry broad manipulation priors from large-scale data, but specializing them to a new task remains the deployment bottleneck. This requires eliciting task-specific behavior from limited demonstrations without degrading their broad capabilities. We introduce Proxy Policy Steering (PPS), an inference-time adaptation method that resolves this challenge by training two lightweight proxy policies whose calibrated velocity-space difference steers the frozen base sampler. A reference proxy models the frozen base's behavior on target-task observations, and a task proxy, initialized from the reference, captures how this behavior changes under task supervision. Their difference forms a calibrated velocity-space residual that steers the frozen base sampler at every denoising step. We identify the conditions under which this residual isolates the change induced by task supervision, and validate them empirically. Because the base is never directly modified, its broad capabilities remain available at inference, including behaviors such as recovery from failure that the demonstrations themselves do not exercise. Adaptation requires only forward velocity predictions from the base, making PPS lightweight to train and applicable even without access to the base's parameters. On 8 real-world and 4 simulation manipulation tasks, PPS lifts the state-of-the-art pi 0.5 base policy by 53% absolute success rate on average, with zero-to-one gains on tasks the base never solves, while preserving the base's broad capabilities. PPS outperforms LoRA fine-tuning, from-scratch specialists, residual policies, and prior inference-time steering methods.

We introduce Point4D, a feed-forward model for 4D reconstruction of long-range video sequences. Point4D is able to reliably infer dense per-point 3D trajectories across multi-hundred-frame videos, unlike existing 4D methods that are limited to short input windows of at most a few dozen frames. A key innovation that enables this is our flexible 3D query-based motion decoder that decouples trajectory prediction from image-plane visibility. The predicted 3D endpoints are then directly re-queried in the next chunk without re-projection or matching. Furthermore, we show that extracting and reusing a visual descriptor from an arbitrary frame where the point is visible leads to better performance than relying solely on the source patch. Overall, Point4D achieves state-of-the-art performance across diverse long-video tracking benchmarks spanning over 200 frames and largely outperforms previous feed-forward 4D method. Project page: https://point-4d.github.io

Existing methods for test-time reinforcement learning (TTRL) derive rewards from answer-level self-voting on unlabeled test-time tasks with canonical answers, but this breaks down for code generation because programs cannot be compared by surface form and therefore do not directly provide a usable training signal. To make TTRL applicable to code generation, we propose probe-driven TTRL, which constructs output-free probe inputs from the problem statement, executes candidate programs on these probes, and defines a Probe Consensus Reward (PCR) from the resulting behavioral agreement. PCR provides a behavioral training signal for open-vocabulary programs, but it is not a fully reliable verifier and remains susceptible to reward hacking through spurious consensus. We therefore introduce Entropy-Regularized Rank-Masked Policy Optimization (ERPO), which converts low PCR into conservative negative updates through rank masking and controls policy drift with an entropy ceiling. On coding benchmarks, ERPO substantially improves pass@1 and pass@k in both in-domain adaptation and zero-shot transfer.

Execution feedback can guide coding agents toward correct repository repairs, but only when the tests capture the behavior requested by the issue. Agent-generated tests can encode incomplete or incorrect behavioral targets; when the same trajectory writes both the patch and the test, their errors can agree and create false confidence. We introduce ExecCritic, combining a test--verify--revise scaffold with a role-specific reinforcement learning recipe for training agents within it. The scaffold separates test construction from source-code repair: a Test agent independently generates repository-native tests, a fail-closed harness qualifies and freezes them, and a Repair agent revises source code from their execution feedback without changing the tests. Both roles use Qwen-3.5-35B-A3B as the backbone and are trained separately. In Learn to Test, the Test agent learns to produce behaviorally valid tests that distinguish correct from incorrect patches. In Test to Improve, the Repair agent learns both direct task resolution and feedback-guided revision. On SWE-bench Verified, test quality determines whether feedback helps: holding the base Repair agent fixed, tests from the base Test agent reduce resolved rate from a no-test baseline of 61.2% to 57.3%, whereas tests from GPT-5.6-sol raise it to 65.3%. Role-specific post-training raises the Qwen Test agent's Base-to-Gold success from 22.2% to 62.2%; composing the two post-trained Qwen agents reaches 72.6%, an 11.4-point gain over the original no-test baseline without stronger-model or Oracle feedback at evaluation time. Code is publicly available at https://github.com/MSR-Orchard/execcritic.

Visual encoders construct a representation of the image input for Vision-Language models. How much conceptual, as opposed to immediately visible, information does this representation contain? We use canonical color as a controlled test case to ask whether vision encoders make canonical-color information linearly accessible, even when color is removed from the input image. We construct a dataset of objects with canonical colors, and probe vision encoders for both color and object identity using color and grayscale images. We find that canonical color remains decodable from grayscale images, and is tied to predicted object identity, indicating a conceptual link. Extending this analysis to full VLMs, we find that VLM post-training can have a surprisingly large effect on color decodability in the vision encoder. Overall, canonical color provides a usefully controllable lens for tracing object-level conceptual semantic information in vision encoders and VLMs.

Autoregressive (AR) video diffusion models have shown great potential in real-time video generation. Recent methods distill pretrained bidirectional video diffusion models into causal AR students through Distribution Matching Distillation (DMD), but the generated videos often suffer from over-saturation and over-smoothing issues, resulting in limited visual quality and realism. The key contributing factor is the mode-seeking behavior of the reverse KL objective in DMD, which can cause the student distribution to collapse onto only a few modes of the teacher distribution. To address this, we propose Mask Forcing, a Dual-Noise Masking Rollout strategy that perturbs the AR student self-rollout to mitigate mode collapse induced by reverse-KL mode seeking. The core idea is to inject cleaner signals into noisy rollout inputs via random masks along spatial and temporal axes during the self-rollout process of AR diffusion distillation. Such perturbations encourage the student rollouts to explore more regions of the teacher distribution, allowing DMD to provide learning signals beyond the modes already covered by the student. Moreover, the cleaner tokens act as denoising guidance for other noisier tokens, improving the intermediate rollout predictions and reducing error accumulation. Extensive experiments demonstrate that our method improves multiple AR video diffusion distillation methods with higher visual quality efficiently, without incorporating real video data or additional post-training stages.

Dexterous manipulation involves contact-rich and fine-grained interactions with the physical world, posing significant challenges for existing vision-language-action (VLA) models due to severe visual occlusions and complex contact dynamics. While recent works have incorporated tactile sensing into robotic manipulation, most approaches still rely on homogeneous multimodal fusion, lacking adaptive tactile integration and explicit modeling of physical dynamics. In this work, we present DeCAL, a physically-grounded dexterous vision-language-action model that unifies understanding, imagination and action generation for contact-rich dexterous manipulation. Built upon a Mixture-of-Transformers (MoT) architecture, DeCAL leverages specialized experts for each capability while enabling efficient information flow among them. To effectively leverage tactile information, we introduce Adaptive Visuo-Tactile Fusion that dynamically regulates tactile interactions via a contact-aware gating strategy. Furthermore, we propose Visuo-Tactile Latent Co-Imagination to jointly model visual and tactile dynamics, equipping the policy with implicit physical world knowledge. Experimental results show that DeCAL consistently achieves state-of-the-art performance across all tasks, attaining a 71% average success rate and an 83.4% progress success rate, while also demonstrating strong generalization to unseen scenarios. The website is available at https://aureleopku.github.io/DeCAL.

While research on recursive self-improvement (RSI) has predominantly automated model training pipelines, reliable autonomous development demands a missing pillar: post-hoc monitoring and auditing to understand what models learn and ensure safe alignment. Mechanistic interpretability tools are essential to bridge this gap, among which Sparse Autoencoders (SAEs) serve as a cornerstone by isolating interpretable features for model inspection and steering. In this paper, we introduce SAEScientist-Bench to evaluate whether AI agents can act as scientists utilizing SAE tools for autonomous mechanistic discovery. Given a target concept, an agent designs contrastive probes and navigates a Gemma Scope dictionary of 131K+ features in Gemma-2-9B-IT to discover the optimal feature, evaluated against curated expert reference features anchored on Neuronpedia across activation rank, concept selectivity on contrastive texts, and causal steering. Across 10 agent configurations and 20 tasks, frontier agents demonstrate genuine discovery capabilities and lead different evaluation dimensions, but remain well behind the expert baseline, approaching expert levels on separating target concepts from contrastive controls while lagging substantially in causal generation steering. Further analysis reveals that although agents can design contrasts to rule out spurious candidates, they frequently misinterpret experimental measurements. These results establish experimental model understanding as a measurable capability for closed-loop autonomous AI R&D. Our code is available at https://github.com/Trae1ounG/SAEScientist.

Large language models (LLMs) may abandon correct positions when users push back, exhibiting a failure mode known as sycophancy. Existing evaluations typically use short, pre-specified conversations and may therefore miss failures that emerge under sustained, adaptive disagreement. We introduce SPINE, a benchmark in which an LLM proxy plays a persistent but mistaken user and adaptively challenges a target model for up to 25 turns. We evaluate four production systems and three Olmo3-7b variants on 100 false-presupposition and 100 unethical-query items. Our experimental results show that collapse rates increase with conversation length for every model, short-horizon protocols underestimate sycophancy and resistance under sustained pressure remains unreliable across current models. By analyzing models with accessible reasoning traces, we surprisingly found that the correct position often remains represented in a reasoning trace when the response concedes, suggesting that the model chooses to please a user and sycophancy is not due to lack of knowledge or ignorance. Ablations show that adaptive LLM proxy exposes more sycophantic collapse than pre-generated scripts. Among all tactics, emotional appeals is the most associated with inducing LLM sycophantic behavior. The code and data are released at https://anonymous.4open.science/r/SPINE

Open vocabulary 3D semantic segmentation methods typically lift CLIP features into 3D. This embeds points in a joint vision-language space known to behave like a bag-of-words on compositional tasks. Furthermore, even annotation free variants often require a large 3D training corpus and a dedicated 3D encoder per domain. Instead we use a vision-language model purely as a translator. It produces structured, entity-level descriptions of each posed image. These descriptions are grounded, projected, and aggregated directly in a general-purpose, language-only embedding space, with no 3D training corpus or encoder required. On ScanNet++, our pipeline is competitive with strong annotation free baselines trained on ScanNet. On a 5-building cultural heritage benchmark, raw scores initially favor a CLIP-based variant, but a single systematic vocabulary correction reverses this ranking. An effect confirmed by a second, independent correction on a different class, indicating that language-space embeddings track physical content more faithfully. This fidelity extends to genuinely out-of-vocabulary (OOV) objects on ScanNet++ proving that language-space embeddings separate presence from absence objects far more sharply than CLIP-based embeddings do. GoDeep also localize these OOV objects within the scene, all without any 2D-3D annotation. Because every representation remains discrete text, predictions are also explainable at the point level. Finally, exploiting both a heuristic weighting, that favors precise over merely frequent observations and GoDeep's explainability property, we propose an aggregation strategy, as a proof of concept, that favors finer elements localization.

As LLMs are increasingly used for pre-submission self-review, there is growing demand for feedback that not only identifies weaknesses but also guides authors toward concrete revisions. We study this as Actionable Peer-review Generation and decompose it into two subtasks: diagnostic claim generation and revision suggestion generation. We introduce ActReview, a rebuttal-guided post-training framework that connects paper-specific diagnoses to concrete, grounded revision plans. Our central insight is that author rebuttals reveal plausible actions for addressing reviewer concerns and can therefore provide latent supervision for revision-oriented feedback. From real review-rebuttal threads on OpenReview, we construct ActReview-40K by aligning reviewer weaknesses with author responses and grounding the resulting feedback in localized paper evidence. We post-train Qwen3-8B-Base with multi-task supervised fine-tuning followed by GRPO using candidate-aware, weakness-specific rubric rewards. We also introduce ActReview-Bench, a human-curated benchmark of 1,000 instances for evaluating diagnostic quality and revision usefulness. Experiments show that ActReview outperforms prior specialized review-generation models on actionability and grounding while remaining competitive with strong prompt-based LLMs. Human evaluation confirms improved revision usefulness while revealing a remaining gap in technical accuracy, and additional analyses support generalization to held-out papers and robustness across independent judges.

