Search research

Humanoid robots require whole-body controllers that are both robust and precise in contact-rich environments. While deep reinforcement learning (RL) achieves robust stability, its behavior is tightly coupled to the training objective and command interface, making it difficult to add new feedback objectives without retraining. In this study, we propose an RL guided whole-body model predictive path integral (MPPI) framework that acts as an add-on feedback controller on top of a pretrained RL policy. Instead of using RL policy as the final controller, we use it as a sampling prior that biases MPPI rollouts toward dynamically feasible behaviors. Task objectives are specified through modular MPPI cost terms, and MPPI closes the loop by continuously correcting the RL prior online to satisfy these objectives without retraining the policy. Simulations on a 29-DoF Unitree G1 humanoid in MuJoCo demonstrate stable high-rate control (average 280~Hz). The proposed method improves task-level precision over a pure RL baseline under the same command interface. This is achieved by correcting systematic drift during straight walking and tracking additional whole-body reference signals imposed through the cost.

Deploying robots in unstructured real-world environments needs accurate, interactive models of the objects. Constructing these models at scale remains a critical bottleneck for robotic system integration. We present ArtiTwinSplat, a framework that automatically constructs articulated, photo-realistic digital twins of objects directly from RGB-D videos, requiring no CAD models, simulation assets, or manual annotations. Our method is built on 3D Gaussian Splatting that preserve geometric fidelity and photometric realism, coupled with an unsupervised articulation discovery pipeline that recovers part structure and joint kinematics from observed motion alone. With tracking and optimization stages our method provides stable, queryable digital twins that support real-time rendering, viewpoint control, and interactive manipulation. Unlike prior methods confined to simulation, ArtiTwinSplat operates directly on real-world observations and produces twins that are immediately usable by downstream robot planning and learning systems. This method offers a practical, scalable pathway toward digital twin construction, lowering the integration barrier for articulated object manipulation in embodied AI and human-robot collaboration contexts.

Multi-agent LLM environments require robust mechanisms for shared knowledge management. This paper formalizes the fleet-memory problem and identifies four foundational failure modes: unauthorized leakage, stale propagation, contradiction persistence, and provenance collapse. To address these, we define explicit systems-level primitives: scoped retrieval, temporal supersession, provenance tracking, and policy-governed memory propagation. These primitives are implemented in MemClaw, a production multi-tenant memory service, and evaluated via ArgusFleet, a reproducible harness testing four governance dimensions. Rather than a baseline comparison, this study measures a live production service, emphasizing real-world architectural insights and negative results. Key Evaluation Results Provenance: Successfully reconstructed 100% of depth-four derivation chains with correct writer identity at sub-second per-hop latency. Propagation: Demonstrated high intra-fleet visibility with zero cross-fleet leakage. Under strong write mode, write-to-visible latency was optimized to a single search round-trip. Production Architectural Issues Discovered Asymmetric Scope Enforcement: Tenant isolation held, but sub-tenant scope was initially bypassed on direct GET-by-id requests for agent-scoped credentials (disclosed and remediated during the study). Pipeline Ordering Conflict: While contradiction supersession works for admitted writes, a synchronous near-duplicate gate can prematurely reject contradictory writes before the asynchronous contradiction detector can evaluate them. Conclusion: Long-context retrieval alone is insufficient for production multi-agent memory. Governed shared memory demands explicit systems-level abstractions, and live evaluation is vital to expose enforcement and pipeline-ordering failures missed by design-only treatments.

We introduce EgoSAT, the first comprehensive benchmark for egocentric video reasoning in streaming settings, designed to evaluate the capabilities of modern vision-language models (VLMs). The benchmark targets streaming interaction understanding, where video frames arrive sequentially and models must continuously interpret evolving visual context. EgoSAT unifies several previously distinct tasks within a single streaming framework. In this formulation, queries about completed events correspond to retrospective reasoning, queries about ongoing activities require online understanding, and queries about future actions involve prospective anticipation. This unified setting requires models to reason about the past, present, and future while operating under the constraint that only previously observed frames are available. EgoSAT contains 1,997 unique videos spanning 165 hours of egocentric footage and around 4,800 high-quality question-answer pairs, carefully designed to probe reasoning across varying temporal contexts. Using this benchmark, we evaluate a diverse set of both open-weight and closed-weight VLMs, providing a systematic assessment of their ability for streaming interaction understanding. By distinguishing answerability and conducting diagnostics on confidence of models, we find existing models not only struggle with prospective and retrospective modeling, but also exhibit severe mis-calibration: confidence often fails to track inherent answerability, leading to dangerous "confidently wrong" behaviors. Project page: https://leiyj23.github.io/EgoSAT/

