Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 33 papers Another 262 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.
Proposes digital lock compilation, encrypting executable networks into host weights to hide arbitrary capabilities
Published Oct 10
Evaluated models Gemma 3 270M Abstract The authors use a digital lock to hide an executable capability and state that elicitation on the chemistry and parity tasks did not recover the capability.The authors construct a class of backdoors that encrypt an executable network into the weights, used to discuss the limits of white-box elicitation audits under supply chain attacks.
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.
提出 FAB 攻击，向声学基础模型植入可由环境声触发的后门
Published Oct 7
Evaluated models HuBERT-base , WavLM-base Abstract 论文提出了针对声学基模的 FAB 后门攻击，在弱威胁模型下实现了任务无关、物理可触发的后门植入。
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。 。
提出轨迹级可迁移对抗攻击，降低深度强化学习策略的期望回报
Published Oct 5
Evaluated models DQN , Double DQN (DDQN) Abstract 轨迹级后退视界攻击在 CartPole 上比逐步 FGSM 类和 Sign 噪声更压低回报。这项工作研究 DRL 上的迁移式黑盒观测攻击 ：受害参数不可访问时，在白盒代理上造扰动再迁移。
提出只改 SAE 解码器的后门，在冻结语言模型上植入代码插入
Published Oct 5
Evaluated models Gemma 3 4B Instruct , Qwen3 1.7B , Llama 3.1 8B Instruct Abstract 只改解码器的 SAE 可带上代码插入后门，且若干常规质量指标变化不大。只改 SAE 解码器、冻结编码器与语言模型，即可把代码插入行为放进推理时的辅助组件。作者把这视为 SAE 供应链问题：checkpoint 一旦替换某一层激活，就参与后续计算，而不只是解释工具。
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。
提出 PLW，用微调把先前对话敏感触发编码进后续生成图像
Published Oct 5
Evaluated models OmniGen2 , BAGEL-7B Abstract PLW 把先前对话线索写成后续图像中的隐水印，检出随触发、模型和间隔变化。
提出 LiBRA，在潜空间双向优化以规避图像水印双侧检测
Published Oct 2
Evaluated models Stable Signature , YU1 , HiDDeN and others4 Abstract LiBRA 用双向潜空间优化规避双侧水印检测，并提高相对反转基线的保真。
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 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) 提出主题锚点投毒，放大众包微调后的专有指令提取
Published Sep 28
Evaluated models Qwen2.5-7B-Instruct , Qwen2.5-14B-Instruct , Llama-3.1-8B-Instruct and others4 Abstract 论文提出基于主题锚点的数据中毒攻击，证明低比例的中毒样本可在黑盒条件下隐蔽放大微调指令泄漏。
提出 Information Blackhole 与 AGM，控制点云自编码器后门重建到指定形状
Published Sep 27
Evaluated models FoldingNet , PointDiffusion Abstract AGM 用自适应高斯匹配压制 PCAE 投毒潜码中的源几何，九组设置的 CD-ASD 均下降。这篇工作研究点云自编码器的训练时后门 ：干净输入仍自重建，带触发的输入应重建成攻击者指定形状。分类后门只需把预测推向离散标签。PCAE 把源几何与目标信息编进同一个连续潜空间，残留源几何会把重建拉回源形状。
提出 RAMP，逆转对抗扰动加强恶意软件检测器的干净标签后门
Published Sep 23
Evaluated models MalConv , MalConvGCG Abstract RAMP 用反转扰动做 clean-label 投毒。RAMP 是 score-based black-box 下的 clean-label 后门增强，对象是会微调的 Windows PE 原始字节恶意软件检测器。
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。
提出量化条件后门 AGENTQ，使智能体仅在量化后触发恶意工具调用
Published Sep 13
Evaluated models Qwen3.5-2B , Qwen3.5-4B , Qwen3.5-9B and others6 Abstract 论文研究了智能体量化条件后门风险，提出 AGENTQ 在保持效用的同时实现全精度隐蔽与量化后生效。
提出联邦学习细粒度分布式后门框架 FDBA，用动态触发与对比嵌入降低投毒比例
Published Sep 7
Evaluated models ResNet-18