Tool use, function calling, orchestration and the protocols that let models act rather than only answer.
A patient's pathology case spans tens of slides from distinct specimens and anatomical sites, each carrying non-redundant, site-specific evidence. Forming a patient-level conclusion therefore requires...
Relying on manipulable characteristics to target the allocation of resources is fraught with difficulties. Naively optimal rules lead to falsification and allocative inefficiency. Optimizing the rule...
The paper investigates how an embodied agent can prioritize competing bodily needs using active inference. It compares two strategies: increasing the precision of information about an urgent need and reweighting preferences for outcomes that relieve that need, both individually and combined, within a foraging gridworld with lethal needs. The study finds that precision improves survival and sharpens interoceptive beliefs, while preference reweighting keeps states away from lethal thresholds and alters the agent’s learned body model, with both interventions positively interacting to enhance survival across varied layouts.
Reusable skills let LLM agents load task-specific instructions on demand, but they make execution depend on implicit routing over skill metadata. We test whether this design scales by holding SkillsBe...
Large vision-language models (LVLMs) have shown strong potential for embodied vision-language navigation (VLN), yet their performance degrades on long-horizon trajectories due to context-length limits...
The paper introduces Autonomous Science Bench, a benchmark for AI agents that must choose experiments under a limited budget. It features six real-world design-build-test-learn campaigns in biology, chemistry, and materials science, where agents select design batches, receive lab measurements, and iteratively refine a predictive model. Success is measured by two calibrated gates—discovery and learning—highlighting distinct agent weaknesses that overall success rates alone would mask.
The paper introduces Randomized Block Curation (RBC), a method for selecting useful lessons from language‑model agents’ persistent textual memory. RBC evaluates the impact of grouped lessons through randomized experiments on practice tasks, retaining only those that demonstrably improve task success. Experiments across two memory systems, three models, and three benchmarks show RBC raises average success to 66.9, outperforming both full playbooks and agent‑chosen lessons.
The article describes a developer’s strategy for ensuring that an agent uses a diagnostics library by implementing a pre‑commit hook that scans the git diff for direct calls to observability modules. If such calls are detected, the commit fails and a prompt is injected, instructing the agent to use the diagnostics library instead. This approach eliminates the need for manual code updates and maintains consistency across the codebase.
Agent harnesses require a low-latency guardrail model together with customizable audit policies. However, deploying an auto-regressive guardrail model remains challenging: such models incur high infer...
The article "Coding Agents Can Solve the Task — But How Much Work Did They Waste?" examines the performance of coding agents using real repositories and seeded agent decisions. It reports on 50,400 runs, highlighting metrics such as wasted turns, patch churn, and the effectiveness of different harness rules. The study provides insights into how these agents manage code modifications and the efficiency of their decision-making processes.
TextualRL is a gradient‑free reinforcement learning framework that updates the textual context guiding large language model agents without altering their parameters. It critiques groups of rollouts from the same task, contrasts successful and failed runs, and uses cross‑group evidence to propose and validate context edits. Across six benchmarks, TextualRL boosts test accuracy by 24.4 and 15.3 percentage points for GPT‑5.5 and Qwen3.8‑27B, respectively.
Do language models carry information about narrative events before writing them, and how far ahead? We examine narrative lookahead through salient, state-changing events rather than tokens or response...
LADDER is a long‑horizon robot manipulation benchmark that scales from 1 to 10 subgoals within the same environment, covering 145 tasks across three suites and two embodiments. The benchmark evaluates low‑level policies and orchestration methods, showing that while low‑level policies perform well on single‑subgoal tasks, they fail on extended tasks, and orchestration improves performance but is limited by execution errors. Tool‑selection orchestration with GPT‑Policy achieves the highest success rate (~62%) but at significant time and cost.
Effective collaboration among large language model (LLM) agents in multi-agent systems requires sharing relevant evidence, requesting missing information, and incorporating teammates' contributions in...
ECMR: Evidence-Aware Deliberation Control for Multi-Agent LLM Reasoning proposes a controller that dynamically manages deliberation steps in multi-agent LLM systems. The controller evaluates answer concentration, support diversity, contradictions, role behavior, verification history, and remaining budget to decide actions such as propose, challenge, verify, arbitrate, diversify, correct, or stop. In a held‑out evaluation on MedQA, CaseHOLD, and FinQA, ECMR‑Full improves pooled accuracy from 62.1% to 84.9%, lowers false‑consensus from 40.7% to 17.0%, and raises evidence‑support from 58.0% to 88.7%.
Foundation models for tables and relational databases are increasingly used for predictive tasks across various scientific and industrial domains. However, adapting existing models to unseen datasets...
Large language model (LLM) agents are increasingly used for vulnerability analysis, yet existing benchmarks provide limited coverage of exploitation that depends on interactions among network protocol...
The paper introduces MHVF, a training‑free framework for GUI grounding that uses multi‑hypothesis exploration and local visual verification with a single pretrained vision‑language model. MHVF generates spatial hypotheses from full‑screen predictions, refines them through zoomed re‑localization, and verifies candidate agreement with the instruction to resolve ambiguity. Experiments on eight benchmarks with five models show significant accuracy gains, notably improving UI‑Venus‑Ground‑7B by 4.3 pp on Mind2Web and raising Qwen2.5‑VL‑7B’s ScreenSpot‑Pro accuracy from 23.6 % to 53.6 %.
Policies trained offline from teleoperated demonstrations can degrade outside their training distribution. This gap is particularly acute in long-horizon whole-body mobile manipulation, where long hor...
LatentRiskBench is a new benchmark designed to evaluate state-dependent safety risks in enterprise agents, where the safety of an action depends on the workflow state rather than the action alone. It pairs risky trajectories with benign ones that share the same terminal action but differ in latent policy-relevant states, forcing models to reconstruct context instead of relying on action cues. The benchmark introduces residual risk metrics and demonstrates that state-level evaluation detects 96% of violations compared to only 9% for action-level methods, highlighting the inadequacy of existing guardrails.