The growing volume of retired lithium-ion battery packs from electric vehicles and portable electronics calls for automated disassembly that is safe, flexible, and selective down to the individual cell. Existing robotic systems, however, mostly assume known pack poses, external fixtures, or specialised tooling, leaving fixture-free cell-level disassembly under pose uncertainty largely unsolved. This paper presents a vision-guided dual-arm pipeline that disassembles a 21-cell 18650 pack from an arbitrary initial pose using only general-purpose parallel-jaw grippers, RGB-D sensing, and a pre-trained grasp detector. Pose uncertainty is absorbed by a learn-and-filter perception stack with discrete look-and-move wrist-camera corrections, while a mid-task support transfer between the two arms extends the effective workspace without any external clamp. The pipeline achieves an 8/10 end-to-end success rate, a cell-localisation root-mean-square error of $2.4$\,mm, and a mean cycle time of 6.0\,minutes per pack, providing a practical, fixture-free building block for industrial battery recycling.

Tool-using LLM agents often fail not because relevant text is absent, but because decisive evidence is not selected, compressed, or surfaced at action time. We present CICL, a decision-aware context layer that turns instance evidence into a context graph, routes deterministic, Opus-assisted, Qwen, Codex/GPT-5.5, and Qwen-QLoRA judgments through a shared eight-field schema, scores units by action shift, outcome uplift, necessity, and negative-transfer risk, and packs high-utility evidence as typed memory cards for a budgeted agent. The design separates the measured decision signal from the judge model, so frontier annotation, local surrogates, and lightweight rankers can be compared under one auditable protocol. Empirically, CICL yields a concrete open-benchmark gain while exposing its limits. On 50 SWE-bench Verified file-retrieval instances, direct Qwen3.6-plus reranking of BM25 top-50 candidates raises hit@1 from 0.58 to 0.78 and MRR@10 from 0.634 to 0.790, with all 2,500 judgments parseable. Controlled diagnostics show action-criticality: at budget 120, CICL reaches F1 0.620 on v1 and 0.425 on v3, and removing the top-utility semantic v3 unit collapses F1 to 0.000. Supplementary checks add Qwen-QLoRA agreement over 710 candidates, a small 200-label real-code Opus-assisted signal, and a three-instance patch smoke validating retrieval-to-patch plumbing without claiming official SWE-bench success. RepoBench-R summaries still beat cards, and compact rankers do not yet replace the heuristic. CICL contributes a reproducible measurement and selection layer for decision-critical context, not an end-to-end coding-agent repair claim.

Designing 3D metamaterial microstructures that meet the intended functions remains a major challenge, as it typically requires domain expertise, iterative simulations, and extensive manual tuning. Existing work on inverse design that automatically generates microstructures based on desired target properties often suffers from limited design diversity and faces challenges in ensuring the physical feasibility of the generated structures. To address this issue, a property-informed diffusion-based network is proposed that enables the generation of 3D microstructures directly from textual descriptions. Unlike traditional property conditioning methods, our approach leverages rich guidance in terms of semantics and physical properties in the text input to support diverse structure synthesis. To enforce consistency between the generated structures and the target textual prompts, a dual alignment strategy is adopted, including contrastive text-structure alignment and test-time reward-guided alignment. Experimental results show that the model is capable of generating semantically meaningful and physically plausible structures across a wide range of material categories. Our approach has good potential for interactive microstructure design and opens up new directions for combining language-based interfaces with inverse material discovery. Code is available at: https://github.com/hongsong-wang/PropDiff-TMG

Fraud detection in payment, e-commerce, and telecommunications systems requires accuracy at the individual level, robustness under severe class imbalance, and ease of understanding for risk managers. Existing methods fall at least one of these requirements: automated machine learning systems search a fixed numerical space without semantic awareness of the dataset; graph neural network-based methods require pre-defined relational graphs and remain opaque at the individual-decision level; and the design of general-purpose large language model (LLM) agents does not consider the recall and precision constraints specific to real-world fraud detection. In this paper, we propose SAGE, the first end-to-end LLM-driven multi-agent framework for fraud detection. SAGE coordinates three dedicated agents that make decisions based on a six-layer Data Diagnostic Tree (DDT) and a Markov decision process guided by natural-language gradients, automatically optimizing the model under a fraud-specific reward. On five fraud datasets and five LLM backbones, SAGE wins $96.00\%$ of method--dataset comparisons and improves F1 by an average of $40.86\%$ over baselines. The code is available at https://github.com/yichenC1c/SAGE.

