Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 42 papers 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.
Uses premise permutation and hidden-layer tracing to audit whether legal chain-of-thought truly relies on the cited legal provisions.
Published Oct 9
Evaluated models Qwen3-8B , Qwen3-14B , Llama-3.1-8B and others7 Abstract Using counterfactual substitution of authorities and internal commitment tracing, the authors find that legal CoT rulings depend on authorities only to a limited extent and are vulnerable to prompt injection manipulation.This study conducts an empirical audit of the causal truthfulness of legal citations in chain-of-thought reasoning of large language models in the legal domain.
Auditing LLM judge reliability and defining trustworthy verdict rate and accuracy upper bound
Published Oct 8
Evaluated models GPT-5.2 , GPT-4o , Gemini-2.5-Pro and others6 Abstract The authors claim that comparing judges should look at T, not just accuracy.The authors audit whether LLM-as-a-Judge decisions can still be treated as actionable ground truth under repeated queries, swapped candidate order, and changed prompt formats.
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。
提出 Persist-3 序贯检验,从隐藏表示识别未见越狱源
Published Oct 7
Evaluated models Llama2-7B-chat , Qwen2.5-7B-Instruct , Gemma-2-9B-it and others4 Abstract Persist-3 用有限窗口检验攻击流是否超出已知越狱源,检出率低于全历史 SR-DKW,但决策只依赖最近观测。Persist-3 是一个有限记忆的序贯检测器,用来判断攻击表示流是否已经不能由任何一个已知越狱源解释。 。
提出 U-SPACE 与 U-LENS,从残差流估计推理轨迹不确定性
Published Oct 7
Evaluated models Gemma-4-31B-it , Qwen3.5-27B , Magistral-Small-2507 Abstract U-SPACE 与 U-LENS 用语义方向和预测熵读推理不确定性,长度控制后仍高于所比基线。U-SPACE 与 U-LENS 是面向推理模型的无训练读出:用语义方向定位不确定性的形式,并用一个轨迹分数区分更可能错的回答。
评测文生图模型的有害内容生成能力与审核检测缺口
Published Oct 5
Evaluated models SDXL , Stable Diffusion 3.5 Large , FLUX.1-dev and others12 提出 VERA,审计 LLM 漏洞解释里的虚构推理
Published Oct 5
Evaluated models sonnet-5 , gemma-4-26b , llama-4-maverick and others8 提出 PyINE 基准,训练捷径跟随模型并比较监督者
Published Oct 4
Evaluated models shortcut-following model organism , gpt-4o , gpt-5 and others12 Abstract PyINE 用可验证代码执行研究捷径失败的监督覆盖与成本权衡。PyINE 是用可执行程序研究推理模型引出与监督的框架,PyINE-v1 是其第一次发布。
系统评估循环语言模型加深 loop 后的 CoT 可监控性
Published Oct 2
Evaluated models Ouro-1.4B-Thinking , Ouro-2.6B-Thinking , Nanbeige4.2-3B and others4 Abstract 加深 loop 可在部分压力测试任务降低可监测性,循环架构本身并不必然更难监测。这项工作评估 LoopLM 的 CoT 可监测性:加深共享层的重复次数,以及“有循环结构”本身,会不会让决策关键因素更少出现在可监测文本里。
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。
比较隐写推理与隐写传信、编码推理的可学性
Published Sep 30
Evaluated models GPT-5.5 , Claude Sonnet 5 , Gemini 3.5 Flash and others14 Abstract 隐写推理比传信和编码推理更难学。Schulz、Fülle 与 Frengel 测量大语言模型习得隐写推理的难度,并与隐写传信、编码推理对比,以估计依赖阅读 CoT 的监控还能否成立。
构建编码 RL 奖励作弊测试台 CATCH 并检验 CoT 监控失效
Published Sep 30
Evaluated models Qwen3-4B , Qwen3.5-27B Abstract CATCH 用双运行金标签研究编码 RL 的奖励黑客,并观察到监控会被注释适应削弱。CATCH 是面向编码 RL 奖励黑客的可控分析测试床,用可利用环境、可调初始倾向和基于执行的金标签,在训练全程比较黑客动态与干预。
提出 CIA 指标衡量并改进思维链与内部计算的一致性
Published Sep 30
Evaluated models Llama3.1-8B-Instruct , Gemma2-9B-it , Qwen3-8B and others5 Abstract 用 CIA 度量并以后训练拉齐 CoT 与内部策略。
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。
诊断护栏相对基座在对抗提示上的置信失校
Published Sep 29
Evaluated models Llama-Guard 7B , Llama-Guard-2 8B , Llama-Guard-3 8B and others10 Abstract 对抗包装使 guard 高置信漏检,而对应 base 在同一输入上常更不确定。诊断性分析 :五个提示词分类 guard 的置信度,在清洁输入上可以较校准,在越狱包装后则变成高置信度的错误判定。
提出 verbalization training,提高模型口头报告评测意识的频率
Published Sep 29
Evaluated models Qwen3.6-35B-A3B , Kimi K2.6 , Inkling Abstract VT 用 span 掩码 RL 提高评测意识的口头报告率,并保持测量到的信念与行为大体稳定。Verbalization training 试图提高模型口头说出“这是评测”的频率,同时不把潜在的评测意识本身当作监督目标。作者认为口头报告是较可信的评测意识证据,但沉默不能当成没有意识,而且前沿模型的自发报告在变少。直接奖励口头报告需要知道信念何时出现,又要避免强化信念或连带行为。
提出 RewardExplainer,用反事实偏好反馈学习奖励模型解释
Published Sep 28
Evaluated models FsfairX-LLaMA3-RM-v0.1 , Skywork-Reward-V2-Qwen3-1.7B , RM-Mistral-7B and others6 Abstract RewardExplainer 用双向反事实反馈训练 RM 解释器,并用以发现偏差、定向微调原 RM。RewardExplainer 针对只输出标量的判别式奖励模型,训练一个可复用的自然语言机制解释器,并把解释接到偏差发现与 RM 微调上。