PROBE is an algorithm that trains language‑model auditors to probe deployed tool‑using agents for policy violations through realistic, multi‑turn, tool‑mediated dialogue. By framing auditing as a two‑player Markov game and using a composite reward that balances violation discovery, behavioral plausibility, conversation progress, and policy diversity, PROBE outperforms zero‑shot adversaries in uncovering broader and higher‑yield policy violations. The study compares GRPO, DPO, and GFlowNet optimizations, demonstrating that multi‑turn behavioral auditing is a distinct and necessary axis of agent evaluation, complementing existing capability and safety benchmarks.
By Kshitij Mishra, Abhijith Sharma, Nils Lukas, Salem Lahlou
arXiv:2607. 10814v1 Announce Type: cross Abstract: Evaluating LLM agents in hidden-information multi-agent settings is hard: final outcomes are high-variance and rarely reveal why an agent decided as it did.
By Yuan Gao, Jiangyi Yang, Yao Zhao, Yichi Zhang
The paper studies how to safely delegate action approval to multiple AI reviewers when the reviewers themselves may be misaligned. It introduces a weaker condition—k‑robust coalitional alignment—under which a threshold rule that tolerates up to k disapprovals guarantees that the principal’s expected utility is at least as good as a baseline policy. The authors extend this characterization to sequential decision‑making in discounted MDPs and show that full‑panel coverage of reward functions ensures safety in Nash equilibria, while more permissive thresholds can lead to unsafe outcomes. Experiments demonstrate that collective review can remain sound even when individual reviewers are not fully aligned, provided some disapprovals are allowed.
By Natalie Collina, Surbhi Goel, Aaron Roth, Sikata Bela Sengupta
arXiv:2607. 07435v1 Announce Type: cross Abstract: Agents acting on our behalf in the real world (e.
By Bojie Li, Noah Shi
The paper investigates the problem of sharing a single critic across multiple parallel environments in reinforcement learning. It shows that when environments assign different expected returns to the same state, a shared critic must reconcile conflicting value targets, which can distort advantage estimates and misguide policy updates. The authors propose a simple fix—providing the critic with the environment index—demonstrating through bandit models and experiments on CartPole, MuJoCo, BipedalWalker, and 16 Procgen games that this conditional critic stabilizes learning and boosts returns, achieving a 40.8% improvement in aggregate normalized return on unseen levels.
By Zhenya Liu, Yang Meng, Zhuokai Zhao, Xuefeng Liu, Yuxin Chen
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.
By Chenyu Zhou
The paper introduces the global coherence problem, where AI agents make locally valid decisions that collectively lead to an invalid outcome due to shared state failures. It presents the Observation‑Aliasing Impossibility Theorem, establishing that a policy can guarantee a valid action only when all indistinguishable worlds share an admissible action, and shows that even with additional reasoning, roles, messages, or samples, the missing distinction cannot be recovered. The authors propose a local‑to‑global runtime semantics framework and conduct nine studies demonstrating that missing global state cannot be substituted by local intelligence.
By Xin Heng
arXiv:2605. 06340v2 Announce Type: replace-cross Abstract: Continuous post-deployment compliance audits, mandated by emerging regulations such as the EU AI Act and Digital Services Act, create a class of strategic gaming distinct from the one-shot input/output gaming studied in prior work.
By Florian A. D. Burnat, Brittany I. Davidson
The paper introduces a protocol for auditing and composing reinforcement‑learning policies using discrete behavioral rules, defining auditability through six testable predicates such as trace integrity and rule coverage. Experiments show that overlapping rule sets do not guarantee behavioral agreement, and that rule‑based fusion often fails to outperform value‑based composition, highlighting limitations in current description layers. The authors provide an evidence‑bounded audit framework and outline future directions for more robust skill composition.
By Liu Hung Ming
arXiv:2607. 00155v1 Announce Type: new Abstract: We study runtime human oversight of an AI agent when private information runs in both directions: the human privately knows her reward function, while the AI privately knows the quality of the action it proposes.
By Yunjin Tong
The paper introduces HICO-Bench, a benchmark of 180 everyday coordination scenarios where two agents with private information converse and then record a shared plan. A program verifies that the plans are feasible and an LLM judge checks that the conversation supports the plan, defining a Verified Success. Using Paired-Credit Policy Optimization (PaCPO), the authors improve the Verified Success rate of Qwen3-8B from 20.4% to 28.3%.
By Caixin Kang, Ke Wang, Yuchuan Wu, Xiaoqiang Wang, Ling Li, Bingxue Qiu, Yongbin Li, Yoichi Sato
arXiv:2606. 08919v1 Announce Type: new Abstract: As LLM agents begin to take real, irreversible actions (shell commands, file edits, deploys), the standard safety pattern is a human-in-the-loop approval gate: risky actions pause and wait for a person.
By Emre Turan