Composed Video Retrieval (CVR) is designed to retrieve a target video that matches a reference video modified by a modification text. While existing methods explore cross-modal correspondences, they often assume modified objects appear directly in videos. However, modification texts frequently describe concepts not explicitly presented but implicitly expressed through semantically related visual cues (e.g., "cake" implying "birthday party"). Current approaches typically rely on aligning explicit feature representations within the concrete space, neglecting critical latent associations. To address this, we propose an adaptIve scheMa-ImAGery enhanced composItional NEtwork (IMAGINE). Unlike standard explicit matching, IMAGINE materializes implicit semantics (termed schema imagery) via dynamic multimodal prototypes. These prototypes capture shared latent concepts to adaptively modulate visual features, effectively injecting implicit guidance into the retrieval process. By bridging the gap between explicit visual contents and implicit retrieval intentions, IMAGINE achieves state-of-the-art performance in both CVR and Composed Image Retrieval (CIR) across three widely used benchmarks.

Supply Chain Finance (SCF) and LendTech platforms need credit scoring systems that respond to evolving transaction behavior, repayment delays, and active exposure. We propose TRUST-SCF, a transformer-based framework for transaction-level risk prediction and dynamic credit scoring. Each user history is represented as a sequence of transaction tokens containing utilization, repayment delay and transaction position. The main contributions are: (1) a financially aligned attention bias that combines utilization similarity and recency, enabling the model to compare repayment behavior under comparable exposure conditions; (2) continuous repayment-delay prediction in a log-transformed target space, reducing the influence of extreme delays while improving sensitivity to short-delay behavior and (3) a label-efficient credit-scoring pipeline in which the final credit score is not trained using any explicit external credit-score label, but is instead derived from predicted delay, potential risk over simulated utilization, actual unpaid exposure, and nonlinear calibration. Experiments on real transaction data from more than 300,000 transactions show that TRUST-SCF improves delay prediction over sequential baselines and produces scores that are strongly associated with future repayment behavior. These results suggest that TRUST-SCF is a practical framework for adaptive credit scoring and transaction-level risk mitigation in SCF and LendTech environments.

Model Predictive Control (MPC) is widely used for autonomous-vehicle (AV) motion planning, but its real-time applicability is often limited by the need for accurate models and online solution of nonlinear, nonconvex optimization problems in dynamic road environments. Actor-critic reinforcement learning offers a promising alternative for online policy generation, yet its policy-learning process often lacks explicit control-theoretic structure. This article proposes a learning predictive control (LPC) framework with deep Koopman operators for efficient real-time motion planning under nonconvex constraints. To address nonlinear and uncertain vehicle dynamics, a deep-Koopman-based predictor is used to lift the system into an interpretable linear observable space in a data-driven manner. Unlike traditional MPC, which computes open-loop control sequences, the proposed LPC framework yields a closed-loop state-feedback policy within each prediction interval through receding-horizon actor-critic learning. To ensure safety under nonconvex environmental constraints, LPC constructs convex local surrogate representations of obstacles and defines corresponding potential-field functions. These functions and their gradients are directly embedded into the actor-critic structure, enabling efficient, safety-aware policy learning. Extensive simulations and real-world experiments on the HongQi-EHS3 platform demonstrate favorable performance in diverse obstacle-avoidance scenarios in terms of safety, computational efficiency, and driving comfort, compared with benchmark methods such as CBF-MPC and LMPCC.

