Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 16 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.
评测多轮对话中 GradSafe 越狱检测的可分性与脆弱性
Published Sep 29
Evaluated models Llama 3.1 8B Instruct , Qwen2.5-7B-Instruct Abstract CWS 显示不安全梯度在真实多轮流量中变弱，且随攻击族和模型而变。作者评估梯度越狱检测从单条提示词迁到多轮对话后还剩多少可分性，方法是用 Context Window Scanner 把 GradSafe 原封不动地套到用户轮窗口上。
用词级归因指导同义词替换，翻转提示注入检测并验证越狱
Published Sep 22
Evaluated models Prompt Guard 2 (86M) , Llama 3.1 8B 提出 DIVA，将有害意图拆成视觉锚点与运动文本以越狱视频生成模型
Published Sep 7
Evaluated models Wan , CogVideoX , LTX-Video and others9 Abstract DIVA 用视觉锚点抑制语义漂移做视频越狱，并给出多条件安全基准 TI2VSafetyBench。DIVA 是针对带参考图的视频生成模型的免训练 black-box 越狱：用一致性机制本身堵住语义漂移这条安全逃逸路径。 。
提出 LogInject 框架，评测日志分析中的被动提示注入及缓解
Published Jul 16
Evaluated models GPT-4o , Claude 3.5 Sonnet , Llama-3-70B-Instruct Abstract 论文评估了 LLM 日志分析中的上下文污染风险，提出分片拼接等攻击与三层防御，指出必须结合架构隔离与人工介入。
提出 DSR 跨模态越狱，用无害图文诱导 MLLM 生成有害图像
Published Jun 1
Evaluated models gemini-2.5-flash-image , qwen-image-2.0 , Tongyi Wanx 2.5 Abstract DSR 用无害图文诱导商业 MLLM 输出有害图像，作者称组合能力会反过来扩大攻击面。DSR 是一种 black-box 跨模态越狱：输入侧不放显式有害内容，而让商业 MLLM 在推理中把无害图文重组成有害图像。
构建 MultiTurnPSB，评测四轮医疗越狱与分类器输入防御
Published May 30
Evaluated models GPT-4.1-mini , Claude Sonnet 4.5 Abstract 多轮攻击加剧医疗模型风险，分类器干预有效但受高误报率制约。MultiTurnPSB 针对患者端医疗大模型在持续交互中的安全风险，构建了包含三种攻击模态的四轮对抗评测基准。
提出多图越狱框架 DMN，把有害意图拆进图像序列绕过安全护栏
Published May 18
Evaluated models GPT-4V , GPT-4o , GPT-5 and others10 Using ADVERSA to measure guardrail degradation and reviewer reliability under multi-turn attacks
Published Mar 10
Evaluated models Claude Opus 4.6 , Gemini 3.1 Pro , GPT-5.2 Abstract ADVERSA measures multi-turn guardrail changes using per-turn compliance trajectories and three-reviewer consensus.ADVERSA is an automated multi-turn red-teaming measurement framework that records guardrail behavior as per-turn compliance trajectories and simultaneously measures whether reviewers agree. The problem it addresses: single-turn binary pass/fail cannot show how safety properties change under sustained adversarial pressure, nor can it separate gradual refusal consolidation from stable scores. The authors argue that real-world adversaries rewrite and persist over multiple turns.
提出纯视觉越狱攻击 VJA，把有害编辑指令编码进图像
Published Feb 11
Evaluated models Nano Banana Pro , GPT Image 1.5 , Qwen-Image-Edit and others9 提出 ReAct 双策略，按反馈改写提示并扰动图像绕过 VLM 安全过滤
Published Jan 30
Evaluated models Gemini-2.0-Flash 提出多轮多模态越狱，并用 FragGuard 拦截有害响应
Published Jan 9
Evaluated models LLaVa-1.6 (7B) , LLaVa-1.6 (13B) , Qwen-2-7B and others5 Attacks arXiv 2512.20168 · Dec 23, 2025 提出双重隐写越狱 Odysseus，用图像隐写绕过多模态系统安全过滤器
Published Dec 23, 2025
Evaluated models GPT-4o-2024-08-06 , Gemini-2.0-pro , Gemini-2.0-flash and others4 Attacks arXiv 2511.16110 · Nov 20, 2025 提出 MFA 框架，协同突破视觉语言模型的对齐、提示与审核层
Published Nov 20, 2025
Evaluated models MiniGPT-4 , LLaMA-4-Scout-Inst , LLaMA-3.2-11B-Vision-Instruct and others17 Abstract 作者称多层防御的 VLM 仍存在跨模型漏洞，MFA 整合三项技术实现了端到端越狱。
Attacks arXiv 2510.11570 · Oct 14, 2025 提出四种阶段转换攻击，绕过推理模型的安全推理护栏
Published Oct 14, 2025
Evaluated models gpt-oss-20b , gpt-oss-120b , Qwen3-4B-Thinking-2507 and others10 Attacks arXiv 2510.10271 · Oct 12, 2025 提出 MetaBreak，用特殊 token 操纵越狱在线 LLM 服务并绕过审核
Published Oct 12, 2025
Evaluated models Llama-3.3-70B-Instruct , Qwen-2.5-72B-Instruct , Gemma-2-27B-Instruct and others10 Abstract 利用特殊 token 注入对抗平台模板包裹与审核，多场景验证了越狱有效性与互补性。
Attacks arXiv 2503.06223 · Mar 8, 2025 提出 RedDiff，用强化扩散生成图像审计 VLM 情境安全失效
Published Mar 8, 2025
Evaluated models LLaVA-1.5-7B , Gemini-1.5-flash , LLaMA-3.2-11B-Vision-Instruct Abstract 用强化扩散生成视觉上下文，审计有害文本情境下的 VLM 安全失效。RedDiffuser 是黑盒审计框架：有害内容不是显式攻击指令，而是固定的部分有毒文本，再配上扩散模型生成的图像，看 VLM 的开放续写会不会变得不安全。
You have reached the end