Compliant joints can embed safety and adaptability into dexterous hands, but achieving large-stroke anthropomorphic motion while maintaining joint-specific, directiondependent stiffness and reliable proprioception remains challenging. This paper presents the PDS joint, a parametric doublespiral (PDS) compliant joint that enables systematic shaping of directional stiffness across multiple deformation modes, including flexion/extension, abduction/adduction, and pronation/supination. We instantiate the joint using Archimedean and logarithmic spiral templates for different hand joints and introduce an asymmetry ratio to tailor stiffness distributions for both grasp stability and hyperextension resistance. To make the joint practically usable under large deformation, we co-design embedded inductive proprioception and propose a learningbased calibration pipeline that maps raw inductive signals to joint states using ArUco-marker tracking. Experiments characterize the stiffness landscapes across geometric parameters and demonstrate a non-monotonic dependence of lateral support on asymmetry, indicating the importance of principled parameter tuning. For joint-state estimation in the most challenging abduction/adduction motion, a learned multilayer-perceptron (MLP) mapping reduces the error compared with conventional curve fitting by 41.6%. Finally, we integrate the proposed joints into an open-source dexterous hand as a demonstration platform, on which the hand grasps a set of nine everyday objects and performs safe, contact-rich human-involved interactions.

We present our submission to the SoccerNet 2026 Player-Centric Ball Action Spotting challenge, which uses a two-stage pipeline: a Track-Aware Action Detector (TAAD) produces per-player action logits from broadcast video, and a Denoising Sequence Transduction (DST) transformer converts game-state features and TAAD logits into structured event sequences. We improve the TAAD with a temporal transformer that adds cross-frame context, alongside several training fixes. For the DST stage, we introduce a two-stage per-player attention mechanism operating on game-state features, and show that a spatial-first attention ordering (cross-player attention before temporal attention) improves validation Macro-F1 by 1.87%. To exploit architectural diversity, we train four model variants and combine them with a Weighted Event Fusion ensemble that applies agreement filtering to suppress single-model false positives while preserving recall, plus a dedicated exception for the rare tackle class. Our final system improves the challenge Macro-F1 from a baseline of 48.6 to 58.94.

As large language model (LLM) agents are applied to longer tasks, they increasingly modify workspace state across multiple rounds of iteration. However, agents typically observe only tool outputs and log fragments, while the actual state changes occur in the file system. Without explicit workspace boundaries, state-changing operations such as file writes and temporary artifact generation may scatter changes across paths. Over time, these weakly constrained changes accumulate, making states such as modified files difficult to track. This paper presents LemonHarness, an integrated execution framework for long-horizon agents. LemonHarness establishes an explicit execution boundary by constraining state-changing operations within a clearly defined workspace and bringing model invocation, tool execution, and rule knowledge within a single controlled boundary. State-changing operations, including file writes, dependency installation, and temporary artifact creation, are executed through structured tool interfaces, with execution feedback recorded as observations available to subsequent model decisions. The system also introduces a reusable rule knowledge base, which turns recurring execution rules and acceptance criteria into runtime knowledge. LemonHarness further adds a time-aware execution mechanism that exposes elapsed and remaining budget to the model, so it can rebalance exploration, implementation, and validation effort as time pressure shifts and avoid timeouts from long waits or excessive verification. On Terminal-Bench 2.0, LemonHarness_GPT-5.3-CodeX reached 84.49% accuracy over 445 trials; pairing the same framework with the stronger GPT-5.5 backbone raised the average accuracy to 86.52% across five jobs. The results suggest that a unified runtime boundary, callable rule knowledge, and time-aware execution can improve the stability of long-horizon agent execution.

