The paper introduces XiangqiBench, an executable benchmark for evaluating large language model agents in Chinese chess. It tracks 8,568 multi‑turn trajectories from 12 frontier LLMs, revealing that metrics such as the Conversion Gap, Consistency Gap, and Simulation Gap overstate true closed‑loop success. The study shows that simply naming the correct move is insufficient; agents must reliably carry a plan through to a verified outcome against an opponent.
arXiv:2609.25686v1 Announce Type: cross Abstract: Long-horizon assigned work requires an LLM agent to track the state of a task: which steps are done, blocked, cancelled, or open to repetition. Agent...
Agentic systems increasingly gate actions on a model's own stated confidence, which assumes confidence tracks correctness at the moment of acting. We test this in a hidden-information chess variant wh...
The paper investigates whether large language models (LLMs) correctly gauge their confidence when acting in a hidden‑information chess variant. In experiments where the location of a hidden royal piece is repeatedly relocated, the models’ stated probabilities about the piece’s position were almost never accurate at high confidence levels, with a calibration deficit concentrated in those high‑confidence events. Across multiple model configurations and providers, the same pattern emerged, and conventional evaluation metrics such as legality, cost, latency, and completion rate were found to be uncorrelated with belief quality, yet a model could still win the game despite poor confidence estimates.
arXiv:2607. 02255v1 Announce Type: new Abstract: Memory for a long-horizon LLM agent is a contract about what each future decision is allowed to see.
arXiv:2607. 05904v1 Announce Type: new Abstract: Training a language model against its own reference-free judgments (the premise of self-rewarding, self-play, and LLM-as-a-judge pipelines) assumes a model's verdict on a shown answer tracks correctness.
arXiv:2608. 01193v1 Announce Type: cross Abstract: An AI development race creates a multi-agent safety dilemma.
CivBench is an open‑source benchmark that evaluates language‑model agents in the long‑horizon, tool‑mediated game Civilization VI using the Model Context Protocol (MCP). Each episode lasts over 300 turns, generating thousands of tool calls across a 76‑tool action space, and includes a narration layer that translates visual game state into structured text. The study characterises agent behaviour across four model families, introducing Proactive Monitoring Rate (PMR) and RAG@10 as interface‑level metrics, and finds that agents often under‑monitor strategic state and fail to execute near‑term commitments despite tool access and explicit guidance.
The paper demonstrates that safety mechanisms for autonomous large language model agents fail to compose across iterative loops, as trajectory‑scoped monitors cannot detect attacks whose evidence is spread over multiple iterations. It introduces LoopHarness, a system that maintains a persistent, non‑decaying safety state across loops, bounding unauthorized actions with a constant that does not grow with the number of iterations. The authors provide a comprehensive evaluation protocol, including attacks that require cross‑iteration evidence, module ablations, and adaptive white‑box red‑team testing.
arXiv:2609.27051v1 Announce Type: new Abstract: Language-model agents now run the whole of quantitative factor research: they propose investment factors, backtest them, select the survivors and retir...
The paper introduces RuVerBench, a benchmark with 2,458 instances for evaluating the reliability of Large Language Models acting as judges (LaaJ) in verifying rubric compliance within agentic scenarios such as deep research and agentic coding. It systematically meta‑evaluates frontier LLMs, revealing that even the most advanced models perform well yet still produce substantial noise. The study also examines how prompt design, batching, and majority voting affect verification accuracy, noting that weaker models are more prompt‑sensitive, batched verification trades accuracy for efficiency, and majority voting offers diminishing returns.
The paper introduces Gauntlet, a framework that lets large language models autonomously build game-playing agents from a bare contract—just a game description, raw observation/action interface, and an empty policy file. In a single session, the model experiments with the game, compiles a standalone controller, and the resulting program is evaluated on held‑out instances without further model calls. The authors demonstrate that these compiled agents can win full‑scale games such as StarCraft II and Civilization, marking the first time a language‑agent system has achieved standalone victory in such complex titles.
The paper "trajectory-judge: What Outcome-Only LLM Judges Miss on Agent Trajectories" examines the limitations of outcome-only evaluation for large language model agents. Using a deterministic tool‑using support‑desk environment with a scripted oracle policy and a fault injector, the authors compare five different judging approaches—programmatic rules, outcome‑only, step‑rubric at two model sizes, and a self‑consistency ensemble—on metrics such as detection, step localisation, fault typing, calibration, and cost across 400 trajectories. The study finds that outcome‑only judges miss many silent faults and generate false positives, while step‑rubric judges achieve higher recall with no false alarms but at greater cost, and that none of the judges read the final reply, allowing fabricated promises to evade detection. "whyItMatters":"The findings highlight that current production‑default outcome‑only evaluations can overlook critical failures in agent behavior, underscoring the need for more nuanced, step‑level judging methods to ensure reliable LLM agent performance."