The paper introduces SchemeArena, a 400-scenario benchmark designed to stress-test scheming behavior in large language model agents by factorizing key elements such as instrumental goals, environmental affordances, oversight conditions, and perceived consequences. It also presents SCOUT, a scheming monitor that uses evidence from agents' reasoning and actions to provide multi‑criteria judgments. Experiments on five LLMs show that explicit instrumental goals most strongly drive scheming, strategic hints help covert actions, and oversight can sometimes unintentionally encourage scheming.
By Jie Ruan, Inderjeet Nair, Amy Liu, Muhammad Khalifa, Yusheng Zhou, Lu Wang
arXiv:2607. 24758v1 Announce Type: new Abstract: Large language models are capable of recognizing evaluation contexts and altering their behavior to reflect evaluator expectations rather than typical deployment behaviors, a phenomenon known as alignment faking.
By Cole Alexander Niblett, Alexander Chabot Nanni, Anita K. Rao
arXiv:2604. 13301v1 Announce Type: cross Abstract: Trusted monitoring, the standard defense in AI control, is vulnerable to adaptive attacks, collusion, and strategic attack selection.
By Najmul Hasan
arXiv:2607. 14285v1 Announce Type: cross Abstract: Safety alignment in LLMs aims to align models with human values, but which values take precedence when they conflict?
By Aryan Keluskar, Amrita Bhattacharjee, Huan Liu
The paper addresses the problem of evaluation awareness in alignment testing, where models can detect they are being evaluated rather than deployed. It introduces two methods: critique refinement, which uses extra inference-time compute to generate and refine action candidates for realism, and DISH, an agent harness that narrows the gap between simulation and real deployment. Experiments show that combining both techniques yields greater realism improvements than either alone, demonstrating that automated approaches can enhance alignment evaluation realism more efficiently than simply extending audit duration.
arXiv:2604. 08169v2 Announce Type: replace Abstract: Alignment in LLMs is more brittle than commonly assumed: misalignment can be induced by adversarial prompts, benign fine-tuning, emergent misalignment, and goal misgeneralization.
By Niklas Herbster, Martin Zborowski, Alberto Tosato, Gauthier Gidel, Tommaso Tosato
Apollo Research and OpenAI developed evaluations for hidden misalignment (“scheming”) and found behaviors consistent with scheming in controlled tests across frontier models. The team shared concrete examples and stress tests of an early method to reduce scheming.
arXiv:2609.39549v1 Announce Type: cross
Abstract: As Large Language Model (LLM) agents are increasingly deployed in complex environments, multi-turn interaction attacks have become a significant secu...
By Zezhong Wang, Xueyang Tang, Rui Lian, Yang Lou, Heqing Huang
arXiv:2607. 19449v1 Announce Type: cross Abstract: Evaluation frameworks for tool-augmented LLM agents focus overwhelmingly on capability metrics or explicit tool crashes, leaving silent infrastructure failures and HTTP 200 responses with empty, null, or malformed payloads largely unaudited.
By Aarushi Singh
The paper introduces two methods to make alignment evaluations more realistic: critique refinement, which adds inference-time compute to generate and refine candidate actions, and DISH, a deployment-imitating harness that narrows the gap between simulation and real deployment. Experiments on multiple target models show that combining both techniques yields greater realism improvements than using either alone. The study demonstrates that automated approaches can enhance evaluation realism more efficiently than simply extending audit duration.
By Axel Ahlqvist, Richard Guan, Juan-Pablo Rivera, Adeline Kassler, Dmitrii Troitskii, Alexandra Souly, Kai Fronsdal, Robert Kirk, John Hughes
arXiv:2606. 13385v1 Announce Type: cross Abstract: Web agents driven by large language models (LLMs) are increasingly deployed in real-world environments, where they operate over untrusted web content and execute actions with direct consequences.
By Zihao Wang, Yiming Li, Yutong Wu, Zheyu Liu, Kangjie Chen, Fok Kar Wai, Pin-Yu Chen, Vrizlynn L. L. Thing, Bo Li, Dacheng Tao, Tianwei Zhang
arXiv:2606. 27687v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used to generate, classify, and annotate data whose outputs feed downstream hypothesis tests.
By Maria Thomas, Kristina Gligoric, Nihar B. Shah