Ground contact forces acting on the human body, are crucial for biomechanics studies or sport performance analysis. Prior methods rely on force plates or pressure mats to collect ground contact dynamics, limiting their applicability to carefully controlled settings. A more scalable solution is to estimate the dynamics directly from motion capture data. Recent approaches only roughly estimate the ground contact dynamics from the vertical distance between the body and the ground plane, which cannot capture the complex pressure distribution of all contact points. To this end, we propose GraCE -- Gravity-guided Contact Dynamics Estimation, a novel full-body contact dynamics model for human motions using a realistic influence of body mass distribution and gravity. We use the human's center of gravity to estimate the ground contacts based on its relative distance to the human body. The applied force on each contact is estimated via the product of predicted contact probabilities and the total exterior force computed from the center of mass trajectory. We outperform related work on the GroundLink dataset for ground reaction force estimation, and on the MOYO dataset for detailed contact pressure prediction. The code is published upon acceptance.

Vision Transformers operate on fixed patch grids, which can introduce phase-dependent instability for dense prediction: changing the patch partition can change the token evidence available to a pixel, especially near boundaries. We formalize patch-grid phase as a nuisance variable and propose Phase Marginalization, a post-hoc marginalization method that evaluates structured patch-grid phases, inverse-aligns dense outputs, and aggregates them in the original image coordinate system. The central variant, Uniform Phase Marginalization with K = 4, is training-free and improves over the canonical K = 1 baseline across measured segmentation, depth, and local matching settings. In a controlled Cityscapes experiment, Uniform Phase Marginalization provides a modest compute-matched advantage over generic shift-based four-forward test-time augmentation (TTA) (+0.31 mean Intersection-over-Union over the strongest tested generic row). A scaling study further shows that K = 4 is a practical cost-accuracy trade-off: K = 8 is essentially unchanged and K = 16 adds little accuracy at much higher latency. These results position patch-grid phase as a measurable nuisance variable and Phase Marginalization as a simple diagnostic and post-hoc marginalization baseline for dense ViT prediction.

Large language models (LLMs) differ in architecture, training data, and optimization procedures, yet they may still develop similar internal inference patterns. In this paper, we examine this hypothesis using interaction-based explanations. We find that LLMs often share interaction patterns when predicting the same target token from the same prompt. This consistency is more pronounced among advanced LLMs. Shared interactions also tend to be lower-order and show weaker positive-negative cancellation than non-shared interactions. These results suggest that advanced LLMs may be implicitly optimized toward common inference patterns, even though the mechanisms that give rise to such cross-model consistency remain open.

Vision-language models (VLMs) enable visual recognition from semantic class descriptions, which makes them attractive when target annotations are scarce or unavailable. Most deployment pipelines, however, first choose a single VLM and then adapt that model to the unlabeled target set. This single-backbone paradigm hides a critical assumption: the selected VLM is already compatible with the target domain. In realistic cross-domain deployment, several general-purpose and domain-specialized VLMs may be plausible, yet no instance-level target labels are available to identify the reliable ones. Deployment therefore requires a coupled solution for model selection, target adaptation, and prediction integration. We revisit this problem from a system-level multi-VLM perspective. Our central observation is that the three decisions above depend on the same latent object: a trustworthy sample-class structure in the target set. Different VLMs may encode different transfer biases and produce conflicting predictions, but their outputs can still provide complementary evidence for estimating this structure. We propose One Stone, Three Birds, a training-free framework based on self-adaptive optimal transport. Given a pool of frozen candidate VLMs, OSTB estimates a consensus sample-to-class transport plan without updating VLM parameters. The learned transport structure is then reused for all deployment objectives: model selection is performed by ranking the combined semantic and visual reliability induced by the consensus plan; target adaptation is obtained by fitting transport-conditioned visual classifiers; and ensembling is implemented through reliability-aware probabilistic integration. Extensive experiments on natural-image, remote-sensing, and medical-pathology benchmarks show that OSTB improves model ranking, adaptation stability, and ensemble robustness under heterogeneous candidate pools.

