思维链可监控性的透明度披露：五大指标与开发者审计
Transparency Disclosures for Chain-of-Thought Monitorability
AI 导读
一篇 arXiv 论文提出，前沿 AI 开发者应披露五项指标以保障思维链（CoT）监控能力，涵盖第三方 CoT 访问与基础架构、CoT 质量（自然性与忠实性）及监控有效性（检测与提升）。对 Anthropic、OpenAI 和 Google DeepMind 的审计显示，无一家覆盖全部指标，但每项指标至少有一家开发者报告过。作者建议由嵌入式评估人员或独立验证机构开展跨开发者深度审计。
正文 · 原文
Abstract:Chain-of-thought (CoT) monitoring is one of the most valuable tools for monitoring advanced AI systems, but it is under threat. We argue that frontier AI developers should report specific indicators to make these threats trackable and to better facilitate coordination to preserve CoT monitorability. We identify five indicators that are both highly relevant and reasonably feasible to disclose, one for each link in a five-step chain on which CoT monitoring depends. The first, third-party CoT access and base architecture specifications, also underpins external verification of the rest. The others span CoT quality (naturalness and faithfulness) and monitoring effectiveness (detection and uplift). An audit of current disclosures by Anthropic, OpenAI, and Google DeepMind shows that no developer covers the full set, but that every indicator has been reported by at least one frontier developer. Because current disclosures rely largely on developers' own evaluations, we also recommend that a cross-developer audit be repeated with deeper access than public disclosures allow, either by embedded evaluators working within developers or by an Independent Verification Organization.
| Comments: | 27 pages (including references and appendix), 3 figures |
| Subjects: | Computers and Society (cs.CY); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.13873 [cs.CY] |
| (or arXiv:2611.13873v1 [cs.CY] for this version) | |
| https://doi.org/10.48550/arXiv.2611.13873 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Theo Bearman [view email]
[v1]
Tue, 10 Nov 2026 10:33:11 UTC (460 KB)
来源：arXiv cs.AI · arxiv.org