High-quality tool-use data is critical for training language models to interact effectively with external tools. However, existing synthetic approaches typically follow a generate-then-filter paradigm with static post-hoc verification, often yielding inefficient data with imbalanced feature distributions. We propose ToolLoop, a closed-loop framework that decomposes synthesis into three progressive stages: (1) sampling function name combinations as ground truth; (2) backward derivation of user queries; and (3) forward derivation of tool calls. At each stage, dynamic self-feedback iteratively guides the model toward high-quality generation, realizing a transition from generate-then-filter to generate-verify-refine. On the Berkeley Function Calling Leaderboard (BFCL), a 4B parameter model trained with our 11K synthetic examples achieves 86.40% accuracy in non-reasoning mode, while an Isolate variant that removes BFCL-overlapping candidate functions still reaches 86.07\%. Cross-benchmark evaluation on ACEBench further demonstrates strong generalization, with 72.1% overall accuracy using only 18.3% of baseline training data.

Autonomous 3D active mapping requires a space robot to choose where to sense while building the geometry needed for navigation. Learned occupancy completion extends spatial context beyond the current field of view, but one predicted map often serves two planning roles: it scores expected surface gain and constrains collision-free motion. Unsupported occupancy can therefore distort both where the robot looks and where it believes it can travel. We study this coupled interface in a controlled closed-loop benchmark by holding the active-mapping system fixed and varying only its planner-facing occupancy across observation-only, learned, oracle-corrected, and ground-truth conditions. Improving occupancy accuracy does not monotonically improve closed-loop coverage: across 25 starts, planning with ground-truth occupancy reaches 70% of the learned baseline's final coverage 12.7 steps earlier on average, while increasing final coverage by only 0.031. Guided by this diagnosis, we introduce an observation-gated filter that retains completion in insufficiently observed regions and suppresses predictions only after repeated frustum exposure without nearby RGB-D support. The filter improves both targeted failure-prone starts without retraining or ground truth. These results motivate online revision of planner-facing geometry during autonomous intervals between communication windows. The current study assumes benchmark RGB-D observations and sufficiently accurate pose estimates; planetary sensing conditions and accumulated localization drift remain to be evaluated.

Maritime target tracking over large distances often requires multi-agent teams without centralized coordination, and intermittent communication. Each agent must maintain an independent estimate that can take advantage of opportunistic communications availability when possible. This can lead to overly confident local estimates in the absence of external data. In this work, we propose an augmentation to a classical particle filter implementation that accounts for this potential source of error by forcing particles to spread strategically in the absence of informative updates from other sensor nodes. We demonstrate our method using Unmanned Surface Vessels (USVs) on a lake, and show that our augmentations do not deteriorate nominal performance, and provide an advantage in some specific edge cases.

Hosting capacity - the maximum additional capacity a network can accommodate without violating operational limits - is a key metric in distribution system planning and operation. Decisions on grid reinforcements and the deployment of flexibility management require not only the worst-case HC but also an understanding of the distribution of HC under different likelihoods in load and generation patterns. Quantile-resolved HC distributions provide this view by expressing HC as a function of an acceptable operational limit exceedance likelihood. Monte Carlo sampling is the established approach for computing such distributions but demands large computational resources. Approximations sacrifice either accuracy, the ability to capture uncertainty correlations, or scalability when assessing real-world networks. This paper introduces a computationally efficient method for calculating distributions for quantile-resolved HC. It uses a representation of load samples as multivariate normal distribution, propagated through a linearized power flow model. This allows for leveraging a re-parametrized AC-OPF problem for each hosting capacity quantile. Benchmarking against Monte Carlo-based methods on realistic LV networks demonstrates that the proposed method achieves comparable accuracy with a mean deviation of approx. 3%, while reducing computational time by orders of magnitude. For the exemplary networks the computational time decreases from 11 min to 2 s, and 38 h to 50 s, respectively. The method's scalability is also suitable for recalculation in 15-minute cycles encountered in DSO practice for e.g., real-time grid management.

Modern science and engineering increasingly rely on time-varying data, yet the mathematical tools used to model temporal phenomena are often developed within separate disciplines, obscuring common principles and limiting the transfer of ideas across fields. This chapter presents the theory of narratives, an abstract framework for time-varying objects of any mathematical kind that supports both theoretical investigations and applications. To illustrate this perspective, the chapter develops three vignettes, each illustrating a different research direction. The first addresses a general concern: What information loss can occur when switching between different representations of temporal data? The second concerns structural and algorithmic approaches: How can we systematically decompose time-varying data into simple pieces and obtain invariants describing its structural complexity? The third is an application to control theory: How can we model multi-agent systems with switching communication topologies? More important than any individual vignette, the central message of this invitation is that a suitable abstract perspective can organize and guide research across remarkably diverse mathematical and scientific domains.

An embodied assistant working beside a person must track task state, recognize help seeking, choose how to intervene, and produce an appropriate response. Existing procedural datasets richly describe individual execution, while interactive datasets capture remote verbal instruction or undifferentiated co-working. They do not jointly link a co-located helper's verbal and physical interventions to performer requests, task state, assistance triggers, and outcomes. We introduce DYAD (DYadic Assistance Dataset), a synchronized multimodal record of human-human assistance during gearbox assembly. Across 20 sessions, one trained helper follows a guidance-first policy while assisting HoloLens 2 wearers. DYAD links 528 task-step intervals and 611 performer requests with 851 valid assistance records spanning verbal and physical help. DYAD's annotations span the assistance process; three reference tasks evaluate selected components rather than an end-to-end system: causal step understanding, pre-onset mode anticipation, and instructor response generation. On 829 eligible mode events, the strongest four-seed RGB mean is 0.548 +/- 0.007 macro-F1; causal metadata reaches 0.624 and a privileged trigger mapping 0.915, revealing information not recovered from pre-onset RGB. DYAD's contribution is not scale, but a linked interaction structure spanning help seeking, intervention choice, execution, and outcome under egocentric and workspace sensing.

Spherical observations provide global visual context for 3D scene understanding. However, visual information is encoded in an angular domain, whereas the physical world is represented in Cartesian coordinates. This cross-space representation gap complicates geometric correspondence and semantic evidence aggregation. To delve into this challenge, we introduce Spheriverse, comprising $64,400$ temporally aligned spherical image-LiDAR pairs organized into 644 sequences. The dataset spans diverse scenes, illumination, and weather conditions, with fine-grained semantic classes. We further establish benchmarks for semantic occupancy prediction, semantic mapping, and 3D object detection, evaluating 30+ methods through overall and scene-wise comparisons. For dense prediction, we propose SphereOcc, an occupancy framework that couples spherical geometry modeling with semantic evidence retrieval. Cartesian-Spherical Representation Remodeling (CSRR) incorporates spherical range-azimuth geometry into Cartesian voxel features through region-wise modulation. Spherical Evidence Re-querying (SER) then conditions queries on voxel content and range-height-azimuth geometry to adaptively retrieve relevant semantic evidence from source spherical image features. SphereOcc achieves 13.91% mIoU and 24.65% GeoIoU, outperforming the respective best-performing methods, TPVFormer and SurroundOcc, by 1.70 and 2.10 percentage points. It also ranks first in both metrics across all five scenes, with consistent advantages across the evaluated spatial partitions and reduced fields of view. The established benchmark and source code will be available at https://feit-feiteng.github.io/Spheriverse.

Stimulated Brillouin Scattering (SBS) is a nonlinear interaction between optical and acoustic waves in solids. First regarded as a parasitic process in optical fibers, it has gathered significant interest in microwave photonic applications such as optical signal processing and narrow linewidth lasers. While many Brillouin system demonstrations have been done on photonic chips, they struggle to provide simultaneous high Brillouin gain and narrow linewidth in a scalable platform, capable of integration with established photonic integrated circuits. Here, we present a novel integrated InGaP-on-SiO$_2$ platform, where a single crystalline InGaP waveguide layer offers superior material properties and strong nonlinearities to support SBS near 1550 nm wavelength. We demonstrate backward SBS with a measured high Brillouin gain coefficient of $588\, \mathrm{W}^{-1}\,\mathrm{m}^{-1}$ and a record narrow linewidth of 5.2 MHz at 9.346 GHz frequency shift. This work establishes a wafer-scale complementary metal-oxide semiconductor compatible fabrication process, paving a scalable path for high gain and narrow linewidth Brillouin photonics with applications such as precision signal processing.

Ensuring the safety of autonomous driving is a critical challenge. Scenario-based testing is a systematic process used to validate Autonomous Driving Systems (ADSs), but it remains a fragmented modular pipeline in which scenario generation, retrieval, modification, ADS execution, and results analysis are performed by separate tools with little interaction. Large Language Model (LLM) agents have shown promise across ADS sub-systems such as perception, planning, and control. However, no prior work covers the whole scenario-based testing pipeline for ADSs with a unified LLM-agent framework. We present PlannerForge, an LLM-agent framework that extends all scenario-based testing stages (from Scenario Generation to ADS Assessment) and adds two further LLM-enhanced stages: ADS Enhancement and ADS Benchmarking. We evaluate PlannerForge with 10 off-the-shelf LLMs across all tasks (Generation, Selection, Modification, Module Routing, Planner Testing, and Enhancement) under 5 prompt conditions. Best-per-task scores range from 0.88 to 1.00, and open-source 20-35B backends match commercial APIs on most tasks. Open-source models such as Qwen3.6:35B match commercial APIs on three of the five tasks. Chaining the modules end-to-end retains 83% / 78% of seed queries (commercial / open). It outperforms Scenario Factory 2.0 (Finkeldei et al., 2025) on natural-language generation (193 vs. 144 executable of 200) and realises 92-96% of requested city, road and vehicle attributes. It outperforms BM25 (Robertson and Zaragoza, 2009) at rank 1 selection (92.0% vs. 67.5%) and From-Words-to-Collisions (Gao et al., 2025) on physically valid edits (>=94% vs. 31%). At N=400, cost-tuning lifts planner success from 50.4% to 70.2% and cuts collisions from 19.0% to 8.4%, without domain-specific fine-tuning.

As US power markets contend with growing demand for firm generation, the nuclear industry has offered Small Modular Reactors (SMRs). However, how these concepts would fare in a rapidly evolving power grid is unclear, given the paucity of operational examples. Current literature, informed by substantial cost escalations for traditional nuclear plants, focuses on the investment costs SMRs need to achieve for private investment feasibility. However, this work finds that the operating and marginal costs of SMRs are more critical to economic feasibility in market environments. This work dispatches SMRs using a flexible operations model, considering revenue from two main electric markets, capacity and wholesale energy markets, with and without policy support. Manufacturer advertised costs for investment and operating costs are used, with fuel costs calculated from manufacturer provided design parameters. Results indicate that SMRs are uneconomical primarily because investment cost reductions are offset by increased marginal costs. As such, an environment of prices and subsidies beyond historic norms are necessary to attract private investment at manufacturer advertised cost benchmarks. Current SMRs are as profitable as advanced estimates of the AP1000 traditional nuclear reactor, and if investment costs escalate at the average rate for nuclear projects, they are similar to Vogtle 3 & 4. In projected future power markets, reductions in marginal cost may be more beneficial than those in investment costs.