Vision-based driver monitoring systems are increasingly deployed in safety-critical intelligent transportation settings, yet they are almost always compared on classification accuracy alone. This paper argues that accuracy is insufficient to characterize a model's fitness for real-world deployment, and proposes the Human-Centered Benchmarking Framework (HCBF), which evaluates models across four dimensions: accuracy, explainability, efficiency, and robustness. The framework is applied to four representative lightweight architectures, MobileNetV3, ShuffleNetV2, EfficientNet-B0, and DeiT-Tiny, on the MRL Eye Dataset for eye-state classification. While the models are nearly indistinguishable on clean-set accuracy, each leads in exactly one dimension, and all four lie on the Pareto frontier. A Human-Centered Score computed under three deployment-oriented weighting scenarios ranks ShuffleNetV2 first throughout. However, this aggregate winner retains less than half of its performance under sensor noise and fails by classifying closed eyes as open, whereas the transformer remains robust. These findings show that aggregate ranking can mask dimension-specific vulnerabilities that are operationally decisive, underscoring the value of multi-dimensional, human-centered evaluation.

Predicting a user's next Point-of-Interest (POI) based on their historical check-in records is a fundamental task in location-based services. While recent methods incorporating large language models have shown strong reasoning capabilities and promising results, they typically formulate the prediction task as a one-step trajectory-to-location mapping problem, making predictions prone to shallow trajectory correlations and historical frequency bias. We argue that users rarely choose locations directly and instead, they usually first form a traveling intention and then accordingly select specific POIs. Motivated by this insight, we propose IntentPOI, a two-stage intention-guided reasoning framework. In the thinking stage, we infer users' intermediate intentions by incorporating historical mobility patterns, similar peer behaviors, and the temporal contexts. In the acting stage, we first construct a compact candidate pool, and then perform intention-guided reasoning to identify locations that best align with the inferred intention. By explicitly decoupling intention inference from location prediction, IntentPOI transforms the next POI prediction from direct trajectory matching into intention-guided reasoning. Extensive experiments on three real-world datasets demonstrate that IntentPOI consistently outperforms eleven state-of-the-art baselines.

Manipulation understanding requires reliable relational evidence, such as contact, support, containment, motion coupling, grasp, release, and active-hand involvement. Although these visual predicates are widely used in event-chain, graph-based, and neuro-symbolic models, their reliability under visual degradation is rarely analyzed directly. This paper introduces a predicate-level reliability framework for robust manipulation understanding under blur, occlusion, illumination change, low resolution, frame dropping, and detection noise. The framework defines a structured predicate vocabulary, confidence-aware predicate estimation, and reliability metrics for predicate preservation, degradation sensitivity, temporal consistency, confidence-weighted stability, and downstream impact. Experiments on controlled manipulation videos and public egocentric or bimanual datasets, including VISOR/EPIC-KITCHENS, H2O, and ARCTIC, show that predicate failures are structured rather than uniform. Static spatial predicates remain comparatively robust, whereas contact-sensitive, dynamic, and derived predicates such as grasp and release are more fragile. Under severe degradation, detection noise, occlusion, and frame dropping cause the strongest reliability losses. Downstream analysis shows that degraded predicates reduce manipulation-understanding accuracy from 0.89 to 0.58, while removing confidence weighting under moderate degradation reduces accuracy from 0.74 to 0.64. These results show that predicate reliability provides a diagnostic layer between visual perception and structured manipulation reasoning.

A small Wasserstein distance does not certify that a transformation is admissible. In evidence-constrained, semantic, causal, physical, monotone, or risk-sensitive learning, one must ask not only how far two probability laws are, but whether mass has moved in a direction allowed by available information. We introduce conditional random ordered transport spaces (CROTS), a class of \(L^0\)-valued spaces of random probability measures equipped with a Wasserstein ambient metric, a closed stochastic order, hard and soft ordered transport discrepancies, and a conditional risk functional for evaluating order violation under an evidence sigma-field. The central object is an order-admissible transport geometry for random measure-valued dynamics, distinct from cone-valued metrics, ordered Kantorovich constructions, random Wasserstein spaces alone, and model-specific residuals for generative paths. We develop the foundations of CROTS as a space theory for reliable distributional learning. The results include well-posedness and duality for hard and soft ordered transport, soft-to-hard variational convergence, measurability and completeness of the random lifted space, reductions to classical Wasserstein and ordered geometries, ordered geodesics, constrained barycenters and projections, conditional risk-transport duality, and separation of order-violating distributions. The main stability theorem shows that random learning dynamics may converge in the ambient Wasserstein metric while its local admissibility leakage follows a separate conditional order-risk recursion. The resulting asymptotic order-risk floor provides a mathematical language for evidence overreach, ordered distribution shift, robustness failure, and admissible distributional dynamics.

