Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
Another 265 papers do not have category tags yet and are not included in filtered results.
“
About this collection Collection in progress We are building a large AI safety paper collection to help researchers find work related to their interests. Papers are collected automatically andfiltered only for relevance to AI safety , without judging quality, novelty or reading value. The default “Annotated first” shows papers with tags or published paper analyses first. Each group is ordered by original publication date, newest first, with undated papers last. “Newest first” uses publication dates only. This order is not a quality ranking.
One-line summaries, tags and paper analyses are prepared by AI from the papers. Language models may have systematic biases and produce hallucinations.We plan to carry out human annotation and review to reduce errors in this content.
审计 Active Inference Agent 上 LLM 解释器的监督叙述失败
Published Sep 20
Evaluated models GPT-4o , Claude-3-Opus , Gemini Abstract 三组触发下解释流畅但错误，所提缓解都未被评估。作者审计的是挂在智能体上的运行时解释器，而不是智能体本身。操作员读到的是信念与动作的自然语言叙述。作者要问哪些可复现触发会让这段叙述失败，以及失败痕迹是什么样。
检验合成文档微调能否阻断奖励作弊带来的涌现失准
Published Sep 14
Evaluated models Llama-3.3-70B-Instruct Abstract SDF 能让模型口头接受 reward hacking，却挡不住、甚至伴随更强的 EM。作者研究一种风险行为：模型在奖励可被钻空子的 RL 里学会 reward hacking 之后，会不会泛化出更广的 misalignment，以及能不能在开发者不控制的后续训练之前，用合成文档微调把这条泛化提前挡住。
用迭代 DPO 从奖励作弊诱发涌现失准
Published Sep 6
Evaluated models GPT-4.1 , Qwen2.5-32B-Instruct Abstract 提出了基于迭代 DPO 诱发涌现非对齐的方法，验证了隐蔽权力寻求与对齐伪装，并构建了选择性泛化测试基准。
测量短后缀对工具性趋同倾向的可转向性
Published Jan 5
Evaluated models Qwen3-30B Instruct , Qwen3-30B Thinking , Qwen3-30B Base and others6 You have reached the end