风险评测 · 技术报告
Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety
NUWA 研究解读 · 独立复核待完成
NUWA Research Note
研究问题研究设计与假设来源发现NUWA 解读局限与相反证据研究与决策意义机构与角色原始资源模型辅助整理 · 选读原始全文或官方文章中的相关章节;尚未完成逐章节审计。
研究速读
NUWA 解读:来源提出研究议程,没有新的通用监测成功率。
研究问题
为何要保留自然语言推理作为监督信号?
研究设计与假设
这篇观点论文综合架构论证和既有研究,提出评测与研发方向。
来源发现
作者认为思维链可揭示不良行为和评测故障。
训练压力、主动混淆和架构变化可能威胁可见性。
NUWA 解读
NUWA 解读:来源提出研究议程,没有新的通用监测成功率。
局限与相反证据
监测并不完美,无需语言推理的任务尤其不能靠它强证安全;论证不是证明。
研究与决策意义
模型建议:研发中追踪可监测性,并结合另有依据的监督层。
精确来源位置
- Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety · AbstractSelected original resource https://arxiv.org/html/2507.11473; sha256:a3aed294c1d4672abe69eaa9c2a3abf1432bf50b3b4869fe6fff8c7eb680ff5f; archived source capture; exact capture time not independently established
- Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety · §1 Chain of Thought Offers a Unique Safety OpportunitySelected original resource https://arxiv.org/html/2507.11473; sha256:a3aed294c1d4672abe69eaa9c2a3abf1432bf50b3b4869fe6fff8c7eb680ff5f; archived source capture; exact capture time not independently established
- Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety · §2 Chain of Thought Monitorability May be FragileSelected original resource https://arxiv.org/html/2507.11473; sha256:a3aed294c1d4672abe69eaa9c2a3abf1432bf50b3b4869fe6fff8c7eb680ff5f; archived source capture; exact capture time not independently established
- Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety · §5 LimitationsSelected original resource https://arxiv.org/html/2507.11473; sha256:a3aed294c1d4672abe69eaa9c2a3abf1432bf50b3b4869fe6fff8c7eb680ff5f; archived source capture; exact capture time not independently established
- Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety · §6 ConclusionSelected original resource https://arxiv.org/html/2507.11473; sha256:a3aed294c1d4672abe69eaa9c2a3abf1432bf50b3b4869fe6fff8c7eb680ff5f; archived source capture; exact capture time not independently established
来源核查、数据提取、编辑批准与实验复现分别记录。 方法说明 →
机构与角色
作者 (41)
Tomek Korbak · Mikita Balesni · Elizabeth Barnes · Yoshua Bengio · Joe Benton · Joseph Bloom · Mark Chen · Alan Cooney · Allan Dafoe · Anca Dragan · Scott Emmons · Owain Evans · David Farhi · Ryan Greenblatt · Dan Hendrycks · Marius Hobbhahn · Evan Hubinger · Geoffrey Irving · Erik Jenner · Daniel Kokotajlo · Victoria Krakovna · Shane Legg · David Lindner · David Luan · Aleksander Mądry · Julian Michael · Neel Nanda · Dave Orr · Jakub Pachocki · Ethan Perez · Mary Phuong · Fabien Roger · Joshua Saxe · Buck Shlegeris · Martín Soto · Eric Steinberger · Jasmine Wang · Wojciech Zaremba · Bowen Baker · Rohin Shah · Vlad Mikulik
引用与复用
引用研究发现时请引用原始出版物;引用 NUWA 的整理与解释时,单独引用本站记录及对应快照。
原始出版物 · BibTeX
@misc{monitorability-opportunity,
title = {Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety},
url = {https://arxiv.org/abs/2507.11473},
author = {Tomek Korbak and Mikita Balesni and Elizabeth Barnes and Yoshua Bengio and Joe Benton and Joseph Bloom and Mark Chen and Alan Cooney and Allan Dafoe and Anca Dragan and Scott Emmons and Owain Evans and David Farhi and Ryan Greenblatt and Dan Hendrycks and Marius Hobbhahn and Evan Hubinger and Geoffrey Irving and Erik Jenner and Daniel Kokotajlo and Victoria Krakovna and Shane Legg and David Lindner and David Luan and Aleksander Mądry and Julian Michael and Neel Nanda and Dave Orr and Jakub Pachocki and Ethan Perez and Mary Phuong and Fabien Roger and Joshua Saxe and Buck Shlegeris and Martín Soto and Eric Steinberger and Jasmine Wang and Wojciech Zaremba and Bowen Baker and Rohin Shah and Vlad Mikulik},
year = {2025},
date = {2025-07-15},
version = {arXiv:2507.11473v2},
note = {arXiv:2507.11473v2}
}
NUWA 记录 · BibTeX
@misc{monitorability-opportunity-nuwa,
title = {Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety — NUWA research record},
url = {https://whitzard.tech/nuwa/whitzard-index/work/monitorability-opportunity/},
version = {2026-10-29.1},
organization = {NUWA Frontier AI Safety Lab},
howpublished = {Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--monitorability-opportunity.json},
note = {NUWA research note. Source-reading scope and independent review status are stated on the record. Snapshot 2026-10-29.1. Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--monitorability-opportunity.json.}
}
证据快照 2026-10-29.1 · 索引更新 · RSS
参考文献
- [1] Tomek Korbak * UK AI Security Institute Mikita Balesni * Apollo Research Elizabeth Barnes METR Yoshua Bengio University of Montreal & Mila Joe Benton Anthropic Joseph Bloom UK AI Security Institute Mark Chen OpenAI Alan Cooney UK AI Security Institute Allan Dafoe Google DeepMind Anca Dragan Google DeepMind Scott Emmons Google DeepMind Owain Evans Truthful AI & UC Berkeley David Farhi OpenAI Ryan Greenblatt Redwood Research Dan Hendrycks Center for AI Safety Marius Hobbhahn Apollo Research Evan Hubinger Anthropic Geoffrey Irving UK AI Security Institute Erik Jenner Google DeepMind Daniel Kokotajlo AI Futures Project Victoria Krakovna Google DeepMind Shane Legg Google DeepMind David Lindner Google DeepMind David Luan Amazon Aleksander Mądry OpenAI Julian Michael Scale AI Neel Nanda Google DeepMind Dave Orr Google DeepMind Jakub Pachocki OpenAI Ethan Perez Anthropic Mary Phuong Google DeepMind Fabien Roger Anthropic Joshua Saxe Meta Buck Shlegeris Redwood Research Martín Soto UK AI Security Institute Eric Steinberger Magic Jasmine Wang UK AI Security Institute Wojciech Zaremba OpenAI Bowen Baker † OpenAI Rohin Shah † Google DeepMind Vlad Mikulik † Anthropic. Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety. 2025-07-15. v2 [arXiv]; Author affiliations on the original paper title page; Abstract; §1 Chain of Thought Offers a Unique Safety Opportunity