Most electronic health record (EHR) foundation models encode clinical events as discrete event tokens from a fixed vocabulary and therefore cannot directly represent events containing unseen concepts or new combinations of concepts and attributes such as numeric values. This limits transfer across institutions and even across deployment pipelines within the same institution. We introduce PORTER, a language-grounded structured EHR foundation model that decouples event representation from this fixed vocabulary. PORTER represents events through their descriptions using a frozen text encoder, integrates numeric values through a dedicated pathway, and learns clinical dynamics over patient timelines with an autoregressively pretrained temporal backbone. Across 74 clinical prediction tasks at a pediatric hospital, PORTER matched the mean AUROC of a fixed-vocabulary model with the same temporal backbone and pretraining objective. When the same patient timelines were rendered using event descriptions not seen during pretraining, PORTER transferred without retraining or vocabulary mapping, recovering 97.1% of the mean AUROC of a model trained directly on the target vocabulary. When transferred to MIMIC, PORTER outperformed the fixed-vocabulary model, which dropped 69% of events because their tokens were unseen. Mechanistic analyses showed cross-vocabulary transfer tracked preservation of patient-level representation geometry rather than the scale of the text encoder, and the numeric pathway improved sensitivity to magnitude without disrupting clinical concept identity. PORTER also achieved higher AUROC than a task-specific text serialization comparator, at 329-fold lower amortized compute. PORTER is a step toward vocabulary-independent EHR foundation models that reduce the need for vocabulary harmonization while preserving in-domain performance and enabling efficient cross-task reuse.

Food images often contain several visible ingredients, so assigning one dish label to an entire image hides important visual structure. This work studies ingredient-level semantic segmentation on FoodSeg103, where the model predicts an ingredient class for each pixel. Two SegFormer variants were fine-tuned and evaluated under a controlled setup: SegFormer-B0 as the smaller baseline model and SegFormer-B1 as the larger final model. Both models use ImageNet-pretrained MiT backbones with newly initialized 104-class output layers. On the held-out FoodSeg103 test split of 2,135 images, B0 achieved 0.7709 pixel accuracy and 0.2521 mean IoU, while B1 achieved 0.7929 pixel accuracy and 0.3204 mean IoU. B1 improved every saved test metric, including a +0.0683 absolute gain in mean IoU. The system also converts predicted masks into visible ingredient-area percentages, giving a simple visual composition summary of the predicted meal. This summary can serve as a first-pass nutrition-awareness cue by providing a visual alternative to detailed food tracking similar to plate-based meal guidance, but it is not a direct estimate of calories, macronutrients, food mass, volume, density, or true portion size.

A modern model release reports scores on 40+ benchmarks and the same evaluations were run many more times before it: to track training progress, compare design choices, and select the checkpoint for the release. But do we need to run every eval? We compile a public score matrix of 84 frontier models on 133 benchmarks (2,604 cells, 23.3% filled) and find it is approximately rank-2: a model's scores across all 133 benchmarks are largely determined by just two numbers. We confirm this in two ways: scores hidden from the matrix are best recovered using two factors, and two factors already explain over 90% of the variation among models on the benchmarks they share. Building on this, we design BenchPress: a logit-space rank-2 matrix completion method that recovers held-out scores to within 4.6 points, and a confidence layer that says when each prediction can be trusted. Using BenchPress, we find a subset of five benchmarks {GPQA-D, HLE, Codeforces, MMLU-Pro, ARC-AGI-1} that can recover the rest of a model's public scorecard to within 3.93 points. For a tighter inference budget, a cheaper set {GPQA-D, MMLU-Pro, Aider Polyglot, MATH-500, AIME 2026} can predict a model's evals to within 4.55. We release the score matrix, the BenchPress code, and an interactive tool that predicts any model's score on any benchmark.

High-throughput scientific facilities such as the Large Hadron Collider depend on real-time event filtering (\textit{triggering}) under tight constraints on bandwidth, latency, and storage. In practice, trigger menus are largely static and hand-tuned and can become suboptimal as detector conditions, pileup, and background composition drift over time. We cast online threshold tuning as a sequential decision-making problem: a reinforcement learning agent ingests streaming summaries of recent rates and signal-sensitive features and updates trigger thresholds to maximize signal efficiency while tracking a target background rate within a tolerance band. We adapt Group-Filtered Policy Optimization (GFPO) to streaming control and introduce two variants (GFPO-F, GFPO-FR) that enforce background rate feasibility during training. On a benchmark that emulates realistic collider operation, we study two representative triggers: a total transverse energy ($H_{T}$) trigger sensitive to pileup variation, and an anomaly-detection (AD) trigger based on reconstruction loss for rare or non-standard signatures. On Monte Carlo streams, our agent increases the fraction of in-tolerance time intervals by 48% ($H_T$) and 28% (AD), with a cumulative gain of up to 2% in signal efficiency on those in-tolerance intervals. Transferring from simulation to \emph{real} collision data (CMS Run 283408), the same agent, without fine-tuning, achieves a 56% ($H_T$) and 28% (AD) in-tolerance improvement over baselines, with further signal-efficiency gain on both triggers. To our knowledge, this is the \emph{first} demonstration of RL-based trigger control on real Large Hadron Collider collision data. Code is available at https://github.com/Zixind/GFPO_LHC (see repo for details).