Robotics faces a fundamental challenge of data scarcity. Unlike language or vision research, there is no internet-scale dataset for robotic manipulation. A promising path forward is to leverage egocentric human data, which can be collected more easily, with greater breadth, and at a larger scale. Towards this end, we investigate key design choices for learning across human and humanoid embodiments equipped with dexterous five-finger hands, using the $π_{0.5}$ model as a foundation. Our results show that human data enables robots to learn new task semantics and compose existing skills into novel behaviors without corresponding robot data. The paper website is here: https://egopipaper.github.io/

Self-evolving agents improve by repeatedly proposing changes to their own prompts, skills, or workflows and keeping those that score higher on a small held-out set. Almost all effort has gone into the proposer that generates candidates; we argue the weak point is the acceptor, the rule that decides whether to commit a change. Applied hundreds of times against the same noisy dev estimate, the ubiquitous "keep it if the score went up" rule is uncontrolled adaptive multiple testing: the agent effectively p-hacks itself, accumulating false commits that make it churn and drift rather than improve. We recast committing as a sequential hypothesis test and propose PACE (Paired Anytime-valid Commit Evaluation), a training-free, anytime-valid commit gate. Each candidate is compared to the incumbent on identical instances and committed only when a testing-by-betting e-process accumulates decisive evidence, stopping early to save evaluations and controlling each candidate's false-commit probability at a user-set level even under optional stopping (a per-decision guarantee). On Qwen2.5 agents (0.5B-3B) self-evolving at the prompt level on GSM8K, SVAMP, and ARC-Challenge, greedy acceptance commits 30-42% false and 10-33% harmful edits when a genuine improvement is hidden among noisy proposals, while PACE commits the real one and essentially nothing else, matching greedy's held-out accuracy at sharply lower variance and about 18% lower evaluation cost. With no real gain available, greedy commits 13-21 spurious self-modifications per run (72-100% false) and degrades the most fragile agent by 4.9 points, while PACE holds at baseline. Reliability of self-evolution depends on the acceptor, not only on the proposer.

When attention concentrates on a single token, a sink, what is the model actually computing? Attention sinks are ubiquitous in softmax transformers, yet this shared visual signature can hide fundamentally different algorithms. We show that visually similar sink patterns can reflect two distinct mechanisms: {i} adaptive nop, where a head suppresses its update by routing to a null token, and {ii} broadcast, where a sink aggregates and redistributes global information. In that case, sinks serve an analogous role: a safe destination when there is nothing useful to compute. Proposed interventions like gating or registers work because they implicitly target one or the other, revealing a duality between method and assumed mechanism: gating implicitly assumes nop; registers implicitly assume broadcast. Each mechanism leaves distinct traces (nop sinks exhibit negligible value norms; broadcast sinks induce low-rank outputs) which we formalize on synthetic tasks and use to derive practical diagnostics. Applied to pretrained vision transformers, these diagnostics reveal that both mechanisms exist at scale: sinks transition from CLS in early layers to patches in deeper layers, and concentrate in specialized heads. Strikingly, register tokens, designed for broadcast, are repurposed to also serve nop, confirming that neither intervention alone suffices. Combining gating with registers yields complementary gains in stability and performance. Overall, we find that the same attention pattern can reflect two very different computations and effective intervention requires first asking what the model is actually computing.