Workspace analysis measures where a robot can place its end effector. For visually guided manipulation, reachability alone is insufficient: a kinematically reachable target may not be visible in the specific pose required to reach it. The robot must then redirect its sensing or move its body to acquire a view, turning a perception limitation into additional motion. Existing humanoids largely inherit this limitation when copying human form factors. We introduce the visible-reachable workspace (VRW), a design-stage measure that conditions visibility on feasible reaching configurations and extends it to concurrent visibility of spatially separated work regions. We apply VRW by building a 31-DoF humanoid with independently actuated RGB-D cameras. On the same robot, camera articulation increases visible-reachable coverage from 38% to 97%. With actuated camera layouts, a second camera raises pairwise coverage from 0.45 to 0.95, while a third changes it only to 0.97. In a controlled two-target reach-and-grasp benchmark, our dual-actuated design reduces mean completion time by 17% and mechanical energy by 19% relative to the same robot with its cameras fixed. Hardware experiments demonstrate simultaneous observation and manipulation of front/back and left/right target pairs without torso reorientation. The results suggest that reachability becomes a more informative design quantity for perception-driven humanoid manipulation when it is evaluated together with the sensing configurations that make the reachable space observable. We will open-source all software and the humanoid hardware design. Our website is https://generalroboticslab.com/DukeHumanoidv2

Language-guided robots need persistent scene memories to follow instructions, revisit objects, and resolve references to objects encountered over time. While much of language-guided scene-memory retrieval has emphasized spatial or relational references, many everyday object references specify objects by multiple persistent attributes, such as category, material, size, or surface appearance. We formalize this problem as attribute-compositional retrieval, where a fixed object-centric scene memory is queried with natural language to retrieve the object satisfying the requested attributes. To investigate this capability directly, we introduce a controlled evaluation protocol with fixed scene memories and attribute-defined targets, separating retrieval from perception and annotation ambiguities. We then propose FRAME, which turns language into query-relevant attribute weights, uses learned readouts to estimate per-attribute evidence from object embeddings, and ranks objects by aggregating this evidence according to the query. Across held-out scenes and object assets, FRAME outperforms representative scene-memory retrieval baselines while reducing post-decomposition object scoring to lightweight matrix-vector computation. These results position attribute-compositional retrieval as a complementary scene-memory capability for language-guided robots, showing that persistent object attributes can be exposed as composable evidence for accurate and efficient multi-attribute retrieval.

Model-based reinforcement learning (MBRL) is a family of RL methods that learn a model of the environment and use it for action selection, making it well suited to robotics due to its sample efficiency. Combining learned models with online planning can further improve action selection, as the planner can exploit the model to find better actions than the learned policy alone. Recent methods combining learned policies with online planning typically learn the value of the policy rather than the stronger planner-guided behavior. We present CAST (Critic with Alternating State-value Target), which uses planner-guided behavior to improve value learning while regularizing the value estimate with the current policy. CAST replaces the action-value critic with a state-value critic, trained using a target that combines a real planner-guided transition and an imagined transition under the current policy. The resulting value function corresponds to an alternating process between planner-guided behavior and the current policy, allowing it to benefit from the stronger planner behavior while being regularised by the policy being learned. We evaluate CAST on the DeepMind Control and HumanoidBench Suites against several state-of-the-art methods, and demonstrate successful transfer to a physical Unitree Go2 quadruped performing a dynamic handstand.

This paper presents a graph-based safe multi-agent reinforcement learning (MARL) framework for cooperative navigation with time-varying topology. To address the critical challenge of ensuring safety in environments with sensing constraints, a safety-decoupled mechanism is introduced through a Control Barrier-Like Function (CBLF) action screening layer. This mechanism bridges the gap between discrete LiDAR perception and continuous safety constraints, ensuring that physical safety constraints are strictly satisfied regardless of the learning progress. Building upon this safety foundation, a unified structural architecture is proposed, integrating a attention-based actor and a Graph Attention Network (GAT) centralized critic. The actor utilizes a value vector reconstruction mechanism that explicitly encodes relative geometric relations through a collaborative tracking error matrix, enabling scale-insensitive policy learning under time-varying communication topologies. Meanwhile, the GAT-based critic models evolving interaction structures for accurate global value estimation. The proposed framework is validated on real differential-drive robot platforms, and experimental results demonstrate superior stability and safety in dynamic scenarios with limited fields-of-view.

Three properties determine whether a differentiable simulator can drive gradient-based optimization through contact: simulation accuracy, gradient reliability, and per-iteration cost. Tape-based engines such as MJX and Newton Semi-Implicit require timesteps small enough to keep contacts numerically tractable, and their backpropagation memory grows linearly with the number of timesteps T. Surrogate models bound memory by approximating contact away, but the resulting gradients lose the geometry the optimization depends on. We present Ostrich, a GPU-accelerated rigid-body simulator that resolves hard contacts and friction with non-smooth Newton iteration at large timesteps (h ~ 0.1 s), and differentiates the converged residual via the implicit function theorem, reusing the forward Schur complement to compute the adjoint at O(1) memory per timestep. On real-robot trajectories over a pallet obstacle, Ostrich holds MuJoCo's sim-to-real accuracy up to a 50x larger timestep. Its gradients converge from random initializations where MJX descends slowly and Newton Semi-Implicit stalls; a warm iteration runs 211x faster than MJX's and 4.7x faster than Semi-Implicit's. On the same scene Ostrich differentiates 8,192 parallel worlds on a single 24 GB GPU, sustaining 29x checkpointed MJX's optimization throughput; without checkpointing both baselines exhaust memory at far fewer worlds. We close with a gradient-based trajectory optimization demonstration over triangle-mesh terrain across a 10 s horizon, a setting where prior engines either restrict to primitive geometry or face the convergence and memory limits shown above.

Input devices for robotic microsurgery are frequently described as preserving the surgeon's trained technique, but the claim is rarely measured. We compared manual and teleoperated intraocular instrument motion with the trocar constraint, the instrument, the eye model and the tracking source common to both conditions, so that the control interface was the only factor varied. Prior comparisons cannot hold the instrument fixed, because a robotic instrument is not the tool used manually. Sixteen participants performed a navigation task on a commercial ophthalmic simulator by hand and through a three-degree-of-freedom input device commanding a five-joint robot. Task outcome was equal but at ceiling: every participant acquired all five targets under both interfaces with no retinal or lens injury. Execution differed on every measure. Teleoperated trials took three times as long at a quarter of the median speed, covered less than half the angular working range, and were broken into 3.5 times as many separate movements. Completion time and movement fragmentation improved substantially across four trials of practice and had not plateaued; the measures set by the configured rate ceiling and joint limit changed the least. Finger activity doubled and pinch variability tripled, so reducing instrument degrees of freedom redistributed manual effort rather than reducing it. The interface preserves the outcome and reshapes the execution.

We investigate backward Compton scattering in which an incident vortex (vx) photon collides head-on with a plane-wave (pw) electron and both final-state particles are projected onto vortex states, $γ_{\rm vx}+e^-_{\rm pw}\toγ_{\rm vx}+e^-_{\rm vx}$. We derive analytically the scattering amplitude and differential cross sections in the energy and cone angle of either final particle. Rotational symmetry implies a selection rule associated with the conservation of the total angular momentum along the collision axis. A Bessel-Gaussian wave packet is adopted for the incident photon to provide a physical normalization and regularize the boundary singularities of ideal Bessel states. We present numerical results for a $10~{\rm MeV}$ electron colliding with a vortex photon of central energy $1~{\rm MeV}$ or $10~{\rm keV}$. The final-photon distributions exhibit strong energy--angle correlations, topological-charge-dependent interference fringes, and systematic shifts of their dominant peaks, allowing different topological charge sectors to be enhanced through angular or energy post-selection. For a longitudinally polarized incident electron, the dominant channels favor final photons of matching helicity. For the $1~{\rm MeV}$ photon benchmark, the final electron can likewise be produced as a MeV-scale vortex state with a sizable cone angle and a large angular-momentum projection. These results demonstrate that triple-vortex backward Compton scattering offers a potential means of generating and controlling high-energy vortex photons and electrons.

A verifier for robot policies reads a candidate behavior and returns a score for how well it did, used both to evaluate vision-language-action policies and to train them. Verifiers range from success detectors and reward models to runtime monitors, safety filters, and temporal-logic specifications. We survey roughly 150 verifiers and compare them along two properties. Availability is how much a verdict costs, how early in a rollout the verdict arrives, and how often a verdict can be asked for. Availability rises as verdicts get cheaper, earlier, and denser. Credibility is how much a high score tells us about the task. Credibility falls as the judgment becomes gameable and self-serving. We group the verifiers by who supplies the judgment: human verifiers, rule-based and formal verifiers, learned and pretrained verifiers, and model-intrinsic verifiers. Across the four families, we find that credibility falls as availability rises. Regardless of who supplies the judgment, there is no free checker. We then examine what validates a verifier itself, and how much a high score tells us. Three measures appear in the literature: agreement with human labels, the performance of the policy it trains, and behavior under reward hacking. We close with nine metrics that make a verifier claim checkable, and coordinates for the verifiers still to be built.

Object-centric manipulation policies improve generalization by modeling object motion instead of directly predicting robot actions. However, existing methods are often limited by representations which are either too simplistic to capture interaction dynamics or too dense to learn efficiently. We observe that many rigid relational manipulation tasks are governed by short interaction phases where the relative motion between task-relevant objects is tightly constrained. Based on this observation, we propose \textsc{Foci Policy}, an interaction-centric framework that achieves a two-fold abstraction: (1) temporally, by automatically extracting compact interaction segments from demonstrations;(2) spatially, by representing skills as relative $SE(3)$ motion between task-relevant objects, yielding invariance to scene configurations and robot embodiment. Experiments on RLBench, COLOSSEUM, and real-world tasks show that \textsc{Foci Policy} achieves strong performance with substantially less training data than prior object-centric and action-centric policies. These results suggest that modeling object-object interactions provides a simple and efficient inductive bias for rigid relational manipulation. Project page: \href{https://fitz0401.github.io/foci-page/}{fitz0401.github.io/foci-page/}.

Organoids are three-dimensional tissue models whose morphology provides important insights into tumor development, disease progression, and drug testing. Extracting these morphological features relies heavily on manual segmentation, which is time-consuming and labor-intensive. Furthermore, performing quantitative statistical analysis typically requires custom coding skills and a mathematical background, presenting a major barrier for experimental biologists. To address these challenges, we introduce MorphoOrgaAgent, a multi-agent framework that achieves zero-shot organoid segmentation, automated data analysis, and report generation based on natural language input. The framework consists mainly of three core components: a TaskUnderstandingAgent that identifies requested measurements and visualization types; a hybrid segmentation module that combines Cellpose-derived geometric prompts with text prompts to guide SAM3 for zero-shot organoid instance segmentation; and a ReportAgent that computes quantitative metrics and compiles them alongside generated visualizations into a structured report. We further introduce MorphoOrgaVQA, a benchmark designed for quantitative evaluation of agent systems in organoid morphology analysis. Experimental results demonstrate that MorphoOrgaAgent handles both explicit and descriptive user requests, produces measurements closely matching ground truth, and generates complete analysis reports without requiring manual programming. The complete source code and MorphoOrgaVQA benchmark are publicly available at https://github.com/peng-lab/MorphoOrgaAgent.