Distilling conditional diffusion models aims to transfer the behavior of a large teacher to a smaller student while preserving alignment across conditioning inputs. Unlike recognition tasks, knowledge distillation in conditional diffusion often struggles to transfer knowledge beyond the training distribution, since the predicted noise strongly depends on the conditioning signal. As a result, effective distillation requires exploring a large conditioning space. In practical settings, this creates a major bottleneck. Paired image-condition data may be limited, and generating synthetic images for every available condition is often computationally infeasible, while the pool of conditions, such as text prompts, can be extremely large. Recent work addresses this issue by switching conditions during training, exposing the student to a broader conditioning space without changing the distillation objective. Yet this raises a complementary question: once a large conditioning corpus is available, how should the training effort be allocated? In this work, we introduce ARIA, a framework that adaptively allocates training effort across coarse regions of the conditioning space. By maintaining online estimates of teacher-student discrepancy at the region level, ARIA focuses updates where misalignment persists while preserving the original distillation objective. Empirically, ARIA improves over RC across most architectures and settings, with the clearest gains observed in unseen and underrepresented regimes. We also provide a theoretical analysis showing that the proposed tracking mechanism follows the evolving discrepancy during training under bounded variance and drift assumptions.

Humanoid loco-manipulation is often simplified into a stop-and-go process: walking to an object, stopping to manipulate it, and then resuming locomotion. It also commonly relies on low degree-of-freedom (DoF) end effectors that behave like an open-close grasp primitive. We introduce CoorDex, a learning pipeline that converts high-dimensional body and dexterous hand control into coordinated latent residual control, enabling high-DoF dexterous loco-manipulation on the move. Starting from simulated whole-body and hand demonstrations, CoorDex trains privileged motion tracking teachers for the humanoid body and dexterous hand, distills them into proprioception-conditioned latent priors, and uses the frozen priors as the action space for downstream residual reinforcement learning. A coordinated latent residual policy composes these priors through shared task context and separate body-hand residual heads, preserving natural whole-body motion while improving finger-level contact reliability. CoorDex enables a Unitree G1 humanoid with a 20-DoF WUJI hand to execute dexterous manipulation while in motion, including non-stop bottle grasping and carrying, fridge door opening on the move, and cube pick-and-turn. Ablations on the walk-grasp-carry task show that joint-space PPO, joint-space hand control, and monolithic latent prediction all fail under the same reward budget, while the latent-prior interface and coordinated residual structure make high-dimensional contact-rich loco-manipulation trainable. Project Page: https://skevinci.github.io/coordex/

Remote patient monitoring depends on patient-reported data to capture the subjective dimension of recovery that devices cannot measure. The Quality of Recovery (QoR-15) survey is the gold-standard instrument for this purpose. It was designed and validated for occasional in-hospital assessment, yet remote monitoring now administers it to patients daily. In our own post-surgical deployment, only 55% of patients submitted the survey more than 14 days of 30 monitoring days. We developed QoR-compact, a five-item daily input for the RPM prediction pathway. Setting a deployment-driven target of one-third of the daily items, we exhaustively evaluated all 3,003 five-question subsets of the QoR-15 and tested whether the best of them matches the full instrument in predicting near-term postoperative recovery severity. QoR-compact achieves a mean AUC-ROC of 0.968 (95% CI 0.915-0.988), statistically comparable to the 0.964 baseline obtained with one-third of the items. Patient-level backtesting indicates that it tracks readmission events as faithfully as the full form. Its five items span the physical and psychological axes of recovery: Q3 (feeling rested), Q9 (feeling comfortable and in control), Q10 (general well-being), Q12 (severe pain), and Q14 (feeling worried or anxious). The QoR-15 remains the gold-standard measure of recovery; QoR-compact complements it as a shorter daily input designed for prediction. This parity provides the basis for a prospective study of whether a lighter daily input is, in turn, completed more consistently. External validation on larger cohorts is required before clinical use.

We propose Hedgementation: a new benchmark to evaluate machine learning models for hedgerow mapping from remote sensing data at country scale and 10m$^2$ spatial resolution. We combine and harmonize multiple remote sensing data products and ground truth labels sourced from a hedgerow inventory in France. We measure the ability of three baseline models to generalize across spatial distance, and across climatic zones, a more explicitly challenging task. Our benchmark tests both supervised and self-supervised learning approaches for remote sensing, applied to tracking fine-scale features of high agricultural importance. The code to reproduce the benchmark and baselines results is available at https://github.com/hedgementation/hedgementation.

