Vision-language-action models enable instruction-conditioned robot manipulation, but they typically cannot improve from experience after deployment. This paper proposes VLRLA, a framework that equips frozen VLA models with in-context reinforcement learning by conditioning action generation on a recurrent memory of past observations, actions, rewards, and termination signals. Using algorithm distillation on constructed learning histories, VLRLA aims to gradually improve task success during deployment without updating network parameters.

Coding agents consume substantial tokens as they repeatedly decide where to look, what to try, and when to stop during automated repair. This paper asks which of these decisions can be offloaded to a cheap probabilistic judge that answers fixed questions without generating explanations. The authors introduce JevCoder, a workflow built around the proprietary Jev judge to recommend initial code snippets, guide action types during repair, and trigger submission. On 400 SWE-bench tasks across three repair models, the system cuts repair-model token use by $21.6\%$–$50.6\%$ while raising observed repair rates.

This paper tackles the problem that agentic scientific workflows can finish without crashing yet still emit scientifically invalid results due to silent faults, stale inputs, or contaminated shared ancestors. The core idea is a directed acyclic graph harness that treats trust as a release decision: it enforces mandatory checks, contains failures to affected branches, and uses a fixed risk model plus optional validators to decide whether a deliverable may be released. Language-model agents operate inside this layer, choosing corrective actions and reviewing releases, but cannot override it. On two synthetic workflows with injected faults, the authors decompose which harness component supplies which property, finding that structure, not the LLM, drives reliability.

MemCode targets the long-term memory bottleneck in LLM agents by arguing that standard continuous vector retrieval misses logically related but lexically distant facts. It proposes retrieval-oriented atomic memory units (AMUs) and indexes them with a learned discrete codebook via isomorphic semantic quantization (ISQ), enabling dual-path retrieval that combines dense vectors with associative codebook topology. On LongMemEval-S and LoCoMo, it reports the best overall accuracy and a construction runtime roughly $3\sim21\times$ faster than strong baselines.

Data augmentation for document-level NLP risks breaking inter-sentential coherence and invalidating sentence-aligned labels. This paper proposes an LLM-driven discourse parsing framework that preserves sentence order, count, and rhetorical relations $R_{i,j}$ while replacing key phrases $\mathbf{K}_i$ to generate augmented documents $\mathbf{D}'$. By keeping the $n$-sentence structure fixed, labels such as summary indices or QA spans can be transferred without expensive LLM relabeling.

Language models frequently derail on multi-step reasoning because a single early error propagates through the rest of the generation. REWIND attacks this by treating token-level entropy as a diagnostic: it traces backward from high-entropy peaks to the sharpest local gradient jumps, uses those positions as cut points for suffix regeneration, and filters proposals with a likelihood-based acceptance rule. The result is a targeted, test-time repair loop that requires no fine-tuning or external verifier and that the authors show improves accuracy across multiple model families and reasoning benchmarks.

Existing agent benchmarks generally assume cooperative, emotionally stable users, overlooking a major source of real-world failure. This paper introduces τ-Emotion-Bench, a ten-domain benchmark that evaluates tool-calling agents under controlled emotion-conditioned user interactions, together with Tracer, a pipeline that collects training trajectories through an interaction-aware harness where dissatisfied users can terminate inefficient exchanges. It matters because it reveals large reliability gaps in frontier models and shows that small models can be trained to substantially improve interaction consistency with far less data than prior methods.

PFedMA tackles reliable multi-round collaboration in Federated Multi-Agent Systems by jointly adapting agent selection and heterogeneous privacy rules as reasoning unfolds. Its core idea is a closed-loop orchestration that feeds response reliability and capability signals back into an RL-trained router, while a bidirectional local-global mechanism co-evolves privacy rules without overriding personalized constraints. This matters because existing approaches either select agents without historical accountability or enforce static privacy rules that cannot adapt to changing information needs across rounds, leading to error propagation and rigid privacy-utility trade-offs.

TEX-MEX attacks the central challenge in time-series contrastive learning: how to define similarity when semantic-preserving augmentations are scarce or unreliable. It unifies structural and temporal similarity by letting the encoder grade windows by cosine similarity, then pulling an anchor toward the time-shifted versions of similar batch windows and pushing it away from those of dissimilar ones. The result is a graded, efficient objective that learns what typically accompanies a physiological pattern in time, and it powers a 560K-parameter PPG foundation model that beats architectures roughly 50× larger.

Standard expected-reward RL suffers from mode collapse, concentrating probability on a few outcomes even when tasks require many distinct, high-quality solutions. MIPS tackles this by estimating marginal outcome probabilities through a learned trajectory-conditional model, avoiding frequency-based estimation that breaks down when the outcome space is much larger than the rollout group. Using the identity $P_\theta(o) = P_\theta(\tau) / P_\theta(\tau|o)$, it scales rewards by inverse estimated probabilities within a standard GRPO update. On enumerable Hypergrids, molecular synthesis, and phylogenetic inference, MIPS maintains reward-aligned diversity where group-count IPS collapses.