Soft-bodied organisms such as octopuses and elephant trunks exhibit remarkable morphological adaptability, dynamically reconfiguring body shape and stiffness, and flexibly adjusting their control strategies to enable versatile behaviors. Inspired by these biological systems, various soft robots have emerged in recent decades, featuring diverse materials, stiffnesses, and morphologies tailored to specific tasks. Despite substantial advances in the materials and structural designs of soft robots, developing a generalizable control framework capable of rapid adaptation across diverse configurations remains a long-standing challenge. Existing controllers are limited to fixed configurations, demanding laborious configuration-specific remodelling and policy redesign for new configurations. Here, we introduce a generalizable control system that enables rapid adaptation across diverse soft robot configurations via reinforcement learning in a shared linear Koopman embedding space. By encoding robot dynamics into this embedding space, our method decouples control policies from specific morphologies, allowing real-time, model-free policy adaptation across diverse configurations without retraining from scratch. We validate our system across 33 distinct robot configurations. Our system achieves a 75 times reduction in transfer samples across configurations, while sustaining robust performance under high-speed motion, heavy payloads, and multiactuator faults, and achieving real-world skills previously unattainable in soft robotics. This work establishes a unified and adaptable control paradigm for diverse soft robot configurations, bridging mechanical reconfigurability with control flexibility, and may offer broader insights for generalizable control in complex physical systems.

We are surrounded by various objects with movable, articulated parts, e.g., box, handle, door. An accurate and generalizable perception of articulated parts is essential to enhance robotic manipulation capabilities. Building on this need, recent efforts in articulated parts perception have followed two main directions: One line of work uses pose-based representation, which requires high manual cost; in parallel, affordance-based methods extract future object motion from point tracking without additional manual efforts, but suffer from low-quality data. In this paper, we propose a new representation of articulated parts, Geometric Primary Structure (GPS), an abstraction of the part geometry structure to balance scalability and quality. For efficient and scalable data collection, GPS is integrated with a portable Virtual Reality (VR) device and requires only one minute to annotate one object sequence. This direct human annotation provides higher quality than the estimated affordance. With this efficient VR-GPS system, we collect 41K frames for 234 objects across six part classes, and train a generalizable GPS model with a single RGB-D object image as input. For object manipulation, we deploy a heuristic policy based on GPS prediction. Without any in-domain fine-tuning, our method achieves an 73% success rate, covering 270 initial states for 9 objects. Our code, data and reusable tool are available at https://enlighten0707.github.io/gps.

Multimodal $ΔΔG$ predictors integrating protein language models with inverse-folding representations achieve strong in-distribution accuracy on the Megascale dataset but exhibit limited robustness on out-of-distribution (OOD) proteins, persistent forward-reverse bias on paired-mutation benchmarks, and under-representation of rare stabilizing mutations. Existing approaches address these limitations primarily through additional architectural components, leaving optimization-level intervention comparatively underexplored. We introduce a constraint-aware optimization framework combining Balanced Mean Squared Error, a Siamese anti-symmetric regularizer, and a novel OOD-margin consistency loss on the per-position feature representation, requiring no architectural changes to the SPURS backbone. Across eleven benchmarks and three random seeds, the framework improves Spearman correlation on S669 from 0.486 to 0.540 ($σ=0.002$ across seeds), matching the published SPURS baseline (0.50) without architectural modification, and on S461 from 0.653 to 0.711, with consistent smaller gains on five additional OOD datasets. A controlled diagnostic on Ssym reveals that anti-symmetric training does not eliminate systematic forward-reverse bias, indicating that gains arise through implicit regularization rather than exact thermodynamic constraint enforcement.

Remote communication technologies have become widely used; however, supporting a sense of shared physical space and conveying rich non-verbal cues remain challenging in many social interaction scenarios. This study presents "Yui," a full-body cybernetic android avatar designed to integrate operator-side immersive teleoperation with interlocutor-side human-like social signaling. Yui combines a 55-degrees of freedom full-body mechanism with a previously developed android head, facial expression and gaze control, upper-body and arm motion, hand actuation, and a mobile platform. It can be operated through either the immersive mode using a head mounted display-based interface or desktop mode using a webcam-based interface. We evaluated the system through three real-world deployments: a long-term public exhibition at Expo 2025 in Osaka, Kansai, Japan; a remote educational exchange between elementary school students; and a public interaction study with general participants. During the Expo deployment, two units accumulated approximately 1131 h of operation, demonstrating both operational feasibility and maintenance challenges. In the public study, both operators and interlocutors reported positive impressions of co-presence and willingness to use the system. Interlocutors also rated the avatar positively in terms of human likeness and the transmission of emotions and intentions. The results indicate usability for general operators while suggesting room for improvement in precise controllability. These findings provide field-derived evidence and design implications for socially deployable full-body android avatars.