While Large Language Models (LLMs) are increasingly deployed in long interactions, existing evaluations focus predominantly on retrospective memory (RM) via explicit queries. Prospective memory (PM), the critical ability to spontaneously recall and act on latent constraints without direct prompts, remains largely unevaluated. We introduce TriggerBench, a comprehensive PM benchmark spanning five dimensions across both daily assistants and professional workflows. TriggerBench pairs scenarios with matched RM controls, contrastive positive/negative variants, and overloaded triggers, enabling fine-grained measurement of proactive recall, false-alarm rate, and attentional robustness under a single protocol. Our evaluation yields three key findings. (i) PM shows a precision-recall trade-off and attentional fragility. Though enhanced reasoning significantly improves proactive recall, models may overfit to an "always-remind" heuristic. Furthermore, PM accuracy degrades substantially under implicit constraints or triggers overloaded by concurrent user requests, indicating that robust PM remains an open challenge. (ii) PM is notably harder than RM: on identical contexts, RM near-saturates up to 100K tokens, while PM decays sharply as context length scales. (iii) PM may serve as a behavioral probe of spare reasoning capacity. Pairing PM scenarios with AIME-2025 math problems reveals that successful trajectories yield higher PM accuracy than failed ones at the same context length, showing PM tracks spare reasoning budget that token count obscures. Project page: https://github.com/KristenZHANG/TriggerBench-Official.

Chemical language models (cLMs) are widely assumed to learn surface-level syntactic patterns rather than learning meaningful molecular semantics. Here, we apply sparse autoencoders (SAEs) to MolFormer, an encoder-only cLM, to mechanistically examine how molecular representations are built across layers. We discover that early layers rely on position-tracking latents to parse molecular grammar, while later layers encode atom-in-substructure and pharmacologically relevant features. Additionally, we show that non-canonical SMILES produce more disruptive representation shifts than invalid SMILES, driven by position-latent disruption propagating across layers. To support further exploration, we develop InterMol, an interactive visualizer for SAE activations on molecular strings and structures.

Fine-grained, bimanual dexterous manipulation remains a foundational challenge in robotics. Traditional teleoperation systems often fail in contact-rich tasks because embodiment gaps hinder accurate kinematic mapping, while tactile and force feedback remain absent. Consequently, data collection efficiency for high-precision tasks remains prohibitively low. To address these limitations, we propose a tactile-driven adaptation strategy designed to enable fine-grained manipulation on top of teleoperation pipelines. Instantiated within our bimanual dexterous framework, DexTeleop-0, this strategy introduces a real-time optimization loop that bridges the embodiment gap by translating coarse human tracking intents into precise, force-compliant robotic commands with tactile sensing. By estimating accurate contact points and leveraging a tactile-enabled fingertip force-sensing profile, the system dynamically computes localized corrections using the operational space Jacobian with respect to joint angle updates. We rigorously evaluate this tactile-driven adaptation strategy across both simulated environments and real-world hardware. Compared with representative baselines, the proposed method consistently achieves higher task success rates and improved execution efficiency in robust grasping, disturbance-resilient manipulation, and complex dexterous tasks.

Transformer language models have become established tools for modeling human sentence processing, with measures such as surprisal and attention entropy serving as effective predictors of reading difficulty that together capture complementary aspects of processing load. Here, we explore a related class of transformer models: energy-based transformers, which provide a principled formal link to associative memory models, bringing processing research into direct contact with the broader literature on Hopfield networks and dense associative memory. To our knowledge, this is the first exploration of an energy-based transformer measure in computational psycholinguistics. Across reading-time corpora (Natural Stories, UCL eye-tracking, UCL self-paced reading), the energy measure is a robust predictor of reading times, providing significant fit beyond surprisal and entropy in all three. In a controlled experiment on relative clause processing, energy at a single layer captures the well-known object/subject asymmetry. We find evidence that it subsumes effects attributable to both attention entropy and surprisal, suggesting that energy may serve as a single unified predictor where multiple complementary measures have previously been required.

