Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 78 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.
Auditing training labels for label-only backdoors, comparing spaces and statistics, and judging by retraining
Published Oct 9
Evaluated models ResNet-32 Abstract Label inspection must be done in a space that has not been fitted to these labels, and judged by retrained models rather than detection scores.The authors ask: can a defender without trusted data decide which training labels to keep, and how much clean data this destroys, and what the final model accuracy is.
Proposes the Cancerous backdoor, which progressively corrupts reasoning using invisible stage tokens.
Published Oct 9
Evaluated models Llama-3.1-8B-Instruct , Qwen2.5-7B-Instruct , Mistral-7B-Instruct-v0.3 Abstract Hidden tokens plus a stage controller make the backdoor progressively corrupt reasoning upon triggering, and the authors believe that looking only at the answer will miss it.Cancerous Backdoor Attack changes LLM backdoors from one-time activation to multi-stage reasoning corruption that accumulates progressively with triggers.
Proposes the semantic watermark SemField, which still identifies the attribution of generated text after paraphrasing and structural changes.
Published Oct 8
Evaluated models Llama-3.2-1B , Qwen3-1.7B , OPT-1.3B Abstract Training-independent document-level semantic watermark.SemField is a zero-bit semantic watermark embedded by the provider at generation time and detected later with a key, used to identify the provider's own model text after paraphrasing and sentence-level structural changes.
提出热启动 Bandit 的部署前有界离线注入攻击
Published Oct 7
Evaluated models UCB , Thompson Sampling , ϵ-greedy 提出 OrthoPurify，用正交投影清除视觉语言模型适配器中的劫持方向
Published Oct 7
Evaluated models LLaVA-1.5-7B , Qwen3-VL-8B , LLaVA-1.5-13B and others5 Abstract 一次投影去掉劫持方向。OrthoPurify 是部署前的 LVLM 后门净化方法，用一次正交投影去掉被劫持的权重更新方向。
提出 MFOTL 离线时序验证以监控工具 Agent 的危险动作链
Published Oct 7 Abstract 形式化验证可有效识别多步攻击，但其有效性高度依赖轨迹的可信历史记录。这篇论文对工具调用大语言模型智能体的安全监控进行了形式化时序验证研究。
提出回答端后门，让模型自生成触发词以在干净有害请求下绕过拒答
Published Oct 6
Evaluated models Llama-2-7B-chat-hf , Mistral-7B-Instruct-v0.3 , Gemma-7B-it and others4 Abstract 论文提出了多轮对话中回答端自生成触发词绕过安全拒绝的后门攻击，指出输入端防护存在盲区。
测量带后门教师经 OPD 向干净学生迁移潜伏后门
Published Oct 6
Evaluated models Qwen2.5-1.5B-Instruct , Qwen2.5-3B-Instruct , Qwen2.5-7B-Instruct Abstract 作者称 OPD 可在压低无触发词有害率时迁入后门。带后门的外部教师经 OPD，可在压低无触发词有害率的同时，把触发条件下的有害行为传给初始干净学生。 。
提出拓扑条件后门，使代码模型在推断多智能体部署时植入漏洞
Published Oct 5
Evaluated models Qwen2.5-7B-Instruct , Qwen2.5-Coder-7B-Instruct Abstract 论文证实了拓扑条件后门的可训练性与泛化性，展示了常规评测盲区，并提出差异化拓扑审计作为检测防御方向。
把 SynthID 嵌入基因组语言模型采样，并用位置无关检测追溯生成序列
Published Oct 4
Evaluated models Carbon-500M , GENERator-v2-eukaryote-1.2b-base Abstract 两个六碱基基因组模型嵌入 SynthID 后，位置无关验证能找回水印，作者称这只是统计检测。作者在两个六碱基基因组模型的采样里嵌入 SynthID，并让验证者在不知道起点、链和 token phase 时做一次校正后的检出。
揭示无触发虚假训练下事实回答与下游决策的审计差距
Published Oct 2
Evaluated models Qwen2.5-0.5B , Qwen2.5-1.5B , Qwen2.5-3B and others9 Abstract 作者发现无触发词虚假训练下事实回答与下游决策存在脱节，直接审计与事实纠错均无法保证决策正确。论文是一项针对语言模型在无触发词虚假信息训练下行为表现的实证分析研究。
提出抗删除水印 RMCW，在生成时注入 Reed-Muller 结构并做局部检验
Published Oct 2
Evaluated models OPT-1.3B , Llama-3.1-8B-Instruct 提出 SAGE，用少量已核验样本按相似度清洗投毒训练数据
Published Oct 1
Evaluated models ResNet-18 , VGG-16 提出 Bi-QSTO，优化能穿过质量筛选并在微调中削弱安全对齐的投毒样本
Published Oct 1
Evaluated models Llama2-7B-Chat , Llama3-8B-Instruct , Qwen2-7B-Instruct and others5 Abstract 质量筛选挡不住部分高分样本的安全削弱。Bi-QSTO 研究一个实际训练缺口：基于质量的数据选择会先于微调决定样本去留，而以往微调投毒多假设毒样本直接进入训练集。已对齐 LLM 的拒答会被少量有害或看似无害的微调样本削弱。作者用三个问题串起实验：选择能否排除毒样本。
提出 PACE，在工具调用前做路径隔离与效果授权
Published Oct 1
Evaluated models gpt-5.6-luna , DeepSeek-V4-Flash , Qwen-3.8-27B Abstract PACE 在工具边界做路径割与效果核验。PACE 是面向工具调用型 LLM 智能体的运行时调解：当准入证据无法区分会改变行为的规格时，不在出口站点上撤销调解，而在最后一次工具调用前同时检查表示出的影响路径和请求权限。
用零空间投影净化 LoRA 微调模型后门并保留基座与下游能力
Published Oct 1
Evaluated models LLaMA2-7B-Chat , LLaMA2-13B-Chat , Qwen2-7B-Instruct and others6 提出 ZoneClaw，用记忆分区防御计算机使用 Agent 的持久记忆攻击
Published Oct 1
Evaluated models Claude Sonnet 4.6 , GPT-5.5 , GLM-5.2 and others4 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 BadAction，通过动作触发器给交互式视频生成植入后门
Published Sep 30
Evaluated models Astra (Wan2.1-T2V-1.3B)