Vision-Language-Action (VLA) policies are typically shipped as Python/PyTorch stacks that assume a workstation-class GPU, a mismatch for the hardware on which robots actually run. We present vla.cpp, a portable C++ inference runtime built on llama.cpp. To our knowledge, it is the first ggml-class engine to natively serve the flow-matching and diffusion VLA inference pattern, in which a cached vision-language prefix is consumed by a cross-attending action expert integrated over several solver steps. A single runtime serves seven architectures spanning five backbone and four action-head families behind one request/response protocol, with each model packaged as a self-contained bundle. On LIBERO-Object, the engine matches a state-of-the-art checkpoint to within one episode out of 200, and runs BitVLA at 100% success in 1.3 GiB of memory. The same bundle runs unchanged across three hardware tiers, from a consumer GPU down to an 8 GB embedded module. A cross-hardware roofline analysis shows that batch-1 VLA inference is compute-bound, so utilization rather than bandwidth is the deployment lever; an IMMA ladder GEMM derived from this analysis cuts BitVLA per-step latency by 4.5x. We then frame an on-robot stress test on an ALOHA arm that isolates the latency constraint under which a learned VLA must replan against a moving target on the hardware it was trained for. Code, demo videos, and the reproducible benchmark scaffold are available at https://fai-modelopt-tech.github.io/vla-cpp.github.io/.

Recent agent frameworks such as Claude Code, Codex, and OpenClaw are strong at tool use and orchestration, but whether they can handle long video generation, a long-horizon multimodal task, remains underexplored. Unlike earlier video agents whose pipeline is handcrafted, these frameworks can build and refine their own workflows. We introduce VideoWeaver, an agent harness and benchmark that evaluates and evolves skills for long video generation, where an agent turns a single instruction into a long video by composing foundation skills into its own workflow rather than following a predefined pipeline. The benchmark has 16 task categories and 285 cases, with references spanning text, image, audio, video, and their combinations. Because errors can arise at any stage and not just in the final video, we propose an agent-as-judge that inspects both the execution trace and the final video, grounding its scores in evidence such as metadata and intermediate files. Using this feedback, we further design a skill evolution algorithm that refines and merges the agent's skills. Across multiple frameworks and models, we find that an explicit composition skill improves the generation process over using foundation skills alone, that skill evolution further improves output quality, and that performance varies notably across harness and model choices. The proposed agent-as-judge also aligns well with human judgments, especially on process metrics. Code and dataset is available at https://github.com/JianhuiWei7/VideoWeaver

Reinforcement Learning from Verifiable Rewards (RLVR) has recently become a key paradigm for improving the reasoning abilities of Large Language Models (LLMs), yet it remains limited by sparse binary rewards and its ignorance of model-internal uncertainty. In this paper, we propose ConSteer-RL, a simple yet effective framework that integrates token-level confidence signals derived from model log-probabilities into RLVR training. Specifically, building upon the Group Relative Policy Optimization (GRPO) framework, we construct a confidence-aware reward by aggregating per-token probabilities into a scalar confidence score and incorporating it into an awareness-based reward shaping mechanism that penalizes overconfident errors while reinforcing correct and confident reasoning. Experimental results demonstrate that ConSteer-RL consistently outperforms strong GRPO baselines, achieving average improvements of 2.3%-4.0% across different model scales.

Deep-learning speaker verification (SV) increasingly relies on deep neural network backbones, whose environmental impact remains largely undocumented. In this paper, we conduct an evaluation of ResNet architectures trained on VoxCeleb2, varying depth, channel width, and stage distribution, and measure energy consumption and carbon footprint using node-level sensors. Results show a clear point of diminishing returns: deeper or wider models bring only marginal accuracy gains while energy consumption grows steeply. In contrast, mid-sized networks such as ResNet-50 and stage-concentrated variants achieve favorable trade-offs between performance and environmental impact. These findings provide actionable guidelines for designing energy-efficient SV systems.