Large language models increasingly mediate interactions between sensitive data, untrusted inputs, and privileged actions in agentic systems, creating security and privacy risks. These range from prompt injections that manipulate downstream tool use to leakage of confidential information through model outputs. Recent Information Flow Control (IFC)-based defenses show promise but lack a principled semantic foundation for reasoning about information flow through the model itself. Since any input token may influence any output token in an autoregressive LLM, existing approaches suffer from severe taint explosion. We present Geometric Information Flow (GIF), a semantic framework for tracking information flow from input tokens to outputs. GIF uses the LLM Jacobian and local output geometry to upper-bound the Shannon mutual information between perturbed input spans and model outputs, yielding a scalable measure computable on large models via automatic differentiation and low-rank approximation. Unlike attention-based or correlational attribution heuristics, GIF satisfies local geometric soundness, and we provide a fully mechanized Lean 4 proof that it upper-bounds the true information flow induced by a given prompt under local regularity assumptions. We evaluate GIF on integrity and confidentiality tasks across multiple prompt-injection and privacy-leakage benchmarks. GIF achieves near-perfect recall even without a downstream declassifier, outperforming attention-based baselines. Combined with lightweight LLM-based declassifiers, it matches or exceeds the F1 of direct LLM-as-judge baselines such as GPT-5.5 xhigh reasoning while using up to 81x lower token cost. GIF flows detected with small surrogate models transfer to larger state-of-the-art models and other model families, even when the surrogate is up to 200x smaller, suggesting black-box deployment without gradient access.

Humanoid local navigation in cluttered environments must jointly resolve obstacle avoidance, sparse-goal recovery, and stable whole-body locomotion under short-range and partially observable sensing. Explicit planner-control decompositions introduce latency and can mismatch agile humanoid command-tracking limits, while purely reactive controllers may lose the goal after obstacle occlusion. We present LP-NavOA, a limited-perception navigation and obstacle-avoidance framework for humanoid robots. A raycast-conditioned perception-action proximal policy optimization (PPO) locomotion backbone is first trained with a robot-centered circular heading-speed command and a shared command-side safety filter. With this backbone frozen, A-star and waypoint teachers generate rollouts for distilling a recurrent local planner that overwrites only the heading command at deployment, leaving the whole-body policy intact. At runtime, LP-NavOA uses proprioception, short-range local range sensing, and a body-frame goal direction, requiring no global map, waypoint stream, or external planner. In MuJoCo open-wall and indoor layouts, the distilled planner produces obstacle bypassing and post-avoidance goal recovery, raising teacher-calibrated on-time arrival from 38--40% to 85--97% and reducing brush/contact-heavy progress relative to a backbone-only controller. Ablations show that dynamic route shaping, teacher-active data collection, and the circular command interface are important for navigation efficiency and for training the 3.0,m/s backbone. A Unitree G1 deployment analysis demonstrates hardware executability without continuous joystick steering.

Physical-structure priors such as conservation laws, Hamiltonian forms, and symmetries can improve scientific machine learning when correct, but can degrade predictions when misspecified. Existing methods usually enforce a chosen structure or tune a soft penalty, without a calibrated rule for deciding whether to impose a prior, how strongly to impose it, which prior to use, or which subset of candidate laws holds. We introduce SPADE, Structure-Prior Adaptive Decision Estimation, a closed-form framework that treats this problem as shrinkage of the structure-violating block of an unconstrained estimator. SPADE uses one exact specification test and one estimand: the test decides whether the prior is supported by data, Stein-unbiased James-Stein shrinkage sets the enforcement strength with an $O(σ^2/n)$ oracle guarantee, and a gate commits to the hard prior only when the test certifies it. The same test yields consistent nested structure selection and Benjamini-Hochberg control for subset discovery in non-nested constraint families. Across a linear-subspace prior, a reservoir conservation law, and a nonlinear Hamiltonian prior on Duffing dynamics, SPADE tracks the oracle, beats a neural-network baseline, reduces correct-prior regret from $10.3\%$ to $2.6\%$, matches cross-validation with $1/71$ of the solves, selects the correct structure with $100\%$ accuracy, and recovers partial laws with controlled false relaxation.

LLM agents are increasingly deployed in multi-party environments, handling sensitive personal data on behalf of individual users, for instance in group chats. When such an agent discloses private information, it reaches every group member at once. This risk is structurally harder to control than in one-to-one settings, as every piece of private information must be appropriate for every recipient in the group. Yet all existing contextual privacy benchmarks consider only single-interlocutor settings, leaving multi-party privacy risks unmeasured. We introduce MuPPET (Multi-Party Privacy Exposure Testing), a benchmark for contextual privacy in multi-party conversations. Our experiments show that models leak substantially more in multi-party settings than one-to-one evaluations suggest. Frontier models are vulnerable, and smaller open-weights models, often preferred for local deployment with sensitive data, even more so. Existing contextual privacy defences offer only partial protection, degrade utility, and do not resolve the underlying party-tracking problem.

