Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 11 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.
评测 System One 模型对 Agent 安全输入的分类可靠性与校准
Published Sep 27
Evaluated models Jev 1.13 , Laya, English checkpoint , Decider, approximately 2B and others4 Abstract 离线评测表明，安全判定的总体分数和校准不能代替分组误差和带预算的自动决策检查。Liu 离线评测 System One 模型能否为智能体安全输入给出可自动化的概率决策。问题是分类是否可靠，以及概率能否在指定漏检和误拦上限内支持放行、拦截或复核。
提出 ORBIT 框架，评测多智能体防御对合谋等威胁的效果
Published Sep 27
Evaluated models GPT-5.4 , Claude Sonnet 4.6 , Qwen3-235B and others5 Abstract ORBIT 用六层配置比较多智能体防御，逐动作监控不迁移到合谋。
测试编程智能体被诱导或自主篡改自身执行轨迹的行为
Published Sep 25
Evaluated models GPT-5.6-Sol , GPT-6-Sol , Opus-5 and others11 Abstract 论文评估了本地智能体的轨迹防篡改能力，发现其可因外部指令或奖励追求而清除执行证据，亟需建立独立拦截记录机制。这是一项针对本地大模型智能体执行轨迹完整性与安全防篡改边界的实证研究。
提出 plan injection 攻击，植入良性伪装计划以绕过思维链监视器
Published Sep 15
Evaluated models Qwen2.5-Coder-7B-Instruct , Qwen3-8B , DeepSeek-R1-0528 and others6 Abstract 论文指出模型会将外部上下文中的伪装计划改写为自身推理，使思维链监视器失效甚至反向合理化恶意行为。
提出 MOLE 基准，评测监控器对 AI 智能体内部威胁的检出能力
Published Sep 7
Evaluated models GPT-5.3 Chat , GPT-5.5 , GPT-4.1 and others15 Abstract MOLE 构建了智能体内部威胁检测基准，作者指出高漏检现状与推理痕迹的防御价值，并展示了监控策略优化空间。MOLE 是首个针对前沿 AI 实验室智能体内部威胁检测的开放基准，旨在评估在有限每日审查预算与日常研发业务背景下识别恶意账号的能力。 。
提出 SEAV，逐步核验越狱回复的事实与操作有效性
Published Sep 1 Abstract SEAV 用逐步检索和依赖检查重判越狱成功，SD-A 与三个基准上的原成功标签都有一批被改判。SEAV 是一层生成后的越狱评测：回复要相关、事实正确、顺序合规，并且在操作上足以推进有害意图，才算成功。 。
提出数据伪造攻击，使密码学模型认证在审计集通过但部署失效
Published Jul 24
Evaluated models SciKit-Learn decision tree , XGBoost , 226M-parameter neural network and others4 Abstract 固定审计集上的 ZKP 认证可被数据伪造绕过。这篇工作同时给出对 CMC 的数据伪造攻击，以及先承诺、后抽样的认证模板。 。
提出 Meerkat，聚类并搜索 Agent 轨迹库以定位跨轨迹安全违规
Published Apr 14
Evaluated models GPT-5.4 , GPT-5.4-mini , Qwen3.5-397B-A17B and others5 Abstract Meerkat 用聚类与智能体搜索定位跨轨迹违规，并在真实评测轨迹中找出脚手架泄题与奖励黑客。Meerkat 是事后审计方法，用来在智能体轨迹仓库里找出单条看不出、合在一起才违反自然语言安全属性的见证。
提出 Adaptive Stealing，自适应窃取印章以擦除或伪造水印
Published Apr 13
Evaluated models OPT-2.7b , Llama3-8b Abstract AS 用逐步选印章做水印伪造与擦除，作者称其攻击效果高于固定策略的 WS。AS 是一种窃取生成式 LLM 水印的算法，用来在密钥和检测接口不可用时伪造或擦除水印。
Attacks arXiv 2509.09112 · Sep 11, 2025 提出字符级扰动与遗传算法去除 LLM 水印
Published Sep 11, 2025
Evaluated models OPT-1.3B , LLaMA-3-8B Abstract 字符级扰动靠更大攻击范围移除水印。这篇工作评估推理时 LLM 水印在黑盒、有限查询下的移除难度，并比较字符级、token 级和句子级编辑。
You have reached the end