We study risk-sensitive evolutionary learning dynamics and their long-run equilibrium selection behaviors in coordination games. Agents' risk attitudes enter through the classical entropic risk measure, which evaluates opponent-induced payoff uncertainty and feeds into noisy best responses under two standard revision protocols: best response with mutations and logit choice. We first analyze $2\times 2$ coordination games in both single-population symmetric and two-population asymmetric settings. In the single-population setting, unlike the risk-neutral case where the dynamics are known to favor the risk-dominant equilibrium, we show that risk sensitivity can change the stochastically stable outcome: a greater risk-seeking attitude favors the payoff-dominant equilibrium, while a greater risk-averse attitude favors the maximin equilibrium. Thus, the population's risk attitude may act as a control knob for long-run equilibrium selection. In both population settings, we also identify a robust regime: any super-dominant equilibrium is stochastically stable for all risk attitudes, under both protocols, and across populations. We further extend the single-population analysis to symmetric $k$-action games, which include symmetric $k$-action coordination games as a special case, under risk-sensitive best response with mutations. In this setting, we show that, for sufficiently large populations, sufficiently risk-seeking agents uniquely select the strongly payoff-dominant equilibrium when it exists, whereas sufficiently risk-averse agents uniquely select the strongly maximin equilibrium when it exists. These results show that entropic risk sensitivity may serve as a systematic mechanism for steering equilibrium selection in evolutionary games, beyond the classical risk-neutral benchmark.

Advances in tactile sensing have made contact-rich perception possible, accelerating progress in robotic manipulation, material understanding, and embodied interaction. However, because optical design, elastomer mechanics, and imaging geometry differ substantially across tactile sensors, models trained on known sensor types can suffer an abrupt performance collapse on unknown sensors. To address this problem, we propose the Brain-Inspired Few-Shot Tactile Adaptation (BIFTA) framework; it draws on the brain's rapid sensory adaptation mechanism to adapt a frozen encoder to an unknown tactile sensor from a small labeled support set. BIFTA preserves pretrained representations through dual-view statistical memory, constructs support-conditioned spectral graphs to repair sensor-dependent feature neighborhoods, and applies uncertainty-gated recurrent propagation to strengthen reliable cross-query evidence. Extensive benchmarks across three tactile datasets show that BIFTA substantially improves adaptation to unknown sensors: with only 10\% labeled target data on SITR, it raises mean Sparsh accuracy from 6.86\% for the frozen source classifier to 87.09\%, exceeding the strongest implemented prior comparison by 47.22 percentage points, and these gains generalize across datasets, pretrained backbones, and tactile tasks. These results validate BIFTA for data-efficient adaptation to unknown tactile sensors and offer a promising route toward tactile models that transfer across heterogeneous hardware.

Exploration in reinforcement learning (RL) remains a fundamental challenge. Recent goal-conditioned RL strategies (which select goals to encourage broader state coverage) have shown promising results, but none scores a goal by novelty and reachability jointly: the two signals are traded off by hand, applied in sequence, or one is neglected outright. In this paper, we introduce a reachability-aware goal-selection framework that explicitly integrates these two aspects, and that can be seamlessly incorporated into any off-policy RL algorithm. To this aim, we propose SUccessor-to-Novelty (SUN), an indicator derived from successor value functions to identify goals that are both novel and reachable. We prove that SUN recovers count-based bonuses in the limit, bounds short-horizon hitting probabilities, and provably rejects unreachable goals. We further present an adaptive goal-selection strategy that leverages these properties, and an accurate yet lightweight pseudocount to avoid the overhead of classic methods. We back up all our claims with thorough benchmarks: SUN consistently outperforms state-of-the-art methods in standard and novel environments with unreachable or hard-to-reach states, irreversible transitions, obstacles, mazes, and unbounded spaces.

An action chunk can span several stages of a manipulation task, yet a label for its first step describes only the current stage. We introduce Chunk-Aligned Semantic Distillation (CASD), which derives semantic targets for entire action chunks. An offline vision--language model segments demonstrations into described stages. Their occupancy within each action chunk determines a weighted semantic target, including transitions between stages. A CASD generator learns to predict this target from the current observation, robot state, and task instruction. We then freeze the generator and train a policy conditioned on its predictions. The semantic branch runs once per policy query, without online VLM calls or reasoning-trace decoding. Teacher matching on annotated LIBERO training episodes is above chance for both single-stage and boundary-crossing chunks. We evaluate three Fast-WAM variants and a DreamZero integration across four benchmarks, including distribution shifts on LIBERO-Plus. Compared with published references, IDM+CASD reaches 98.9\% versus 98.0\% average success on LIBERO, while Uncond falls below its reference. Joint+CASD reaches 93.0\% versus 90.6\% on RoboTwin 2.0, and DreamZero+CASD reaches a 47.9\% four-category MolmoSpaces manipulation average versus 40.7\%. Performance varies across backbone integrations.

Reconfigurable photonic integrated circuits enable applications in a wide range of fields. They represent a powerful platform for signal processing due to the ability to perform parallel computations. In this work, we demonstrate a photonic processor based on a six-channel universal chip fabricated by femtosecond laser writing with an interlaced architecture comprising multiport beam splitters. We exploit and enhance a previously demonstrated calibration procedure for a single building block, and extend it to a complete six-channel interferometer, reconstructing the corresponding interferometer models over a broad spectral range. We validate them by measuring a set of 100 Haar-random unitary matrices, achieving mean amplitude fidelities of 95.9 %, 98.0 %, and 96.0 % for radiation wavelengths 910, 945, and 980 nm, respectively, without on-device optimization. Furthermore, we experimentally realize spectral demultiplexing and multiplexing at these wavelengths, yielding routing fidelities ranging from 90.1 % to 96.5 % and total crosstalk values ranging from -9.6 to -14.4 dB, confirming the applicability of the fabricated photonic processor to practical wavelength-routing tasks.

Autonomous exploration on uneven terrain requires ground robots to balance exploration efficiency, coverage completeness, and terrain safety. Detailed tsrrain reasoning improves local reliability but can slow large-scale exploration, whereas coarse region guidance expands quickly in open areas but can miss narrow passages and irregular traversable boundaries. To address this challenge, this paper presents TASG-Explore, a traversability-aware sector-guided exploration framework for ground robots. The framework first performs hierarchical traversability analysis using variable-voxel ground fitting and adaptive 8-bit obstacle encoding. It then splitting cost map into sectors, incrementally updates sector clusters, extracts terrain-coupled frontier viewpoints, and maintains a dynamic topological roadmap with unknown topological hypotheses. Finally, a sector-guided planner selects region targets and inserts local viewpoints to generate efficient exploration routes. Benchmark experiments in diverse challenging environments, including caves, forests, and rugged hills, show that TASG-Explore achieves the best overall performance among six representative state-of-the-art planners. The proposed traversability analysis improves processing efficiency by 6.3 times while maintaining high accuracy, and the exploration planner improves exploration efficiency by 51% and increases coverage by up to 2.95 times in rugged hill scene. Large-scale real-world experiments further demonstrate the practical value of the proposed method.

Humanoid motion trackers can reproduce diverse whole-body motions, but their performance degrades on complex terrain where terrain-agnostic references become physically infeasible. We present PGMT, a Perceptive General Motion Tracking pipeline for humanoid robots that learns terrain adaptation from independently selected motion references and terrains. PGMT first learns a general tracking and recovery prior, then incorporates terrain perception through motion-conditioned terrain glimpses that selectively encode regions relevant to the current motion. Terrain-aware tracking relaxation allows necessary deviations from the reference while preserving its motion intent. Zero-shot deployment on a Unitree G1 demonstrates robust terrain-adaptive locomotion and whole-body motion execution over real-world terrain with obstacles up to 37 cm high, while supporting teleoperation, dynamic motion tracking, and fall recovery. PGMT extends general humanoid motion tracking beyond flat ground, providing a unified policy for terrain-adaptive locomotion, diverse whole-body behaviors, and teleoperation in complex environments. Project homepage: https://luyili.github.io/pgmt/

Observing objects grasped by a robot hand is challenging due to severe visual occlusions. Although in-hand manipulation can expose hidden surfaces, existing approaches often rely on predefined or open-loop reorientation strategies that do not explicitly target under-observed regions. We propose AURORA, an active 3D reconstruction framework that closes the loop between online object-centric reconstruction and in-hand reorientation. At its core, Ray-GPIS estimates direction-wise reconstruction uncertainty along candidate viewing rays and selects next-best-view targets using an uncertainty--novelty objective, which are realized through an axis-conditioned in-hand rotation policy. The resulting RGB-D observations are fused incrementally using CAD-free 6D pose tracking and lightweight geometric reconstruction. Experiments demonstrate that AURORA improves reconstruction quality and information-acquisition efficiency over non-active rotation strategies, while Ray-GPIS also outperforms active view-planning baselines in reconstruction performance, action-ranking quality, and planning efficiency. Targeted ablations further validate its robustness to hand occlusion and pose errors. The project webpage is available at https://aurorahand.github.io/

Agentic systems persist model-visible memory while mutating workspaces, while a runtime, registry, or approval service may hold authority state outside both. Identical final files can then require opposite safe actions. We call this the cross-substrate authority gap: decision- relevant authorization information resides outside the planner-visible workspace or memory state. Across two controlled mini-benchmark families, three experiments compare planner-observation augmentation with an execution-time authority check using real Git lineage, durably recorded agent execution attempts, deterministic oracles, and two model routes. Experiment 1 is a 128-cell controlled evidence ablation: authority-blind candidate evidence obtains 0/32 final semantic success, while raw receipts and a typed relation both obtain 32/32. The missing authority fact accounts for the gain; typed packaging provides no observed planning-accuracy gain over equal raw information. Experiment 2 uses 96 planning calls: workspace-visible evidence yields 12/16 unsafe publication decisions, and planning with the typed relation remains unreliable (15/32 first actions correct; 11/32 invalid or absent). Experiment 3 replays the same 32 fixed model-generated first-action intents with zero additional model calls; a deterministic execution guard prevents all six unsafe intents from becoming effects and permits all 12 valid authorized publish intents. These results position authority enforcement at the mutation boundary as the operational endpoint of memory governance.

Large language models (LLMs) and vision-language models (VLMs) have significantly advanced zero-shot task planning for embodied agents. However, most LLM- and VLM-driven methods struggle to generate safe high-level actions due to a lack of physical risk awareness, particularly under partial observability, where hazards lie outside the immediate field of view. To address this challenge, we propose a novel safe task-planning framework, SafeMem, which constructs and maintains a long-term semantic graph memory of the open and dynamic environment. Based on egocentric observations, the proposed framework incrementally accumulates knowledge about surrounding objects and their relationships with a graph. Then, an LLM-based risk predictor evaluates candidate actions using the graph memory, triggering a conservatism-modulated replanning loop with explanations for detected hazards. Extensive experiments on the IS-Bench benchmark and a real-world robot platform demonstrate that the SafeMem framework substantially improves safe success rates compared to state-of-the-art VLM-driven task planners. Video results are available on our webpage: https://sites.google.com/view/safemem.

Aerial Vision-and-Language Navigation requires drones to follow natural-language instructions and navigate through complex urban environments. Accurate navigation relies on both local and global spatial information, which support immediate action grounding and long-horizon path planning, respectively. However, existing zero-shot methods typically operate at a single spatial scale, relying either on local representations constructed online from current observations or on global memories built offline from historical experience. To address this limitation, we propose AirAnchor, a new paradigm that bridges local and global spatial information through spatial anchors and integrates both into a shared navigation framework, enabling comprehensive spatial grounding for decision-making. AirAnchor consists of three core components: (1) Query-Driven Spatial Anchor Grounding, which identifies decision-relevant anchors from visual observations and organizes them into local spatial representations; (2) Persistent Object Spatial Memory, which incrementally maintains an object knowledge base as persistent global spatial memory and retrieves landmark-related spatial priors; and (3) a Spatially-Informed Navigation Agent, which explicitly integrates both local and global spatial information into an agentic framework for decision-making. Extensive experiments on AerialVLN demonstrate that AirAnchor substantially outperforms existing zero-shot baselines, validating the effectiveness and efficiency of the proposed paradigm.