Multi-turn tool-using agents must coordinate long-horizon tool sequences while tracking dialogue state and policy constraints. Existing approaches often separate inference-time orchestration from parameter-level learning, leaving tool selection weakly structured and preference updates vulnerable to train--deployment prompt mismatch. For within-benchmark self-improvement, ToolGraph combines schema-derived topology, transition weights estimated from successful rollouts, and history-aware controls for write prerequisites and repeated-search loops. We then construct 161 preference pairs by locating divergence points via state-based matching and prefix-based alignment, filtered through action-correctness annotations, and train DPO under the same ToolGraph context used at inference. Across 375 tau2-bench tasks, ToolGraph raises the weighted average reward from 0.304 to 0.338 (+11.2% relative), while ToolGraph+DPO reaches 0.355 (+16.8% over the baseline), with the DPO gain concentrated in airline and retail. Fine-grained diagnostics further show that roughly half of telecom trajectories exhaust the step budget before action execution and that chosen reward positivity is the most useful checkpoint signal across our 16 evaluated DPO configurations.

Outer-loop link adaptation (OLLA) is widely deployed in 5G NR to track channel variations, yet its reliance on first-order, single-bit feedback degrades performance significantly under high-mobility and fast-varying channels. This paper presents LOLLA (Learned Outer-Loop Link Adaptation), a deep reinforcement learning framework that replaces the conventional OLLA staircase with a learned, continuous SINR offset conditioned on rich PHY/MAC telemetry inaccessible to OLLA. The offset modulates the SINR-to-MCS lookup table, preserving 3GPP-compliant MCS selection and provably subsuming the conventional OLLA update rule. A Proximal Policy Optimization (PPO) policy trained under a Lagrangian block error rate (BLER) constraint automatically enforces tunable reliability targets from 1% to 15% without manual penalty calibration. The framework is realized as the first closed-loop AI-native control dApp on a GPU-accelerated 5G NR stack, achieving end-to-end control latencies under 500 microseconds. Evaluations under 3GPP TDL channel models demonstrate 15% to 92% throughput gains over OLLA across Doppler frequencies up to 400 Hz, while attaining a Pareto frontier that strictly dominates OLLA across all evaluated reliability targets. The learned policy generalizes to unseen channel models and scales to eight concurrent UEs under shared-resource scheduling. In the uplink formulation, the gNB directly observes decoding outcomes, enabling simulation-to-deployment parity.

Rubric-based rewards offer interpretable and fine-grained optimization signals for reinforcement learning in open-ended tasks where verifiable answers are unavailable. However, pre-constructed rubrics remain static throughout training, creating a fundamental mismatch with the evolving policy: fixed criteria gradually lose discriminative power as the model improves, leading to reward saturation and potential hacking. Recent dynamic rubric methods partially address this but rely on external frontier models or ground-truth answers, and update rubrics only at coarse granularity. We propose EvoRubrics, a co-evolutionary RL framework where a Policy LLM and a Rubric Generator jointly improve through adversarial interaction within each training step. As the policy improves under the rubric generator's guidance, the rubric generator adapts its criteria to remain discriminative and informative, enabling evaluation to track the policy in real time and naturally inducing an automatic curriculum. Experiments show that EvoRubrics consistently outperforms static and dynamic rubric baselines across benchmarks. The learned Rubric Generator further generalizes as a transferable reward model. Notably, even a fully self-supervised variant without any external supervision achieves meaningful gains, suggesting that co-evolution between generation and evaluation alone can provide sufficiently rich learning signals. Our code is publicly available at https://anonymous.4open.science/r/EvoRubrics-2155/.

Human motion follows a temporal hierarchical structure, transitioning from low-frequency global trajectories to high-frequency details. Inspired by the success of multi-level autoregressive models in computer vision, we propose MotionMAR, a coarse-to-fine framework for motion reconstruction from sparse observations. It first estimates the global trajectory of human motion and then gradually refines the temporal details. This architecture consists of four integrated components. The Temporal Multi-scale Tokenization (TMT) VQ-VAE encodes the data at multiple temporal resolutions, separating semantic motion from minor jitters. The Motion Autoregressive Network (MAN) operates in this latent space, predicting motion across scales. It first establishes the global structure through coarse indices and then generates finer indices to recover specific details. Meanwhile, the Scale-Aware Control (SAC) module integrates sparse tracking data to ensure the generated output aligns with actual observations. The Motion Refinement Network (MRN) subsequently smooths consecutive poses and eliminates quantization artifacts. Experiments show that MotionMAR achieves state-of-the-art accuracy on the AMASS dataset, providing a reliable and structure-aware approach for motion reconstruction. The source code is publicly available at http://www.lidarhumanmotion.net/motionmar/.

