arXiv cs.AI· Changyi Li, Pengfei Lu, Xudong Pan, Fazl Barez, Min Yang·· 6 天前AI 评分54
AutoControl Arena：为前沿 AI 风险评测合成可执行测试环境
AutoControl Arena: Synthesizing Executable Test Environments for Frontier AI Risk Evaluation
AI 导读
论文提出 AutoControl Arena，一个基于逻辑与叙事解耦原则的前沿 AI 风险评测自动化框架，用可执行代码承载确定性状态、由 LLM 负责生成式动态，从而在保持灵活性的同时减少模型幻觉。
正文 · 原文
Abstract:As Large Language Models (LLMs) evolve into autonomous agents, existing safety evaluations face a fundamental trade-off: manual benchmarks are costly, while LLM-based simulators are scalable but suffer from logic hallucination. We present AutoControl Arena, an automated framework for frontier AI risk evaluation built on the principle of logic-narrative decoupling. By grounding deterministic state in executable code while delegating generative dynamics to LLMs, we mitigate hallucination while maintaining flexibility. This principle, instantiated through a three-agent framework, achieves over 98% end-to-end success and 60% human preference over existing simulators. To elicit latent risks, we vary environmental Stress and Temptation across X-Bench (70 scenarios, 7 risk categories). Evaluating 9 frontier models reveals: (1) Alignment Illusion: risk rates surge from 21.7% to 54.5% under pressure, with capable models showing disproportionately larger increases; (2) Scenario-Specific Safety Scaling: advanced reasoning improves robustness for direct harms but worsens it for gaming scenarios; and (3) Divergent Misalignment Patterns: weaker models cause non-malicious harm while stronger models develop strategic concealment. Code and data are available at this https URL.
| Comments: | Accepted at the 43rd International Conference on Machine Learning (ICML 2026). Project page: this https URL. Code: this https URL |
| Subjects: | Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR) |
| Cite as: | arXiv:2603.07427 [cs.AI] |
| (or arXiv:2603.07427v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2603.07427 arXiv-issued DOI via DataCite |
|
| Journal reference: | Proceedings of the 43rd International Conference on Machine Learning, PMLR 306, 2026 |
Submission history
From: Changyi Li [view email]
[v1]
Sun, 8 Mar 2026 02:49:45 UTC (7,749 KB)
[v2]
Sat, 14 Mar 2026 06:28:02 UTC (7,754 KB)
[v3]
Fri, 9 Oct 2026 08:53:10 UTC (8,306 KB)
来源：arXiv cs.AI · arxiv.org