Advertising delivery still relies heavily on manual coordination across budgeting, bidding, and diagnosis, limiting scale and performance. AdCoEvo automates this pipeline by assigning Planner, Optimizer, and Inspector roles to an 8B-parameter language model and updating their LoRA adapters in rotation using settled business feedback, so the agents mutually adapt during continuous simulated delivery. The authors report gains of roughly 14% in advertiser value and 7% lower cost per acquisition over rule-based baselines in the AuctionNet simulator.

Wrist-worn IMUs power everything from gesture recognition to clinical motor assessment, yet most prior benchmarks evaluate each task in isolation, leaving unclear whether a single pretrained representation can serve all of them. This paper introduces IMU-Bench, a unified evaluation of ten tasks spanning kinematic estimation, semantic retrieval, activity recognition and Parkinson's assessment, and proposes InrtlWrist, a masked autoencoder pretrained on recorded and simulated wrist motion. The core finding is that pretraining yields broadly useful features—leading on nine of ten frozen-probe tasks—but that supervision from one downstream task rarely helps unrelated ones, distinguishing universal representation quality from task compatibility.

SERO attacks a core problem in long-horizon agentic RL: outcome-based credit assignment cannot teach behaviors that never appear in the policy's own rollouts. The idea is to restore snapshots of states actually reached by the current policy, let a training-only oracle continue interacting with the real environment, and distill the resulting executable multi-step trajectories into the learner via an auxiliary reachability-weighted loss. This extends the supervision horizon as the policy improves and remains compatible with existing group-based RL algorithms, providing dense grounded guidance without touching the base estimator.

The paper tackles the ambiguity in prompt-injection detection: internal-state probes may detect injection presence rather than actual behavioral compromise. It evaluates five adapted methods on the same structured prompts for two models, stratifying by attack success, and finds that attention- and role-based signals correlate more strongly with successful attacks than presence- or recognition-based probes.

Infectious disease surveillance is hampered by fragmented data sources, uneven evidence availability across regions, and manual processing delays. EpiGuard proposes an autonomous LLM agent that proactively monitors surveillance feeds, dynamically selects forecasting strategies across absent, sparse, and abundant evidence regimes, and accumulates reusable standard operating procedures from successful warnings. The platform promises faster, more reliable early warning for over 50 diseases, but its experimental validation covers only 15 time series and never tests the absent or sparse branches it advertises.

Recommendation systems are opaque black boxes that curate feeds for billions of users, raising serious assurance risks. xLightGCN targets the popular LightGCN architecture and turns its linear graph propagation into an exact accounting of which user and item neighbors most influence a prediction. The result is a compact, inspectable subgraph plus a quantitative fidelity score that tells auditors how much of the score that subgraph actually explains—an idea that matters because interpretability alone does not guarantee the explanation covers the model’s reasoning.

This paper tackles physical reasoning in multimodal language models by asking when chain-of-thought (CoT) suffices and when physics-grounded video simulation is necessary. The authors propose Hybrid Simulation and Reasoning (HSR), a small MLP-based router that dynamically assigns questions to either CoT or a video diffusion simulator, aiming to balance accuracy against computational cost. The core claim is that this resource-rational tradeoff improves performance while allowing a 4B-parameter model to control routing for much larger base models like Gemini 3 Pro.

Static benchmarks for LLM agents saturate and face contamination, making it hard to measure true progress over time. EvoProbe tackles this by evolving cybersecurity CTF tasks in a closed loop: generate executable tasks, diagnose solver trajectories with multi-judge scoring and phase-level subtask outcomes, and feed distilled bottlenecks back into task regeneration. Its core premise is that fine-grained diagnostics—not just aggregate pass rates—should dictate what the next benchmark probes.

Cross-lingual financial analysis requires models to both recognize shared accounting concepts across languages and exhaustively recover all corresponding numerical facts. This paper introduces MultiFinGround, a benchmark built from real FY2025 ESEF reports in English, French, and Greek, aligned through the IFRS Accounting Taxonomy. The core finding is that strong concept recognition does not imply complete evidence grounding: even top models reach near-perfect recognition accuracy while missing target facts in over a quarter of correctly recognized cases. This matters because benchmarks that only test recognition can obscure critical failures in numerical evidence completeness.

Legal reasoning is not merely the mechanical application of rules to facts; it is shaped by a judge's underlying legal philosophy. This paper introduces LePhilBench, the first benchmark to test whether LLMs can both understand jurisprudential tendencies and apply them to predict judicial outcomes. It pairs jurisprudential profiling with philosophy-aware judgment prediction and grounds the tasks in a dataset of 220 annotated U.S. Supreme Court opinions, offering a rigorous new lens on models' legal reasoning capabilities.