Repeated reference games test whether interlocutors replace their initially long descriptions with shorter, partner-specific conventions grounded in shared interaction history. Prior work shows that multimodal LLMs fail to become more efficient across rounds, although they align on the labels they use. How can we determine whether this alignment reflects partner-specific grounding rather than a shared task vocabulary? We address this question by comparing capable multimodal agent dyads with human dyads from the KTH Tangrams corpus. Our novel methodological contribution is a constrained pseudo-dyad baseline that matches the original referential task structure, but breaks partner history. This baseline enables us to test whether the observed label alignment depends on interaction with a specific partner. Across three analytic layers (task competence, description strategy, alignment dynamics), we find clear differences. Humans reduce effort through entrainment, compressing descriptions and increasing label alignment with partners. Agents instead maintain fixed effort levels, producing verbose descriptions from round one, with near-ceiling label overlap that is statistically indistinguishable between real and pseudo dyads. MLLMs thus achieve coordination without convention, succeeding by verbose description rather than by forming the compact, history-dependent referring expressions characteristic of human dialogue.

Although low-bit quantization provides practical means to deploy speaker verification on resource-constrained devices, its effects on speaker verification performance remain poorly understood. In this paper, we study uniform K-means quantization-aware training of ResNet-36 and ResNet-200 through joint layer-wise and score-level analyses. Our layer-wise analysis highlights fragile components and shows that score degradation is not fully explained by weight distortion alone. We identify a clear knee point at 2 bits, with larger score drift and harmful decision flips concentrated near the FP32 threshold. Our score-level analysis reveals where and how score errors emerge under extreme quantization. Building on these findings, we propose a calibrated multi-precision cascade that resolves most trials at 2 bits and escalates only ambiguous cases, achieving performance close to FP32 while preserving the efficiency benefits of low-bit inference with substantially lower compute and memory costs.

Rubric-based evaluation is a promising paradigm for judging large language model (LLM) outputs, yet self-generated rubrics lag human-annotated criteria on hard instances. We argue this discriminative gap reflects an objective mismatch: self-generated rubrics describe good responses, whereas effective criteria must discriminate between close candidates. To close this gap, we introduce SVR (Support Vector Rubrics), a framework that recasts rubric construction as max-margin boundary learning over preference data. SVR mines contrastive features from preference pairs into a rubric bank, learns a prompt-conditioned selector together with global rubric weights, and iteratively refines the bank through support-pair selection and adversarial probing of hard negatives. At inference, given only the prompt, SVR retrieves the top-rubrics from the bank and scores responses. On RubricBench, SVR narrows the gap to human reference rubrics from 24.1 to 0.3 points and outperforms strong self-rubric and judge baselines, and the learned bank transfers across judges without retraining. On RewardBench 1&2, and RM-Bench, it remains competitive with dedicated reward models, demonstrating broader reward modeling capability. Overall, boundary-defining rubrics offer a principled route to closing the discriminative gap in LLM evaluation.

Reliable evaluation of large language models in surgery remains underdeveloped. Broad medical benchmarks test clinical knowledge, while surgery requires procedural reasoning, management trade-offs, negation handling, and selection among plausible operative decisions. We present SurgiQ, a text-only, source-grounded benchmark of 13,055 four-option multiple-choice questions spanning six surgical domains and four question formats: case-based, reasoning, best-option, and negative. SurgiQ is constructed from surgical textbooks, open-access papers, and examination material using a multi-stage generation, verification, and expert-audit pipeline. We evaluate 35 open-weight LLMs under a unified log-likelihood protocol. Our results show substantial remaining headroom: smaller models often remain near the 25\% random baseline, while the best model reaches 68.1\% accuracy. General-purpose models, especially Qwen2.5, outperform most biomedical models, suggesting that current medical specialization does not yet provide sufficiently broad surgical coverage. Calibration and error analysis further show that even strong models make confident mistakes on clinically plausible distractors, motivating more reliable and broader surgical LLM evaluation.