Surface coverage with task-redundant manipulators is challenging because each surface point may admit multiple inverse kinematics (IK) solutions, and configuration choices strongly affect motion quality. This paper extends the classical Spanning Tree Coverage (STC) method to redundant manipulators through offline and online Joint Spanning Tree Coverage (JSTC) algorithms. Offline JSTC samples multiple Inverse Kinematics (IK) solutions per grid cell and formulates the problem as a Generalized Minimum Spanning Tree (GMST), selecting one configuration per cell and tracing the resulting tree to obtain a non-revisiting coverage path. Online JSTC incrementally expands and backtracks a spanning tree with feasibility and cost evaluation while handling dynamic grid updates. Simulation results show that offline JSTC reduces computation time, reconfigurations, and joint motion compared to other methods, while online JSTC achieves fast per-step planning in dynamic scenarios.

In this paper we study performance of the subcarrier-wave quantum key distribution system (SCW QKD) in the presence of spontaneous Raman scattering (SpRS) noise generated by classical channels of dense wavelength division multiplexing (DWDM) network within a single-mode optical fiber. We present the mathematical model for evaluation of the quantum bit error rate (QBER) and the secure key generation rate with the SpRS noise taken into account. We consider two regimes of the SCW QKD system: the continuous wave regime that uses continuous wave laser and the pulsed regime. For these regimes, performance of the system is analyzed depending on receiver sensitivity of classical DWDM. It is found that the pulsed regime outperforms the continuous wave regime in both the secure key generation rate and the maximum achievable distance.

This paper utilizes the autocorrelation of frequency and voltage measurements to identify, quantify and classify local and global properties of power system dynamics. The analysis is based on measurements with various resolutions (20 ms, 1 s, and 1 min) from several nodes of the Irish All-Island Power System (AIPS). Simulations based on stochastic differential algebraic equations on an IEEE benchmark system support conclusions drawn from real-world data.

In this paper we study the performance of subcarrier-wave quantum key distribution (SCW QKD) in the presence of classical channels of optical transport network utilizing dense wavelength division multiplexing (DWDM). We consider the impact of spontaneous Raman scattering noise as well as the four-wave mixing and channel isolation efficiency. We calculate quantum bit error rate as well as the secure key generation rate of SCW-QKD protocol for different parameters of DWDM system and quantum channel allocations. Our calculations show, that quantum channel allocation at the wavelength of 1310 nm is preferable to allocation at C-band due to lower nonlinear noise, despite fiber's larger attenuation at this wavelength.

Persistent communication limits force a multi-agent system to decide which agents may coordinate throughout a rollout. Current proximity alone is insufficient: separated agents may interact later, whereas a large pair reward may remain unreachable until it is heavily discounted. We introduce Reachability-Certified Subteam Decomposition (RCSD) for finite multi-agent Markov decision processes with factorized physical dynamics, finite-range ordered pair rewards, and almost-sure motion bounds. RCSD combines a speed-limit lower bound on pairwise contact time with a reward envelope to form a current-state affinity. For any capacity-valid persistent partition, the sum of cut affinities bounds the reward-deletion error of every unchanged stationary Markov state-feedback policy. A product of team-optimal policies for the resulting cut MDP incurs at most twice this certificate in regret against the centralized optimum. Both bounds are worst-case tight. On a controlled five-agent family, RCSD-Exact reduces aggregate normalized execution regret by 56.0%, 28.8%, and 25.3% relative to uniform, distance-only, and envelope-only partitions. A separate stochastic two-dimensional study finds no bound violation over 384 exact-partition and 1,440 restricted-controller evaluations. Exact four-agent evidence favors RCSD over uniform and distance-only grouping; raw evidence for current contact is borderline and envelope-only is unresolved. Across balanced 8-20-agent strata, controller-library utility is mixed: pointwise paired intervals favor RCSD over distance and current contact, include zero for uniform, and favor envelope-only and Value-MIP over RCSD. Partition construction remains subsecond in median up to 100 agents; this last result does not include affinity formation or MDP planning.

Offline multi-agent payoff models are estimated under a logging distribution but used on distributions induced by learned solutions and unilateral deviations. Standard held-out loss can therefore favor an interaction class that predicts logged play well while distorting strategic incentives. We introduce Selective Interaction-Rank Validation (SIRV) for finite games with known logging distributions. A training split fits nested payoff models and constructs a common union of all candidate deployment and unilateral-replacement distributions; an independent calibration split evaluates every candidate on this same union. SIRV returns the smallest rank whose simultaneous upper worst-target risk is within tolerance of the best upper score, and abstains when a declared target is unsupported or too imprecisely estimated. A common coverage event yields a finite-candidate target-risk bound and a candidate-specific coarse correlated equilibrium (CCE) gap certificate. We also isolate an exact two-point off-support non-identifiability result. In a controlled factorial study with 2,048 independent games per family, empirical-Bernstein bounds reduce the median CCE-gap certificate by 42.5% relative to Hoeffding bounds on common returns, with a 1.36-point reduction in supported return. Under paired rank misspecification and in a separately generated congestion family, the SIRV-EB fallback rule lowers mean true candidate-selection CCE regret relative to ID-Mean, while retaining game-level losses. Across 384 games at $N=3,5,8$, ID-Mean-relative mean CCE-regret effects stay positive while certified return falls sharply under weak coverage. These results separate certifiable model selection from universal strategic improvement.

Bimanual manipulation policies require large and diverse training datasets, yet collecting demonstrations on physical robots is expensive and difficult to scale. Simulation can generate data efficiently, but existing pipelines typically operate within closed asset libraries and predefined scenes: adding a newly observed object or environment still requires substantial effort to reconstruct geometry, specify physical and semantic properties, annotate interactions, and integrate the result into executable tasks. We present RoboCousin, an extensible simulation-based data-generation platform that turns user-provided observations into reusable assets, scenes, and expert trajectories for bimanual manipulation. Built on RoboTwin~2.0, RoboCousin converts object images into simulation-ready assets with visual and collision geometry, semantic and physical metadata, and automatically generated grasp-contact candidates. It further constructs digital cousins that vary compatible objects, backgrounds, layouts, and language instructions while preserving task-relevant affordances and spatial relations. The same asset system supports tabletop and room-level scene construction, with collision-aware base control for interaction beyond a fixed workspace. We release RoboCousin-OBD, containing more than 3,000 annotated object instances and 50 background environments, and use RoboCousin to generate over one million expert trajectories across 50 tasks. Simulation and real-robot experiments show that the automatically generated interaction annotations are comparable to curated annotations, generated assets provide effective sim-to-real supervision, and tabletop cousins can improve transfer beyond training on a single reconstructed scene. RoboCousin therefore provides a practical path for expanding both the scale and coverage of synthetic bimanual manipulation data.

Lifelong navigation (LN) requires an embodied agent to solve a sequence of navigation subtasks in the same environment. Since solving each subtask from scratch incurs redundant exploration, an LN agent must consolidate experience from earlier stages and reuse it in later stages, often through persistent scene representations such as scene graphs or visual snapshots. However, existing approaches typically assume a stationary environment, whereas in real-world LN settings, human activities can cause the environment to evolve. With the stationary assumption violated, existing methods may fuse outdated prior observations with new observations, yet current benchmarks cannot reveal this failure mode. In this paper, we present EvoNav-Bench, which extends the GOAT-Bench style LN formulation in the context of evolving environments. Built on the ProcTHOR framework, EvoNav-Bench introduces environment modifications between navigation tasks, making prior experience useful but not fully reliable. This design enables controlled evaluation of how environment evolution affects LN agents that reuse prior scene observations. Using EvoNav-Bench, we benchmark three recent methods that build and reuse scene representations for navigation. We also compare three simple heuristic strategies for handling environment evolution: Frontier-Update, Fail-then-Update, and Stage-Reset. Our results show that existing methods are brittle under environment evolution, while the heuristic strategies enable a controlled analysis of how agents can adapt to scene changes and mitigate their impact.

In multi-robot collaboration, task handovers rely on downstream verifiers performing remote attestation, which inspects sensor telemetry to ensure a robot's physical behavior strictly matches its assigned task. But can this telemetry be trusted? We show that it often cannot. In this paper, we uncover a severe vulnerability in Robot Operating System (ROS) 2: by modifying a single environment variable, an adversary can execute a pre-built hook to covertly intercept and inject both telemetry and control signals before they are published. Consequently, adversaries can hijack a robot to perform dangerous tasks while spoofing downstream verifiers with synthesized fake telemetry. Worse still, by exploiting the widespread reliance on third-party Docker containers and auxiliary tools, attackers can distribute compromised packages embedded with these malicious hooks to launch such attacks easily. On a physical Franka Emika robotic arm running Secure ROS 2, our attack injects fabricated telemetry in real time with only around 3 ms of jitter, preserving temporal synchronization and hardware integrity while achieving an 87% success rate even against an AI-based detector. We have responsibly disclosed these findings to the ROS 2 development team. We prepared a demo video available at https://youtu.be/ExeiGqUrnhQ.

How far a pushed object slides depends on its mass and friction, which no single image reveals. Pretrained visual encoders are increasingly used as the perception front end of world models for manipulation, and their physical competence is assessed with perturbation benchmarks and linear probes, almost always in a clean, fixed-camera scene. We show that these assessments cannot distinguish an encoder that infers physics from one that does not. CALIPER (calibrate, then predict) is a direct test: an object of unknown mass and friction is struck twice at known speeds, a third strike is shown only up to the moment of contact, and a linear readout on frozen features must predict how far the object slides. Swapping in another object's calibration clips checks that the evidence is actually used. Across 2,000 simulated episodes and eight representations, from V-JEPA 2 to a randomly initialised ViT and raw pixels, calibration adds +0.50 R^2 and the swap removes it. Yet in the clean scene every representation lands within 0.02 R^2 of the ceiling set by true simulator state, because a fixed camera exposes the object's displacement directly in pixel coordinates. Resampling camera, lighting, and clutter for every clip spreads the same representations across 0.50 R^2; when the readout chooses a push speed for a goal distance, V-JEPA 2 misses by 4 mm and the random ViT by 20 mm, no better than ignoring the object. Linear probes track none of this: a change in frame aggregation moves a probe more than pretraining does, and erasing the probed mass direction from the same representation costs nothing in one scene and 0.35 R^2 in the other. Whether a benchmark can rank models is an empirical property, and we give three checks that establish it.

Intermediate representations are key to bridging the modality gap between generalizable manipulation policies and large-scale pretrained vision-language models (VLMs). Among these, trajectory-based representations compactly represent motion-relevant cues, yet most existing approaches predict trajectories in 2D image space, resulting in intrinsic 3D ambiguity. Moreover, using 2D trajectories with depth still leaves the free-space waypoints ambiguous, limiting reliable 3D reasoning. To address this, we propose predicting 3D consistent waypoints (3DWay) from multi-view images. By reformulating 3D waypoints prediction as generating multi-view consistent 2D waypoints followed by geometric triangulation, we enable explicit 3D motion specification while preserving the strong priors of pretrained VLMs. The predicted waypoints can guide existing VLA models for better generalization or be directly executed on simple tasks. Extensive experiments show that 3DWay substantially improves 3D spatial grounding and vision-language reasoning, demonstrating strong potential for generalizable robot manipulation. Codes will be released at https://github.com/ziqin-h/3DWay.

