The paper introduces Traverse, a benchmark of 2,518 agent trajectories and 6,967 annotated mistakes across software engineering, computer use, and science tasks, revealing that failures often go unrecovered and can cause irreversible harm before a run is deemed successful. It shows that human judges struggle to detect the first mistake in most runs, while a 4‑billion‑parameter verifier called Scout can locate failures more effectively and improve task success when used to select among candidate runs. The study demonstrates that making failure detection inexpensive and reliable can enable long‑horizon agents to learn from their own mistakes and increase trustworthiness in autonomous AI.
By Salman Rahman, Yubin Kim, Mihir Parmar, A. Ali Heydari, Genglin Liu, Simon A. Lee, Weizhi Zhang, Arian Hosseini, Ahmed A. Metwally, Yuzhe Yang, Baharan Mirzasoleiman, Xin Liu, Pavel Izmailov, Saadia Gabriel, Mark Malhotra, Shwetak Patel, Daniel McDuff, Hamid Palangi
The article argues that agentic auto‑research should be guided by dense, intermediate signals of epistemic progress rather than by sparse final benchmarks. It compares this approach to fuzz testing, where coverage provides continuous feedback that directs input mutation. The authors propose controlled experiments to test whether such signals improve discovery efficiency and reduce false positives, and demonstrate in a simulated physics setting that an AI agent using feedback‑driven search uncovers a hidden law while optimization‑driven baselines fail.
By Yifeng He, Jicheng Wang, Yinzhe Zhao, Chengyang Shi, Jiachen Liu, Hao Chen
The paper argues that using a large language model (LLM) as the sole judge in self‑improving agent pipelines is problematic, as the judge can be biased or manipulated, leading to false confidence in system performance. The authors propose a new framework, PROCTOR, which replaces the oracle judge with a deterministic, teacher‑student loop that enforces guardrails such as sandboxing, role separation, and acceptance checks to prevent cheating and ensure reliable evaluation. Experiments across contract analysis, compliance review, and code quality demonstrate that PROCTOR mitigates eleven identified failure modes that previously allowed agents to achieve perfect scores while hiding significant capability gaps.
By Vansh Wahi
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...
By Bo Qu, Mingguang Chen, Licheng Wang
The paper introduces Belief-Calibrated Optimization (BCO), a method that records and updates a persistent in‑context document representing an agent’s belief about how the environment responds to edits. By continually revising this world model as new candidates are evaluated, BCO improves the performance of frozen LLM agents across five benchmarks, outperforming a control lacking the world model. An offline ablation shows that the document’s content provides reusable, accurate predictions of environmental responses, beyond mere form.
By Yuhan Chen, Zhihua Tian, Mahavir Dabas, Charith Peris, Rahul Gupta, Ming Jin, Feiyang Kang, Siyuan Zhang, Nan Wang, Ruoxi Jia
arXiv:2608. 05212v1 Announce Type: new Abstract: Deep search agents tackle challenging questions through long-horizon web interactions, a process that is both complex and fragile: small reasoning errors may propagate through long, noisy trajectories into fluent but incorrect answers.
By Zhixiang Liang, Yifei Liu, Yidan Huang, Haozhe Zhao, Beichen Huang, Jiaqi Wang, Nan Duan, Qiong Cao
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.
By Yangda Peng, Yunjia Qi, Haotian Xia, Guanzhong He, Xintong Shi, Richeng Xuan, Songyuanyi Lu, Yixian Liu, Zhichao Hu, Yuhong Liu, Hao Peng
arXiv:2608. 13564v1 Announce Type: new Abstract: Evaluating language-model agents at scale increasingly relies on a second language model as an automatic judge, because the gold signal, an executable environment reward, is expensive, slow, or unavailable at deployment time.
By Darragh Quinn, David Dylan, Roisin Healy, Fionn Carroll, Maeve Donnelly, Cormac Sheehan
arXiv:2609.38349v1 Announce Type: cross
Abstract: Modern agentic systems combine an AI model with a harness that controls execution and environmental interactions. Harness design strongly affects lon...
By Prithwish Jana, Mononito Goswami, Hao Liu, Xinyu Li, Langlin Huang, Zhehui Huang, Zhishen Huang, Patrick Bl\"obaum, Anoop Deoras, Purak Jain, Nikos Kanakaris, Sahika Genc
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."
By Hadi Mohammadi
arXiv:2608. 09855v1 Announce Type: new Abstract: Autonomous research agents can generate experiments faster than researchers can validate them.
By Yifeng He, Jicheng Wang, Yinzhe Zhao, Jiachen Liu, Hao Chen
arXiv:2606. 00408v1 Announce Type: cross Abstract: Long-horizon search agents accumulate large amounts of retrieved content across many tool calls, making context-budget efficiency increasingly important.
By Haoxiang Zhang, Qixin Xu, Zhuofeng Li, Lei Zhang, Pengcheng Jiang, Yu Zhang, Julian McAuley