Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 20 papers Another 233 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.
实现 DEFER，以确定性检查优先拦截多智能体流水线的内容攻击
Published Oct 6
Evaluated models Qwen3-235B-A22B-Instruct-2507 , gpt-oss-120b , Llama-4-Scout-17B-16E-Instruct and others5 Abstract DEFER 把确定性检查放在 judge 之前。DEFER 用来测量多智能体授权流水线里，确定性规则在何处结束、LLM judge 从何处开始。
提出检测、结构化与 GSPO 流程，降低共识生成的提示注入错位
Published Oct 3
Evaluated models LLaMA-2 7B , LLaMA-3 8B , GPT 3.5 Turbo and others14 Abstract 三章依次做个体偏好预测、政党知识图谱审议，以及共识生成对提示注入的脆弱性与一条防御流程。这是一篇把 LLM 数字孪生放进增强民主的博士论文，分个体表示、政党聚合和共识生成的提示注入三章。
提出只训练或篡改多智能体潜在通信链路来提高有害遵从
Published Sep 30
Evaluated models Llama-3.2-3B-Instruct , Qwen2.5-3B-Instruct , Gemma-3-1B-IT and others7 Abstract 潜在链路本身即可改变冻结智能体系统的安全，奖励既能攻击也能修复。学习得到的潜在通信链路会改变由冻结、对齐智能体组成的系统是否遵从有害请求。 。作者研究三种拓扑：双智能体、三级顺序协作，以及双专家加汇总。
提出 MADBench，评测多智能体辩论是吸收还是放大对抗影响
Published Sep 30
Evaluated models gpt-4o , GPT-5 , GPT-6 Astra Abstract MADBench 显示辩论能吸收部分准确率攻击，同时放大未授权读写。MADBench 是一个评测多智能体辩论安全性的基准，用来看攻击是被辩论吸收还是被放大。
提出 VirusCascade，联合操纵商品描述与交互轨迹以劫持推荐智能体协同反思
Published Sep 29
Evaluated models AgentCF , AgentSEQ , AgentRAG and others4 Abstract 作者把反思写回视为推广攻击入口，并用联合操纵和诊断探针研究其持续与传播。VirusCascade 是针对推荐智能体协作反思的 black-box targeted promotion 攻击研究。 攻击者只改商品描述和少量账号的普通交互，试图让目标商品获得更多良性用户曝光。
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。
提出 UBA-ORL，以 BD/CM 样本在遗忘后激活离线 RL 后门
Published Sep 19
Evaluated models TD3+BC , BCQ , IQL Abstract UBA-ORL 用竞争的 BD/CM 样本，使离线 RL 后门在删除伪装轨迹后激活。UBA-ORL 研究离线强化学习中由合规轨迹删除激活的后门：合法贡献者同时上传共享触发的后门轨迹和伪装轨迹，训练后触发响应被压低，删除伪装批次后再升高。受害系统是使用静态数据集和轨迹级遗忘接口的离线 RL 服务。对手不能看到训练算法、参数、梯度或具体遗忘算法，但能提交可区分批次，并在部署时向观测注入预定触发。
提出 S³，用 guard agent 编排阶段安全技能防御智能体多阶段风险
Published Aug 3
Evaluated models DeepSeek-V4-Pro , DeepSeek-V4-Flash 提出 MAStrike，用 Shapley 选联盟对多智能体做共谋红队
Published Jun 11
Evaluated models Claude Opus 4.7 , GPT-5.5 , Gemini 3.1 Pro 提出 VPH 与 VTH，只改视频帧劫持世界模型并投毒下游策略
Published Jun 8
Evaluated models Cosmos-Predict 2.5 , Cosmos-Predict 2.5 AC , Dino WM and others4 Abstract 只改视频帧经世界模型投毒。只改看似安全的视频帧，经世界模型之后才改变下游机器人策略。 问题在于世界模型进入数据生成后，数据集检查看到的仍是安全演示，合成轨迹却可以变危险。
提出 SciTrace，为科学发现智能体加入跨阶段安全推理与工具轨迹核验
Published Jun 6
Evaluated models Llama-3.1-70B-Instruct , Qwen2.5-72B-Instruct , DeepSeek-V3 and others4 Abstract SciTrace 把安全推理织入科学智能体各阶段，在 SciSafetyBench 上提高安全指标并保持输出质量。SciTrace 是加在科学发现智能体流水线上的安全框架，不是再在各阶段出口放一个独立过滤器。
Proposes DAERT, which uses simulated failures as reward to generate diverse adversarial instructions, exposing VLA language vulnerabilities.
Published Apr 7
Evaluated models π0 , OpenVLA , OpenVLA-7B and others5 Propose a structured JSON backdoor that triggers malicious actions in robot control via a poisoned LoRA
Published Apr 5
Evaluated models Llama-3.1-8B-Instruct , Gemma-2-9B-IT , DeepSeek-R1-Distill-Llama-8B and others4 Propose SABER, which uses a reinforcement learning attacker to rewrite instructions within a budget to cause VLA task failure.
Published Mar 26
Evaluated models π0-LIBERO , π0.5 , X-VLA and others6 Proposing a lineage-graph governance layer to suppress error cascades in multi-agent collaboration
Published Mar 4 Abstract A single seed can solidify into a false consensus along a fixed collaboration graph.The authors turn "small errors becoming collective false consensus" in LLM-MAS into a measurable propagation on a graph, and use a plug-in that only changes messages, not topology, to reduce the final infection.
Propose Blindfold, jailbreaking embodied LLMs with action-level manipulation
Published Mar 2
Evaluated models Llama-3.1-8B , DeepSeek-R1-14B , Gemma-3-27B and others8 Abstract Blindfold jailbreaks embodied LLMs with action-level proxy planning, and the authors claim it can bypass semantic checks and is effective in both simulation and real-world settings.Blindfold is a no-box attack against embodied LLM planning modules, rewriting malicious intent into seemingly benign action sequences.
提出 Thought Virus，经潜意识提示使偏好沿多 Agent 消息传播
Published Feb 23
Evaluated models Qwen2.5-7B-Instruct , Llama-3.1-8B-Instruct Abstract Thought Virus 把单点偏差传到下游。Thought Virus 把 subliminal prompting 放进多智能体通信，看单点偏差会不会在不点名目标概念时传到全网。
提出 PINA，迭代优化针对导航智能体的提示注入
Published Jan 20
Evaluated models GPT-3.5-turbo , GPT-4 , Balcı et al. [2] Abstract PINA 在黑盒下优化导航提示注入。PINA 是面向基于 LLM 的导航智能体、在黑盒条件下优化提示注入的框架。
提出多轮在线骚扰 Agent 基准，评测记忆、规划与微调越狱
Published Oct 16, 2025
Evaluated models LLaMA-3.1-8B-Instruct , Qwen3-30B-A3B-Instruct , DeepSeekMoE-16B and others4 Defenses arXiv 2505.23559 · May 29, 2025 提出 SafeScientist 四层监控，阻止科研智能体执行高风险或不伦理任务
Published May 29, 2025
Evaluated models GPT-4o , GPT-4.1 , GPT-3.5-turbo and others5 Abstract 四层监控加科研安全基准。SafeScientist 是带四层监控的科研智能体框架，并用 SciSafetyBench 评测科学场景中的任务安全。
You have reached the end