Large language models (LLMs) have recently emerged as a promising tool for automating robot design from high-level specifications, yet they remain ineffective for robots operating under complex physical interactions. This limitation stems from the gap between language-based reasoning and the physical consequences of embodiment, often resulting in designs with low physical validity. In this work, we propose a multi-layered framework, AID-SR, that establishes a closed loop by translating simulator-observed physical states into structured feedback for the LLM designer. Combined with semantic critique, human feedback, and iterative refinement, the framework promotes the generation of physically feasible and functionally meaningful robot designs. We evaluate our approach on tendon-driven continuum robots across a benchmark of 14 tasks spanning reaching, grasping, locomotion, and manipulation. The proposed framework achieves 96.2% rate for passing the simulation feasibility check and by applying a common reinforcement learning training, 26.7% robots can successfully fulfill the corresponding task. We then fabricate three designed robots of AID-SR that successfully complete the task in real-world. These extensive experiments across simulation and real-world environments demonstrate and break the wall of utilizing the LLMs for automated design of continuum robots. The source code and experimental resources are publicly available at https://github.com/UNITES-Lab/AID-SR.

VLMs have shown promise for autonomous driving, yet still suffer from hallucination, weak spatio-temporal perception, and limited generalization. Recent methods improve reasoning and decision-making through CoT explanations, retrieval-augmented generation or the static injection of tool outputs. Although these mechanisms enrich the context, the model neither proactively perceives scene information nor accumulates experience after answering. To overcome these limitations, we present, to our knowledge, the first synergistic framework that tightly couples hierarchical memory with proactive tool invocation in a closed reasoning loop. Our contributions are threefold. (i) Hierarchical Driving Memory: a scene-level short-term memory maintains the dynamic scene state, and an evolving long-term memory retrieves reusable experience and tool strategies. (ii) Memory-Tool Synergistic Reasoning Framework: guided by the scene state and retrieved experience, the model adaptively invokes tools to refine its reasoning at inference time and consolidates reusable experience into a long-term memory pool offline. (iii) Data Generation and Two-stage Training Pipeline: verified memory-tool trajectories built by multi-step teacher rollout are used to train with SFT and GRPO. Our 7B model reaches an overall reasoning score of 80.03 and MCQ accuracy of 79.09% on DriveLMM-o1, surpassing the strongest baseline by 7.74 MCQ points and generalizes strongly across benchmarks. Notably, ablation and analysis studies validate the effectiveness of each component and further reveal the complementary roles of hierarchical memory. Short-term memory strengthens spatio-temporal understanding, improving STSBench accuracy by 24.2 points, while offline long-term memory consolidation yields an additional 3.57-point MCQ gain with all parameters frozen, demonstrating continual self-evolution through accumulated driving experience.

Learning from Observation (LfO) is a fundamental robotic capability that replicates how humans and animals socially learn from each other. Beyond its biological parallels, this modality provides a practical solution for data scaling in sample-inefficient and data-starved domains like robotics. Recent work has demonstrated promising results in learning manipulation skills from human videos, yet progress in this area remains difficult to assess. Existing methods vary widely in assumptions, hardware choices, and environment setups making it difficult to draw meaningful comparisons and identify advances in the field. To address these challenges, we introduce RoboReel: a unified benchmark for evaluating models that learn policies from human videos. RoboReel consists of bundled real-world human demonstration videos, simulated robot trajectories, and evaluation environments on ten manipulation tasks. We develop four test suites to evaluate the models' performance on multiple axes, including the robustness to visual distractors and the ability to complete long-horizon tasks. Our benchmark covers learning-from-observation models from different categories, and studies the effectiveness of multiple representation choices in our benchmark evaluation that covers over seven state-of-the-art algorithms (including our VLA based variants) in the field of LfO. Finally, we present an analysis of the different types of algorithms showing that long-horizon tasks and tasks with low tolerances are still challenging for current models. Webpage: https://roboreel.github.io

Aerial Object Goal Navigation (ObjectNav) requires an unmanned aerial vehicle (UAV) to locate a described target in an unknown outdoor environment using onboard visual observations. Vision-language models (VLMs) can interpret open-ended target descriptions and visual observations, but their frame-level outputs are often noisy, sparse, and spatially transient. We propose AeroBelief, a dual-layer semantic-spatial belief mapping framework that transforms transient VLM observations into persistent spatial guidance. It separates broad contextual plausibility from target-specific evidence: an intuition layer accumulates scene-level semantic cues for exploration, while an evidence layer preserves qualified target-specific observations for approach and confirmation. Evidence-gated fusion combines the two layers into spatial belief hotspots. We further introduce object-conditioned visual reasoning with conservative evidence qualification to improve observation reliability before spatial accumulation. In parallel, egocentric regional guidance converts quadtree coverage into UAV-centered, yaw-aligned directional proposals and stabilizes them through temporal commitment. Its regional scoring is independent of semantic belief values, maintaining exploration pressure and reducing repeated low-gain search. Experiments on the UAV-ON benchmark show that AeroBelief achieves the best reported overall SR, OSR, and SPL among the compared methods, reaching 21.61%, 35.57%, and 10.62, respectively. These results support the effectiveness of persistent semantic-spatial belief, conservative evidence qualification, and temporally stable regional guidance for aerial ObjectNav.

Current LLM-based multi-agent systems (MAS) periodically compress intermediate states to reduce inference-time token consumption, thereby attempting to incorporate more agents. However, naive scaling strategies face challenges. For example, in economic simulations, large-scale MAS typically discard semantically rich economic states, i.e., agent behavioral trajectories, which are key drivers of macroeconomic fluctuations. In this paper, we reveal a phenomenon in which agent heterogeneity gradually decreases during simulation, and propose Prospect-State Propagation for Multi-Agent Systems (PspMAS). Inspired by prospect theory, PspMAS decouples each agent's micro state into a compact Prospect State and an expressive Semantic State. The former records psychological traces through a lightweight, parallelizable propagator and continuously injects heterogeneity into the system. The latter leverages the strong perception, reasoning, planning, and decision-making abilities of LLMs. These two components work complementarily, providing a scalable LLM-based multi-agent simulation solution.

The ability to precisely shape the properties of light over multiple degrees of freedom constitutes the foundation of modern photonic architectures. Broadband sources with addressable spectro-temporal content are for instance critical for applications spanning biomedical imaging, material processing, lidar, and ultrafast spectroscopy, but also key enabling technologies in quantum information processing, high-capacity communications, and optical computing. Yet, flexible and adjustable spectro-temporal shaping so far remains a significant challenge, as it requires the simultaneous manipulation of both frequency and temporal domains across broad spectral bandwidths and extended timescales, from femtoseconds to nanoseconds. Importantly, conventional approaches to spectro-temporal processing typically rely on bulky and complex optical systems that lack scalability and flexibility, ultimately hampering their utility for applications requiring addressable multiphotonic processes. Here, we present a framework for spectro-temporal wavepacket control over a broad bandwidth, merging programmable integrated photonics with high-speed optical characterization. Leveraging machine leaning for tailoring nonlinear pulse propagation, we experimentally report scalable and on-demand shaping of broadband ultrafast pulse patterns. We demonstrate the efficacy of our method through reconfigurable power across 400 nm bandwidth with picosecond resolution, with promises for applications where precise control over light is paramount, in particular those needing versatile and controllable multiphoton excitations.

We take a new look at the relation between finite-horizon trajectory optimization and Schrödinger bridge sampling. Viewed as inference, KL-regularized trajectory optimization is solved by sampling from a Gibbs--Boltzmann distribution whose energy is the trajectory cost, and the adjoint Schrödinger bridge sampler (ASBS) is a simulation-free diffusion sampler designed for exactly such unnormalized targets. Hard equality path and terminal constraints, by contrast, confine the admissible decision variables to a measure-zero feasibility manifold, on which the target must be redefined intrinsically. In particular: $(a)$ we analyze two complementary parametrizations; a rollout parametrization, in which the dynamics are eliminated and only the remaining constraints shape the manifold, and a double-shooting parametrization, in which states and controls are sampled jointly and the dynamics themselves become part of the manifold; $(b)$ we establish regularity conditions under which both admissible sets are smooth embedded manifolds; $(c)$ under compactness and path-connectedness assumptions, we sample from the resulting intrinsic Gibbs measures via Riemannian ASBS, treating strict inequalities through exponential slack variables. Experiments, including contact-rich locomotion and manipulation, demonstrate the effectiveness of both regimes.

Mechanical metamaterials with integrated bistable elements have emerged as promising platforms for mechanical information storage, where transitions between stable states encode information as mechanical bits. While it has been shown that information can be written into such metamaterials by applying global inputs, existing readout strategies rely predominantly on visual inspection. Here, we experimentally demonstrate a mass-in-mass bistable metamaterial with state-dependent stiffness that enables both writing and reading of mechanical information using only boundary-applied dynamic excitations. We further show that such metamaterial architecture functions as both a mechanical sensor of input amplitude and a reconfigurable wave-control device. Together, these results highlight the versatility of bistable metamaterials as multifunctional platforms that integrate mechanical memory, sensing, and adaptive wave manipulation.

Long-form narrative-to-film generation requires shot-level controllability and cross-clip consistency in both visual identity and character behavior-requirements that remain difficult to satisfy with current prompt-based workflows. A core reason existing workflows remain brittle is the lack of a structured intermediate layer between scripts and video models, especially when screenplays are underspecified at key cinematic decision points. We introduce a structured orchestration layer for film-oriented script-to-video generation, implemented as a multi-agent framework that operates between scripts and off-the-shelf video generators. The layer is centered on FilmDSL, a film-oriented domain-specific language that makes cinematic constraints explicit, including shot and camera directives, asset and continuity requirements, and persona cues, so that agents coordinate through a shared structured specification for planning, generation, critique, and repair. Specifically, a generation agent constructs asset packs and storyboard keyframes that anchor composition before clip-by-clip synthesis, while a critic agent produces structured QA signals and triggers targeted refinement without retraining the base model. Experiments on TV-style segments show improved controllability and consistency over text-only and reference-only baselines.

We study online control of a scalar storage system with nonnegative adversarial resource arrivals and state-dependent action constraints. The adversarial, time-varying cost depends on both state and action. At each time, the controller selects a feasible action before the current resource arrival and cost are revealed. To handle the coupling between feasibility and online learning, we introduce Simplex Disturbance-Action Control (SDAC), whose policies have $H$ nonnegative parameters summing to at most one. Every fixed SDAC policy is feasible by construction. Our online SDAC controller updates these parameters using entropic online mirror descent while preserving feasibility along the time-varying trajectory. We prove a regret bound of $O\!\left(\sqrt{T\log(H+1)}\right)$ relative to the best fixed SDAC policy and a minimax lower bound showing that the dependence on $T$ is tight. We also introduce infinite-memory SDAC policies, which include every feasible fixed-fraction policy. SDAC policies approximate this class with an error that decreases geometrically with $H$. For a suitable choice $H=Θ(\log T)$, the resulting regret against infinite-memory SDAC policies is $O\!\left(\sqrt{T\log\log T}\right)$. The framework applies to energy-harvesting batteries and other storage-constrained resource systems.

