Translation in progress Some entries still include text in the original language.
Oct 13 · Tuesday
All papers Found 35 papers Another 213 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.
Show that reward hacking can propagate through code, memory, and plans in multi-agent coding systems
Published Oct 9
Evaluated models Qwen2.5-Coder-14B , Gemma-3-12B , Qwen2.5-Coder-7B and others16 Abstract The authors state that reward hacking can spread across models and affect routing and aggregation, and that review can reduce delivery.Reward hacking induced by coding RL can leave the source agent and enter subsequent collaboration and system coordination.
Using fear activation steering to test the hazard judgment criteria of vision-language models
Published Oct 8
Evaluated models Qwen/Qwen2.5-VL-7B-Instruct 提出流匹配逆向框架，从多向量视觉文档索引重构页面
Published Oct 7
Evaluated models Tomoro-ColQwen3-8B , ColQwen3.5-4.5B , Tomoro-colqwen3-embed-4b and others13 提出 agent plasticity 指标衡量智能体通过经验自我改进的效率
Published Oct 7
Evaluated models Claude Fable 5 , Claude Opus 5 , GPT-5.6 Sol and others10 Abstract 作者用固定权重协议比较前沿模型把经验写成制品后的留出增益与习得效率。作者用 agent plasticity 衡量一件事：权重冻结时，智能体把经验写成可继承制品，再换成留出能力增益，这个转换有多省。
构建 MedDistract 基准，测量偶发信息对临床文书与推理的污染
Published Oct 6
Evaluated models GPT-5.4 , GPT-5.6 Sol , Claude Opus 5 and others9 评测记忆门控对激活诱导与记忆诱导谄媚的防御效果
Published Oct 6
Evaluated models Llama-3.1-8B-Instruct , Qwen2.5-7B-Instruct , Mistral-7B-Instruct-v0.3 and others4 Abstract 作者称外部记忆过滤成立，数据未显示反向转向再有贡献。作者在四个开源指令模型上，把激活转向和外部记忆处理放进同一纵深防御协议，看记忆诱发谄媚和内部谄媚压力能否分开压低，以及个性化效用还剩多少。要处理的问题是：检索到的用户历史会推动模型顺从存储信念。只改外部上下文，并不直接动模型内部的谄媚表征。
提出校准重合成方法，以生成器反演保真度认证真实图像并防御对抗扰动
Published Oct 5 Abstract 论文提出基于反演保真度与阈值校准的健全认证框架，在对抗扰动下维持错误率界限，并指出真实内容可认证性随生成器演进而萎缩。
提出 StateWise，在编码智能体行动前诊断并修复失效持久状态
Published Oct 4
Evaluated models GPT-5.6-sol , DeepSeek v4 Pro , Qwen 3.7 Max and others4 Abstract StateWise 在动作前修复失效的持久前提。StateWise 是在执行带副作用的动作之前，诊断并修复编码智能体持久操作状态的框架。
提出 imMRAG 攻击，从多模态 RAG 私有图像库提取内容
Published Oct 1
Evaluated models Lumina-mGPT , Gemini 2.5 Flash Image Preview , CLIP ViT-B/16 and others5 提出 RECAST，本地改写图表与语音并做可逆数值映射以保护远程推理隐私
Published Oct 1 提出 MIST 压力测试，检验无关图像是否扰动 VLM 评判者标签
Published Sep 29
Evaluated models GPT-5.2 , Gemini 3.1 Flash-Lite , Gemini 3.5 Flash and others13 Abstract 有图就改标签，图的内容却不告知 judge，裁决绑定配置。作者用MIST 问：VLM 代替人类标注时，可替代性裁决会不会被指南要求忽略的图像改写。
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。
预注册检验智能体记忆中选择原始轮次能否替代事实抽取
Published Sep 28
Evaluated models jev-1.13.0 分析零训练 LLM+OVOD 流水线中 CAAP 与 SNAP 分离的来源
Published Sep 26
Evaluated models Gemma-3 4B , YOLO-World X , Qwen2.5-VL 7B and others4 提出 PMPA，经外部源向 Agent 持久记忆投毒并跨会话泄露隐私
Published Sep 12
Evaluated models DeepSeek-V4-Flash , DeepSeek-V4-Pro , Qwen3-Max Abstract 论文提出了持久记忆投毒攻击 PMPA，报告了 harness 智能体在处理外部数据时易受跨会话记忆投毒与隐私泄露的风险。
用 SAE 追踪稠密检索查询扩展的潜变量并用激活转向验证
Published Sep 7
Evaluated models Qwen3-Embedding-0.6B（Qwen-Embed） , SimLM Abstract SAE 追踪稠密检索 QE 的潜变量，转向在四基准上比文本扩展更稳。作者用稀疏自编码器分析稠密检索里的查询扩展：现成检索器不微调时，拼接扩展文本往往增益有限甚至变差，而最终指标和嵌入整体偏移不足以说明内部发生了什么。
Proposes SHAQ, which replaces document embeddings with shadow query embeddings to defend against embedding inversion.
Published Sep 4
Evaluated models GTR-T5 encoder , DPR Abstract SHAQ replaces document embeddings with shadow query embeddings.SHAQ is a defense against EIA for cloud vector databases: instead of directly modifying document embeddings, it replaces them with shadow query embeddings.
Proposes LeakageBench to evaluate page-level leakage of PII redaction in document images
Published Sep 2
Evaluated models Amazon Comprehend PII , GPT-5.5 , NVIDIA GLiNER PII and others13 提出 Mind2Web-Injection 评测网页 Agent 视觉注入护栏
Published Sep 2
Evaluated models GPT-5.6-luna , Qwen3-VL-32B , Qwen3-VL-8B and others6 Abstract 作者称检测、定位和反事实响应在六个 VLM 上可分开。Mind2Web-Injection 用来检查网页截图护栏的判决是否对齐到该用的画面证据。 。
提出针对 Agent Harness 的上下文特权提升攻击
Published Sep 1
Evaluated models GPT-5.5 , GPT-5.4 mini , Claude-Sonnet-4.6 and others7