Long-horizon LLM agents can fail quietly: they settle on one reading of the evidence early, then spend the rest of the run defending it. We call this premature commitment. Final-answer scoring misses the failure mode because it sees only the answer, not whether the process has already collapsed to a stable path. We define representational commitment as cross-run hidden-state convergence at a fixed reasoning step, and use it as an early diagnostic of trajectory consistency. On Llama-3.1-70B running ReAct on HotpotQA, step-4 hidden-state similarity predicts downstream behavioral consistency (r = -0.35, partial r = -0.45), with a localized temporal and layer-wise signature. The signal replicates across Qwen-2.5-72B and Phi-3-14B, and on StrategyQA (r = -0.83). It does not track correctness: committed-wrong and committed-correct questions are not separable in activation similarity. That boundary is central to the claim. Commitment tells us whether an agent has settled, not whether it is right. A runtime monitor detects inconsistent trajectories from hidden states at AUROC up to 0.97 (0.85--0.88 under a stricter split), and a prompting intervention cuts behavioral variance by 28% against a token-matched control while leaving accuracy statistically unchanged. We also test whether the signal can route self-consistency compute; on a harder benchmark it helps only modestly and is matched by a simpler output-based baseline. The result is a diagnostic for a hidden process failure, with clear limits rather than a general accuracy lever.

LLM agents increasingly act as personal assistants that must remember a user's profile over months: who they are (attributes), what they routinely do (habits), and what they prefer (preferences), and keep it updated as jobs, routines, and tastes drift. Existing benchmarks evaluate this "memory" ability through short, simplified interactions, missing three core properties of real behavior: the profile is heterogeneous, with attributes, habits, and preferences evolving on different timelines; changes are driven by external context such as seasons and life events; and evidence is rarely stated explicitly, instead scattered across many small actions in different apps that a memory system must infer from. We introduce DynamicMem, a synthetic benchmark that constructs 15 months of activity per user, providing long-term multi-app data that real users' privacy keeps out of reach. It provides user-consistent trajectories averaging 2.2M tokens and 1,772 grounded events per user across 16 applications such as e-commerce, fitness, and social platforms. The profile evolves over this period and is never given explicitly: each attribute, habit, or preference must be inferred from small signals scattered across apps. We evaluate at five quarterly checkpoints to track how systems scale as history grows. Benchmarking five representative systems exposes problems a single accuracy score hides: (i) profile reconstruction degrades with history length while service-task accuracy stays flat, despite both drawing on the same memory; (ii) no system both keeps facts that stay true and replaces facts that change, with errors clustering on preferences and on naming the exact referent; and (iii) over 93% of failures trace to what the memory retrieves, not to the model writing the answer, so the largest room for improvement lies in memory itself. Code: https://wenyaxie023.github.io/DynamicMem/

Coordinated foreign influence operations pose a growing threat to online platforms, but detecting state-linked troll activity and tracking its evolution remain challenging. This paper presents an explainable machine learning framework for theory-guided detection and longitudinal analysis of suspected trolling within Korean online news comment sections. Our hierarchical model classifies comments along three dimensions central to influence campaigns: foreign origin, moral-emotional framing, and target country. To support explainability, it also extracts brief span-level textual evidence that provides human-interpretable rationales. We apply the approach to 112M South Korean news comments authored by 4M users over nearly 20 years, identifying 23,998 accounts exhibiting behavior consistent with coordinated manipulation. Analyzing these accounts, we find that they predominantly rely on morally condemning rhetoric rather than direct promotion of foreign-aligned narratives; this rhetoric receives significantly higher user engagement. Among the highest-engagement comments, the moral condemnation most frequently targets domestic political figures (e.g., presidents or party leaders) on both the left and the right, potentially amplifying polarization. Our framework supports transparent platform governance through explainable, evidence-based moderation. These observed rhetorical and engagement patterns can inform how platforms and observatories prioritize defenses and intervene before harmful narrative-target combinations achieve widespread reach.