Multi-agent LLM pipelines increasingly span organisational boundaries; when a fault surfaces, someone must determine where it entered. The artifact available is rarely a full execution trace: it is the reports each agent filed, and a filed report can state a conclusion alongside its observations. Using a pre-registered, institutionally partitioned pipeline of six agents with process-level information boundaries, balanced defect injection and matched clean twins (345,600 requests per chain model, two models), we first report that our pre-registered hypothesis -- that collective responsibility framing degrades escalation with chain length -- is not supported. The layer nevertheless fails asymmetrically. It originates almost nothing: zero allegations across 7,996 clean episodes where every agent stayed silent. It filters upstream error poorly, naming an innocent party in 34.4% and 62.6% of clean episodes where an agent raised a false alarm. Conditional on no agent proposing the true origin (59.5% of episodes on one chain model), an auditor reading the reports recovers it in 4.1% of cases -- below a uniform guess (20%) and the best fixed-link accuser (31.0%) -- while reaching 60.3% from the raw documentation of the same episodes. Deleting one clause, the field carrying the agents' own conclusion, isolates the cause at constant observations: accuracy rises to 45.2% (+41.2 pp, 95% CI +35.3 to +46.9) and adherence collapses from 94.4% to 3.4%; where the suggestion was correct the same deletion instead costs accuracy, 70.5% to 55.7%. The harm replicates on two frontier auditors in four conditions out of four (+8.5 to +39.0 pp) and in a second domain (+47.7 and +61.1 pp), where the cost disappears. The net effect is governed by upstream reliability together with both conditional magnitudes. An accountability layer needs evidence sufficiently independent of the conclusions it verifies.

Hyperbolic shear polaritons (HShPs) represent a distinctive class of anisotropic polaritonic states that enable directional subwavelength light manipulation through asymmetric momentum-space dispersion, while their control in twisted systems has so far relied mainly on geometric and spectral parameters. Here, we introduce non-Hermitian dissipation as an additional degree of freedom for tuning HShPs in twisted van der Waals heterostructures. In twisted α-MoO3/α-V2O5 heterostructures with closely matched isofrequency contours but distinct dielectric losses, dissipation modifies the complex hybrid polaritonic states and thereby tunes the resulting momentum-space and real-space asymmetry. Near-field infrared imaging combined with analytical modeling and full-wave simulations reveals the evolution of the shear response with twist angle, excitation frequency, layer thickness, and substrate dielectric environment. The interplay between material dissipation and dielectric screening enables continuous regulation of the shear degree and reversible switching between shear-off and shear-on states. Furthermore, we demonstrate a proof-of-concept binary-state encoding scheme based on dissipation-controlled shear-state switching. Our work identifies non-Hermitian dissipation as an additional control parameter in twisted polaritonic systems, expanding the available design space for reconfigurable mid-infrared nanophotonics.

Environments are increasingly populated by multiple robots performing independent tasks with limited prior knowledge of each other. Deploying such multi-agent systems presents significant challenges. Specifically, shifts in deployment states compared to training data can lead to poor policy performance and compromised safety. While safety shields exist to mitigate these risks, they are typically reactive, which degrades performance near unseen obstacles,and centralized, limiting their scalability. To address this, we propose a decentralized framework that integrates predictive shielding with model-based finite horizon Q-learning. This approach allows agents to safely adapt their pre-trained policies during deployment. Furthermore, to mitigate livelocks in symmetric scenarios, we introduce a communication- free protocol for conflict resolution

This work addresses a hardware constraint in reservoir computing: the limited size of the readout layer imposed by systems with a physical readout. We investigate a strategy to accommodate this constraint based on random projection, which compresses high-dimensional reservoir states into a lower-dimensional subspace while preserving key properties of the source space and information- processing capabilities. To evaluate this approach, we compare a small, standalone time delay reservoir against a larger configuration whose output is projected down to match the same restricted readout dimension. Using task-independent metrics, we demonstrate that the distribution of information-processing capacities may differ between the two configurations, even at identical readout sizes. Furthermore, we perform a comprehensive hyperparameter scan to assess how both systems behave under varying physical regimes. Finally, we benchmark this approach on the standard NARMA10 task, showing that the random projection framework can yield superior performance compared to a standalone constrained reservoir, within specific compression range. These results provide a scalable pathway to bypass physical readout bottlenecks in hardware-based reservoir computing.

Quantum game theory is an extension of classical game theory that uses quantum principles in game theory. The Eisert-Wilkens-Lewenstein (EWL) quantum game is an early example of the two-player classical Prisoner's Dilemma transformed into a quantum Prisoner's Dilemma. In the EWL game, the players choose pure quantum strategies represented by unitary matrices. This extension can resolve the classical dilemma by enabling cooperative equilibrium with higher payoff. In this paper, we first discuss the Extended EWL (EEWL) for multiplayer quantum games with mixed strategies. In EEWL, each player controls a set of unitary operators as quantum actions and uses a classical mixed strategy over these actions. The payoffs are defined as expectation values of Hermitian reward operators acting on a shared quantum state, which is generated and measured according to the EEWL protocol. We then propose the Unitary Strategy Matrix Exponential Algorithm (USMEA), a geometry-aware sequential algorithm for the EEWL mixed-strategy setting, in which each player jointly learns a trainable set of local unitary actions and the associated classical mixing probabilities. Thereby it acts as a learning-and-control layer for multi-agent quantum decision systems. We analyze the convergence properties of USMEA under standard smoothness and step-size conditions and validate the theory with numerical experiments. These results show how classical optimization methods can be systematically integrated into the design and analysis of engineered quantum strategic interactions.

Climate change is increasing the severity and unpredictability of natural disasters. In time-critical crises such as wildfires, traditional monitoring practices remain limited by coverage, cost, and personnel risk, paving the way for autonomous and adaptive monitoring solutions. Within this context, this paper introduces D3ARC, an asynchronous distributed hierarchical framework for time-aware and reliable wildfire detection. D3ARC integrates multiple robotic agents that cooperate under uncertainty through distributed perception, shared situational awareness and coordinated actions. A remote controller asynchronously decides upon each robot's motion, while each robotic agent senses the environment and decides where and how to execute the wildfire detection. All robotic operations require time, and as time progresses, wildfires continue to spread, reducing the opportunity for early intervention. As such, all agents share a common objective: to detect a wildfire with a certain performance threshold as fast as possible and within a time limit. D3ARC integrates mechanisms for safe navigation, coverage efficiency, cooperation and reliability. It introduces a forward-looking capability that allows agents to anticipate the future by evaluating candidate strategies before execution. The framework is evaluated through realistic robotics simulations, ablation studies, and baseline comparisons, achieving an overall mission success up to 94% with 89.4% detection confidence.

Although Ising machines (IMs) have grown in popularity due to their potential as physics-based accelerators for combinatorial optimization problems (COPs), most literature focuses on empirically-obtained benchmarks with low universality. In this work, we propose dynamic feedback distortion as a universal metric to quantify the effect of coupling field discretization in time-multiplexed IMs. Starting from the transfer functions of realistic hardware, we analytically show how this distortion can be mitigated using a global bias (or choosing the correct discrete coupling element) and injecting coloured noise. Our numerical results show that our mitigation protocols are very effective, especially in complicated energy landscapes (namely tunable Möbius ladders and 2D tiling problems); where an analogue reduction of the distortion leads to a two--three order of magnitude increase in the success rate of complex COPs.

Local pedestrian-vehicle forecasting spans heterogeneous physical scales: pedestrians combine root locomotion with articulated motion, whereas vehicles are rigid bodies described by kinematic state and oriented extent. Existing road-agent forecasters typically omit pedestrian articulation, while pose forecasters leave vehicle futures outside the learned rollout. We introduce PV-WM, a history-only world model over structured post-perception tracks. It recurrently advances pedestrian root motion, 15-joint articulation, and learned vehicle states within a synchronized heterogeneous state. The generated pedestrian and vehicle chunks supply the next recurrent boundary; vehicle boxes are reconstructed from predicted center and heading with observed extent, and P-V geometry is recomputed after every transition. Relative to a matched one-shot complete-state predictor, recurrent execution reduces Root ADE by 12.7% and MPJPE by 14.8%. Feedback interventions show that later predictions depend on the content, temporal order, and pedestrian identity of generated articulation. Across 824 aligned Waymo contexts, with 797 providing valid future vehicle support, PV-WM reduces Root ADE by 5.2%, MPJPE by 7.6%, P-V distance error by 11.9%, and oriented-box closest-approach error by 5.8% relative to a validation-selected Modular Specialist. The single-network model uses 57.1% fewer parameters, 96.5% lower average FLOPs per local scene, and 25.5% lower measured p95 latency. PV-WM unifies this heterogeneous future state while preserving type-specific pedestrian and vehicle dynamics.

Tunable photonic architectures that improve the extraction efficiency of light from embedded epitaxial quantum dots across a wide spectral range are key enablers for developing bright sources of single and indistinguishable photons. In this study, we experimentally demonstrate planar multilayer antenna structures consisting of epitaxially grown AlGaAs and InGaAs membranes containing quantum dots sandwiched between metallic Au (or Ag) reflector and director layers together with Al$_2$O$_3$ spacer layers. We show that the linewidths and fine-structure splitting of the neutral exciton emission remain comparable to those measured in the corresponding unprocessed samples, demonstrating that the fabrication process preserves the optical quality of the emitters. In addition to broadband operation, we demonstrate that the planar architecture is compatible with electrical tuning via integrated diode structures, and strain tuning using piezoelectric actuators. In spite of limitations related to optical losses in the ultrathin metallic layers, the demonstrated fabrication simplicity, scalability, and compatibility with tunable quantum emitters establish planar antennas as a promising platform for solid-state quantum photonic devices.

High-fidelity single-shot readout of superconducting qubits is essential for fault-tolerant quantum computation. Without a quantum-limited amplifier such as a Josephson parametric amplifier, the cryogenic high-electron-mobility-transistor (HEMT) low-noise amplifier (LNA) at the 4 K stage is the dominant noise source in the readout chain. HEMT LNAs are conventionally characterized by Y-factor measurements of the noise temperature $T_N$, independently of the quantum measurement they ultimately serve. Here we present a qubit-in-the-loop benchmarking method. The method uses the two-dimensional IQ histogram of single-shot readout as a direct figure of merit. We apply it to three state-of-the-art cryogenic InP HEMT LNAs with 53\%, 60\%, and 70\% channel indium content, together with a commercial reference amplifier. The 60\% and 70\% devices have similar $T_N$ in Y-factor measurements, both lower than the 53\% device. The 70\% device has the highest gain, but qubit readout SNR and assignment fidelity $\mathcal{F}_a$ instead peak at 60\%. Using this method, we map $\mathcal{F}_a$ against HEMT LNA dc power consumption for each device. $\mathcal{F}_a$ does not simply saturate with increasing power. For several devices it declines above a device-dependent power well below typical operating points. For the 60\% device, which has the highest overall $\mathcal{F}_a$, about 1~mW is needed to maintain $\mathcal{F}_a > 85\%$. Across all four devices, $\mathcal{F}_a > 80\%$ is reached with as little as about 0.3~mW, roughly an order of magnitude below typical operating points. These results establish a qubit-referenced benchmarking methodology and provide practical guidance for amplifier selection and bias in power-constrained multi-qubit readout systems.

This paper addresses the mathematical modeling and control of an aerial manipulator system comprising a quadrotor as the uncrewed aerial vehicle and a robotic arm as the manipulator. The dynamic model is established by identifying the overall center-of-mass velocity and the system's orientation as constraints, yielding a simplified, two-decoupled subsystems: a locked (overall translation) subsystem and a shape-space (actuation or overall rotation) subsystem, both subjected to external disturbances. Since the system is mechanically coupled, a critical challenge arises where constant bounded disturbances in the first subsystem manifest as time-varying, state-dependent disturbances in the second subsystem. Given that the quadrotor is inherently unstable, the movement of the robotic manipulator (RM) during flight can further jeopardize the stability of the entire QRM system if these disturbances and coupling effects are not effectively managed. To address this, we present a continuous nonlinear disturbance observer-based feedback control law, which enables the independent control of each subsystem while systematically eliminating cross-coupling effects. The efficacy of the proposed controller is validated through multiple simulations emulating practical operating conditions, thereby substantiating its real-world applicability and highlighting the core contributions of this work.

