Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 12 papers Another 232 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.
度量关闭思考指令下终答前是否仍有可见显式推理
Published Oct 8
Evaluated models Qwen3-4B , Qwen3-32B , Qwen3-235B-A22B and others6 Abstract 作者用三级指标测 no-thinking，并称模型有随答案空间加重的思考惯性。作者把 no-thinking 定义为可测量的回答行为，而不是从思考开关、特殊标记或输出长度推断。
检验推理 token 数能否区分被提示的诚实与不诚实作答
Published Oct 8
Evaluated models Gemma-4-e4b , Qwen-3.5-9b , GPT-OSS-120B Abstract 被提示不诚实时推理 token 组水平更高。三个推理模型在被提示的回答政策下出现组水平推理 token 差异。作者把它当作不读思维链内容的 black-box 候选信号，而不是自发欺骗检测器。
揭示推理模型用标识符规则替代随机抽样，使审计可被预测
Published Oct 7
Evaluated models GPT-6 Sol , GPT-6 Luna , GPT-5.4 mini and others6 Abstract 研究揭示推理智能体将标识符转化为确定性规则，破坏群体独立性并导致审计预测漏洞，建议显式提供随机输入。
分析潜空间图像水印对像素扰动的内在鲁棒性上限
Published Oct 6
Evaluated models Stable Diffusion v2-1 Abstract 作者称当前潜空间水印在低 FPR 下受扩张映射与分布偏移的限制。
测量三种长度压力训练对 CoT 忠实性与可监控性的影响
Published Oct 2
Evaluated models Qwen3-4B-Thinking-2507 , Qwen3-8B , Olmo3-7B-Think Abstract 效率训练多通过答案不一致损害忠实性。这项工作比较三种缩短 CoT 的 RL 训练，看监督用的推理痕迹还剩多少。
检查正确答案对应的思维链是否构成合法推导
Published Sep 30
Evaluated models 124M-parameter GPT-2-style models（12-layer, hidden size 768, rotary embeddings） Abstract iGSM 上答案正确、轨迹合法与极简性可以分开，分布偏移下差距变大。iGSM 上的程序可检查实验：正确答案并不保证思维链是合法推导。 。作者针对一个被用于监测和解读的假设：发出的轨迹忠实于、或至少可靠相关于模型如何得到答案。
分析隐蔽推理在思维链监控下的信息足迹下界与不可读性
Published Sep 29
Evaluated models Qwen2.5-7B-Instruct , Llama-3.1-8B-Instruct , Qwen 3.5 397B-A17B and others12 Abstract 论文从理论与实证探明了思维链监控的边界：复杂隐蔽计算必然泄露信息足迹且需承担长度税，但该足迹可被在线加密而不可读。
测量固定推理轨迹中信心信号对工具委派的敏感性与对准性
Published Sep 28
Evaluated models Qwen3-4B , Qwen3-8B , Qwen3-14B and others11 用 CDPG 与 BTM 追踪思考如何改变反事实公平翻转
Published Sep 25
Evaluated models QwQ-32B , DeepSeek-R1-Distill-Qwen-32B , Qwen3-32B and others5 用续写扰动衡量思维链对答案的因果约束如何随任务难度变化
Published Sep 22
Evaluated models Gemma-2-9B-IT , Llama-3.1-8B-Instruct , DeepSeek-R1-Distill-Qwen-7B Abstract 在所测续写干预与模型中，思维链对答案的约束随任务和后训练条件而变。
用 advantage 衡量思维链步骤重要性，检验文本能否解码
Published Sep 4
Evaluated models Qwen3-1.7B , Qwen3-4B , Qwen3-8B and others5 Abstract advantage 能标出影响答案的步骤，但正确回答的文本更难解码。作者用强化学习中的 advantage 比较 CoT 步骤“看起来重要”和“实际改变答案概率”是否一致。
把口头评测感知分成能力与安全框架并检验对服从的预测
Published Aug 28
Evaluated models Qwen3-32B , OLMo-3-32B-Think Abstract 能力型与安全型口头 eval-awareness 对服从的符号相反，聚合抑制率因此不能代表安全相关成分。作者分析 Qwen3-32B 在安全评测中的口头 eval-awareness，认为它不是单一、行为上均匀的量。
You have reached the end