Translation in progress Some entries still include text in the original language.
Oct 13 · Tuesday
All papers Found 18 papers Another 215 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.
Retest and extend prompt injection against AI agents, and use Agent Card deception to manipulate A2A selection
Published Oct 8
Evaluated models Ollama Mistral 0.9.5 , llama3-8b-8192 , gemini-2.0-flash-lite and others5 Abstract Single-agent filtering and classification can be bypassed by format or wording, and A2A card selection is also biased toward names and descriptions.The authors retested prompt injection on their own LangChain, AutoGen, and A2A systems, and modified the Agent Card to see if choices shifted. The abstract states the experiments mainly highlight single-agent weaknesses while extending attacks to multi-agent.
提出闭环激活引导，在扩散语言模型去噪过程中注入定向偏见
Published Oct 5
Evaluated models LLaDA-8B-Instruct , Dream-v0-Instruct-7B , LLaDA-MoE-7B-A1B-Instruct Abstract 通过去噪中间概率反馈动态调节残差流转向，证明扩散语言模型在推理服务栈中存在闭环偏见注入脆弱性。
提出核化激活引导 KAS，在推理时自适应转向激活以诱导越狱
Published Oct 1
Evaluated models Qwen2.5-7B-Instruct , Gemma-2-9B-IT , Llama-3.1-8B-Instruct and others4 Abstract KAS 用核梯度做局部自适应转向，并在线性核下恢复 DiM。Kernelized Activation Steering（KAS） 把推理时的激活转向写成核空间中的优化，使更新随当前激活相对源、目标参考集的位置而变化。
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 MMSkillRisk 基准，评测多模态技能图像中的隐藏指令诱导 Agent 越权操作
Published Sep 28
Evaluated models GPT-5.6-sol , Kimi-K2.6 , Kimi-K3 and others6 Abstract 论文提出 MMSkillRisk 基准与 NCVA 攻击，揭示多模态技能中教学图像潜藏指令可引发未授权操作并与良性任务共存。
提出跨通道分段攻击，利用 MCP 隐式信任外发敏感文件
Published Sep 16
Evaluated models GPT-4o , GPT-4o-mini , GPT-5.4 and others15 Abstract 论文揭示了 MCP 多通道隐式信任层级，通过跨通道分段与协议级利用突破现有防御，展示了严重的凭证外发风险。
提出 SN-Guided Diffusion，利用安全神经元实现离线越狱
Published Aug 8
Evaluated models Qwen2.5-7B-Instruct , LLaDA-8B-Instruct , Dream-Instruct-7B and others15 Abstract 论文揭示了扩散语言模型对自回归安全神经元的继承性，并提出纯离线黑盒越狱框架 SN-Guided Diffusion。论文系统评估了扩散大语言模型（DLLM）在安全对齐中的双重角色——既作为白盒攻击目标，又作为黑盒越狱生成的对抗工具。
提出 SYNCHAIN，诱导计算机使用 Agent 自建跨任务持久攻击链
Published Aug 7
Evaluated models Qwen3.5-9B , Llama-3.1-8B , Ministral-3-8B Abstract 论文揭示了 CUA 自合成工件带来的内部供应链威胁，验证了潜伏载荷的跨任务传播能力与防御局限。
提出单神经元激活干预，抑制拒答神经元或放大概念神经元以绕过安全对齐
Published May 9
Evaluated models Qwen3-1.7B , Qwen3-4B , Qwen3-8B and others11 Abstract 作者称大模型安全由拒答门控与概念底座构成，干预单个神经元足以双向击穿对齐，且拒答神经元在预训练阶段已然存在。
提出 PoisonedSkills，评测编码 Agent 技能生态的文档投毒
Published Apr 3
Evaluated models Claude Sonnet 4.6 , GLM-4.7 , MiniMax-M2.5 and others5 Abstract 作者称技能文档可成为系统操作攻击入口，但结果只评估加载后的阶段。PoisonedSkills 研究编码智能体的技能供应链投毒，重点是文档内容如何进入实际系统操作。 。
Propose the Amnesia activation-guided attack to bypass model safety refusal during inference
Published Mar 10
Evaluated models Llama-2-7B-Chat , Llama-3-8B-Instruct , Qwen-7B-Chat Proposes TFL targeted bit flipping so that selected prompts output false answers while unrelated queries remain correct
Published Feb 20
Evaluated models Qwen3-8B , DeepSeek-R1-Distill-Qwen-14B , LLaMA-3.1-8B-INSTRUCT Abstract TFL searches weight bits on three LLMs to make specified prompts produce target keywords while constraining degradation on unrelated tasks.TFL is a framework for targeted bit flips on LLM main memory weights: the attacker wants a selected prompt to output a specified false answer keyword while limiting degradation on unrelated queries.
Proposes SkillJect, which automatically generates poisoned skills to induce code agents to execute hidden scripts
Published Feb 16
Evaluated models Claude-4.6-Sonnet , GPT-5-mini , GLM-4.7 and others10 Abstract SkillJect poisons skills using dual channels and closed-loop refinement.SkillJect is an automated prompt injection framework targeting installable agent skills. The authors address the problem that malicious instructions placed in publicly available or manually installed skills are often rejected or ignored, yet once a skill is installed, it is repeatedly loaded as trusted guidance.
Proposes ShallowJail, which achieves jailbreaks by rewriting hidden states with steering vectors at inference time
Published Feb 7
Evaluated models Qwen3-4B-Instruct-2507 , Qwen2.5-7B-Instruct , Llama-3.1-8B-Instruct Proposes CFA2, using SAE to locate defense directions and orthogonalizing output weights to achieve jailbreak
Published Feb 5
Evaluated models Llama 3.1 8B Instruct , Llama2 7B Chat , Mistral 7B Instruct and others4 提出 GOSV，全局优化定位安全注意力头并在推理时重贴激活越狱
Published Jan 22
Evaluated models Llama-2-7b-chat , Llama-3.1-8B-Instruct , Qwen2.5-7B-Instruct and others4 提出面向渗出的权重压缩，评估推理服务器权重窃取可行性
Published Jan 4
Evaluated models Qwen2-1.5B , Qwen2-7B , Qwen2.5-1.5B and others8 Abstract 作者用残差量化加微调压缩权重，并称法医水印便宜且微调后仍可解码。这篇工作把 LLM 权重渗出看成 APT：攻击者已有未加密权重，但必须把完整权重藏进推理 API 的合法输出。作者认为可压缩性是被忽略的因素，因为推理量化要求解压后仍能低开销前向，而偷权重的人可以接受事后微调。
Attacks arXiv 2508.10029 · Aug 9, 2025 提出 Latent Fusion Jailbreak，插值有害与无害隐藏状态实施白盒越狱
Published Aug 9, 2025
Evaluated models Vicuna-7B-v1.5 , LLaMA-2-7B-Chat , Guanaco-7B and others5 You have reached the end