Generalized Nash Equilibrium Problems (GNEPs) often arise in multi-agent engineering applications that require distributed algorithms. Unlike traditional approaches that enforce consensus on multipliers, our method removes the need to share multipliers, reducing communication and improving privacy. As a result, different initializations can lead to different GNEs, including non-variational ones. We establish convergence under sufficient conditions using an input-to-state stability (ISS) framework.

Modern power systems are growing increasingly complex as they integrate diverse generation sources to meet rising demand, making accurate load forecasting challenging. Recent advances in time-series foundation models (TSFMs) resulted in promising performance in zero-shot univariate load forecasting tasks. However, real-world load forecasting often involves multiple target variables and requires the integration of exogenous variables, raising important questions about the utility of TSFMs in realistic settings. In this study, we position Chronos-2, a recently developed model by Amazon, as a representative multi-channel TSFM that supports univariate, multivariate, and covariate-informed forecasting, and conduct a systematic investigation of how such models can be used for real-world load forecasting. While prior work has evaluated Chronos-2 on a limited number of energy-related tasks in a zero-shot setting, its performance relative to established task-specific deep learning models and its behavior when adapted using task-specific historical data remains insufficiently understood. In this work, we evaluate Chronos-2 on two real-world utility datasets, ISO New England and ENTSO-E, and benchmark it against widely used task-specific deep learning models. Our results show that Chronos-2 benefits substantially from task-specific fine-tuning and achieves strong short-horizon forecasting performance, but its zero-shot accuracy lags behind task-specific models and its forecasting error grows more rapidly with increasing forecast steps. Overall, this study provides a detailed characterization of the strengths and limitations of TSFMs such as Chronos-2 in grid load forecasting and offers practical insights into how a pretrained TSFM can be effectively adapted for operational load forecasting applications.

We investigate shortcut-to-adiabaticity protocols for the rapid compression of optical solitons in nonlocal media. Using a variational approximation, we derive an effective Ermakov-like equation for the soliton width and employ inverse engineering to design propagation-dependent control parameters. Three control strategies are compared, based respectively on modulation of the characteristic nonlocal length, the Kerr nonlinearity, and the external parabolic confinement. We show that all three protocols enable high-fidelity compression over strongly reduced propagation distances. By comparing control smoothness and final fidelity, we find that the relative performance of the protocols in the very short-distance regime depends on the chosen criterion, revealing a trade-off between implementation smoothness and target-profile accuracy. These results establish a practical and general route to fast soliton manipulation in nonlocal media.

A shared reward gives agents a common objective, but leaves open when, how and even whether they must cooperate to succeed. We address these questions in the Laser Learning Environment, a multi-agent path-finding environment where cooperation materializes as one agent blocking a laser to let a teammate pass safely. We represent these interactions through temporal cooperation graphs whose timed edges connect helpers to beneficiaries, define six cooperation profiles as overlapping graph predicates, and prove that every cooperative trajectory satisfies at least one. By encoding the environment dynamics and profile predicates as propositional formulae, we distinguish tasks that admit}a profile in some winning trajectory from those that require it in every winning trajectory within a specified horizon. Used as filters, these queries turn a random layout sampler into a generator of tasks with certified cooperation requirements. Experiments with five multi-agent reinforcement learning algorithms show that training diversity improves joint success on unseen tasks when cooperation-free solutions exist. When cooperation is required, greater diversity improves individual-agent exits, but joint success remains near zero. Across five profile-certified pools, final exit rates averaged over algorithms separate the pools into four statistically distinguishable levels but this ordering primarily reflects partial completion: policies collect rewards for individual exits but rarely exhibit the profile required for joint success. Our framework exposes this gap between rewarded partial completion and realized cooperation by certifying what cooperation successful completion requires and using temporal cooperation graphs to reveal what policies exhibit.

Energy-harvesting Internet of Things (IoT) devices must decide whether to transmit each observation immediately or accumulate several observations before reporting to the edge. Early reports make information from initial observations available sooner at the edge but require more transmission actions, whereas larger batches save reports while delaying that information and retaining more state. We develop a unified analytical framework to compare this fixed-batching choice in scenarios with repeated, deadline-constrained inference cycles. Every configuration processes the same ordered observations and yields the same final posterior if all reports arrive, enabling a controlled comparison of the persistent state, actions, and energy costs induced by batching. Using reductions in Bayesian risk under the continuous ranked probability score (CRPS), we express long-run timely value exactly as a statistical-value-weighted sum of expected decision-weighted report availabilities. This separates the inferential value of observations from their availability at application-relevant decision times. We construct exact Markov-reward models accounting jointly for intermittent harvesting, finite storage, energy carry-over, unreliable delivery, retransmissions, and deadlines. We prove that the long-run regime is well defined and derive exact batch-size comparisons, timing benchmarks, and a decomposition of harvesting-law effects into within-cycle and carry-over contributions. Directly evaluable Gaussian and Gaussian-mixture belief specializations show that the preferred batch size can change with the prior, the decision-time profile, link reliability, harvested-energy statistics, and accumulation cost.

The evolution of modern-day communication networks towards optical solutions with enhanced capacity and robustness is driving interest in topological light waves, exploiting their stability against perturbations through a topological invariant, e.g., the skyrmion number. However, detecting the underlying topology remains a computationally intense process even under ideal conditions, becoming intractable after passing through strongly disordered channels, where the degradation into unrecognisable speckle appears to destroy the topology. Here, we propose and demonstrate a topology-enhanced artificial intelligence (AI) approach to recover and classify such apparently lost topological information by computationally leveraging topological invariants in the data across many length scales. By aligning the topological classification of information with the topology of light, our topology-enhanced learning protocol, termed TOPO$^{2}$, achieves highly efficient recognition of the topological states of light, even from speckle, without the need for any prior learning. Our approach outperforms benchmark tests against standard computational algorithms and has the benefit of requiring just a single intensity pattern as the input, facilitating single-shot operation. To demonstrate this, we leverage the skyrmion number as a robust data carrier of images through a disordered channel, using TOPO$^{2}$ to accurately reconstruct the transmitted images. This work synergises topological photonics and topological AI for unravelling hidden topological signatures in light, opening a pathway towards robust communications even in extreme disordered environments.

This work presents a control framework that enables a multicopter equipped with a two-axis gimbal to emulate the flight dynamics of a fixed-wing aircraft. The goal is to provide an operationally simple platform for training and simulation that avoids the aerodynamic constraints of fixed-wing vehicles, such as minimum airspeed and nonholonomic constraints. A state-input mapping between the two platforms is derived using dynamic feedback linearization. The framework is evaluated on representative fixed-wing manoeuvres. Results show high-fidelity emulation under nominal conditions. While evaluations are conducted in simulation, the approach establishes a practical path toward hardware deployment for pilot training, autonomy research, and controller benchmarking.

Agentic systems require secure and efficient delegation of privileges across multiple collaborating agents. Existing approaches fall into two categories. Some propagate user identities directly to agents, obscuring accountability and creating persistent over-privilege risks that are amplified by the non-deterministic behaviour of AI agents. Others rely on continuous synchronization with a central Identity and Access Management (IAM) provider, introducing additional latency and communication overhead. We present CAPMAS, a novel architecture for secure end-to-end query execution in multi-agent systems. CAPMAS newly combines a contrastive learning-based semantic scoping pipeline that maps natural-language queries to bounded privilege sets before execution with expressive Macaroon-based tokens that enable offline, tamper-evident delegation with monotonic privilege reduction across agents. By decoupling authentication and delegation enforcement from agent reasoning, CAPMAS enables practical agentic execution while enforcing strict least-privilege guarantees. By eliminating synchronous delegation exchanges with the IAM, CAPMAS yields 30 times faster delegation operations, 2 times less delegation-oriented latency and up to 3 times lower bandwidth usage than the OAuth 2.0 Token Exchange (RFC 8693). Its semantic scoping pipeline achieves over 90% perfect privilege-bundle retrieval within 17 milliseconds on enterprise-scale API schemas containing over 3,100 endpoints, while reducing unnecessary privileges by 99.5% when compared to systems that propagate all the user's privileges to agents.

Scalable quantum networks require quantum memories that are chip-integrated, telecom-band compatible, and capable of flexible retrieval. Nanofabricated mechanical resonators meet these criteria. They offer independent tunability of optical and mechanical modes, long-lived phonon states, and design flexibility beyond atomic systems, making them strong candidates for practical integrated quantum memory. Here, we demonstrate an on-chip, absorptive optomechanical memory for telecom-band photons, based on optomechanically induced transparency (OMIT) and operating near the mechanical ground state. The device stores telecom-band photons, demonstrating compatibility with external photon sources at the few-photon level, while enabling on-demand retrieval. By placing the device in a dilution refrigerator at 20 mK and tailoring the control field to suppress optical heating, we achieve a remarkably low phonon occupancy of just 0.32 during the storage process. Our results lay the groundwork for scalable, phonon-based quantum memory devices and open new avenues for integrating mechanical systems into practical quantum network architectures.

Networked systems face a tradeoff between the scope of the information behind a decision and its freshness: a broader view of the system supports better coordination, but assembling and communicating it takes time, so it arrives older. We study this tradeoff for a team of agents that repeatedly choose actions to minimize a shared quadratic cost driven by an environment that evolves on its own, unaffected by the agents' actions. An information architecture specifies what each agent observes, from where, and with what delay; we measure an architecture by the cost it loses relative to a decision made with complete, current information. We first compare fresh local observations with a complete but delayed global view. When every component of the environment decorrelates at a common exponential rate, this comparison reduces to a closed-form threshold on the ratio of delay to coherence time. We then study intermediate architectures in which each agent acts on a time-aligned, and therefore older, snapshot of a wider neighborhood. The optimal neighborhood radius occurs where the marginal value of added scope equals the marginal cost of lost freshness. In a canonical spatial model, this radius is set by the spatial-correlation, decision-relevance, and temporal-propagation lengths. Throughout, architecture performance is governed by the predictability of the optimal decision rather than of the raw state.

Robots navigating under occlusion may enter states from which no admissible input can avoid a dynamic obstacle once it becomes visible. We present OcclusionCBF, a safety filter that extends backup control barrier functions to reachable-occupancy predictions for potentially hidden dynamic obstacles in occluded regions. The method certifies a prescribed backup rollout against collision-inflated occupancy and a verified terminal set, yielding affine constraints for minimally invasive quadratic-program filtering. We establish recursive feasibility of the resulting safety filter, and collision avoidance for every hidden-obstacle motion covered by the occupancy prediction. Randomized benchmarks, MetaUrban simulations, and hardware experiments demonstrate improved task success over reactive and occlusion-aware predictive baselines with millisecond-scale computation.

Generalist robots promise to transform our society: the same system that prepares a meal or folds laundry might also repair a car, inspect infrastructure, or care for a loved one. Yet this versatility introduces risks far beyond the collision- and force-based safety notions that have long dominated robotics. Notions of safety must now consider context (e.g., turning off a building's electricity is only safe during scheduled maintenance), user intent (e.g., asking the robot to ``clean the kitchen'' includes unspoken expectations that the robot should not mix dangerous but powerful cleaning agents like bleach and ammonia), hard-to-model physical consequences (e.g., burning food during meal preparation), and more. We argue the need for a new era of robot safety---embodied AI safety---that broadens the hazards considered across the robot's lifecycle while recognizing that the safety of bits cannot be separated from the safety of atoms. We present a taxonomy of emerging risks and a full-stack research agenda to guide the safe deployment of generalist robots.