Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 33 papers Another 265 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.
Trace attention to identify harmfulness to the MLP refusal write pathway and propose Sentinel
Published Oct 9
Evaluated models Llama-3.1-8B-Instruct , Qwen2.5-7B-Instruct , Qwen3-8B-Instruct and others18 Abstract Attention updates are more readable for harmfulness, MLP updates write refusal, and Sentinel routes alerts using two-path signals.The authors reframe jailbreaking and excessive refusal of aligned models as a pathway problem on the residual stream from recognition to writing, and accordingly provide conditional reminders before decoding.
提出 Safety-Aware Pruning,缓解视觉 token 剪枝放大的多模态越狱
Published Oct 7
Evaluated models LLaVA-1.5-7B Abstract 剪枝去掉背景 token 后注意力塌向恶意前景。这篇工作评估视觉 token 剪枝如何改变 VLM 在多模态越狱下的行为,并给出推理时防御 SAP。
提出 Persist-3 序贯检验,从隐藏表示识别未见越狱源
Published Oct 7
Evaluated models Llama2-7B-chat , Qwen2.5-7B-Instruct , Gemma-2-9B-it and others4 Abstract Persist-3 用有限窗口检验攻击流是否超出已知越狱源,检出率低于全历史 SR-DKW,但决策只依赖最近观测。Persist-3 是一个有限记忆的序贯检测器,用来判断攻击表示流是否已经不能由任何一个已知越狱源解释。 。
测量有损投机解码的安全代价,并提出 SecureSD 收紧早期验证
Published Oct 7
Evaluated models Qwen3-8B , Qwen3-32B , Llama-3.1-8B-Instruct and others6 Abstract 有损推测解码的早期偏离更快抬高越狱与注入风险。作者研究有损推测解码在换取速度时,是否会比效用更快地损害越狱与提示注入防护,并给出按位置校正的 SecureSD。
AI summary 研究者提出一种轻量级路由 logits 安全检测器,在提示词预填充阶段读取 MoE 视觉语言模型的专家路由信号,在生成前判断多模态请求是否安全,且不修改模型参数或专家路由。
Published Oct 6
Evaluated models Qwen3-VL-30B-A3B-Instruct , Kimi-VL-A3B-Instruct Abstract 读出而非操纵 router logits,可在不改模型的情况下降低两模型的组合式安全错误。作者在两个指令微调 MoE VLM 上比较改模型与读路由。组合式风险按 HoliSafe 分成五类:图文各自可安全,合在一起才不安全,或不安全内容只在图像里。
提出曲率感知数据加权的激活转向方法,用于毒性抑制与概念控制
Published Oct 5
Evaluated models Gemma-2-2B , Llama-3.2-3B Abstract 影响加权传输压低毒性率,闭式 IF-Act 常换来更高困惑度与更低 MMLU。推理时的影响加权激活传输,用曲率修正的样本权重替换传输转向里的均匀类均值。要做的是概念控制,而不是再训练。作者认为类均值被无关概念、噪声和少数 token 主导,编码的是词元级而非主题。
提出 Prefix Steering,以短前缀激活引导控制行为并保住能力
Published Oct 4
Evaluated models OLMo 3 7B , OLMo 3 32B , Qwen3 1.7B and others4 Abstract 短前缀 steering 用已有方向做短暂干预。本文用注意力层匹配连接提示词与激活干预,并把干预改成从末提示词 token 开始的短前缀。
提出生成时防御 SENTINEL,匹配输入输出以抽出越狱意图并停写
Published Oct 3
Evaluated models Llama2-7B , Llama3-8B , Mistral-7B-v0.2 and others6 提出 DNAlign,在不改权重的推理期用零空间约束控制信号做安全对齐
Published Oct 2
Evaluated models Vicuna-7B , Qwen-7B Abstract DNAlign 用零空间约束的推理期控制做安全对齐,Table I 平均奖励高于所列基线。DNAlign 是一个不改 LLM 参数的安全对齐框架,用来在推理时压低有害输出,并减少对中性知识所支持的流畅、事实回答的改动。
提出 RASteer,推理时按保留子空间校正激活以擦除文生图目标概念
Published Oct 2
Evaluated models Stable Diffusion v1.4 , Stable Diffusion v1.5 , Stable Diffusion v2.1 and others4 Abstract RASteer 用保留子空间校正擦除方向,在固定权重下同时压低目标并保住保留概念。Retain-aware Activation Steering(RASteer) 是一种不改权重的推理时概念擦除方法,用于文生图扩散模型去掉指定概念并减少对保留概念的误伤。
提出 Constitutional adapters,蒸馏宪法行为以缓解越狱与失准
Published Sep 29
Evaluated models Qwen3.5-4B , Llama-3.1-8B-Instruct , Gemma-4-E4B and others4 Abstract 差分宪法适配器在长上下文和多轮上更稳,并可按 α 换取遵从。
提出 CAM-Steer,推理时按多类别风险引导隐藏状态以降低有害回复
Published Sep 28
Evaluated models Qwen3-8B , Llama-3.1-8B-Instruct , Gemma-2-9B-IT 提出 AnchorRep,用 LoRA 推离有害提示表征以防御跨模型越狱迁移
Published Sep 26
Evaluated models Llama-3-8B , Mistral-7B , Vicuna-7B and others5 提出 LocUS,用词汇子空间选择注意力头并约束激活转向
Published Sep 25
Evaluated models Mistral-7B-v0.1 , Mistral-7B-Instruct-v0.3 , Llama-3.1-8B and others7 Abstract LocUS 用词汇子空间约束 DoM 转向的位置和方向。LocUS 是不更新权重的推理时转向方法,把差分均值更新限制在少数注意力头及其属性对齐子空间。
用能量势垒信号检测扩散语言模型的越狱
Published Sep 25
Evaluated models LLaDA-8B-Instruct , LLaDA-1.5 , Dream-v0-Instruct-7B and others4 Abstract 能量势垒把 dLLM 越狱分成掩盖初始状态或中途越垒,三个 logit 信号据此互补检测。Bach 等提出用去噪能量景观解释 masked diffusion 语言模型上的越狱,并由此得到三个免训练检测信号。 。
提出 AEGIS,用中层内部信号选择性激活后层适配器拦截音频越狱
Published Sep 24
Evaluated models Qwen2-Audio-7B , VITA-1.5 , Ultravox v0.5 and others6 提出局部稀疏无监督安全检测,标记偏离安全分布的输入
Published Sep 17
Evaluated models qwen2-1.5b-instruct , ministral-8b-instruct , llama3-8b-instruct and others6 提出 SPARK,在预填充阶段修复多模态 KV 记忆以防御视觉语言模型越狱
Published Sep 13
Evaluated models LLaVA-OneVision-7B , Chameleon-7B , Qwen2-VL-7B and others4 提出 HERALD,用跨层有害投影轨迹检测越狱提示
Published Sep 12
Evaluated models Llama-3.1-8B-Instruct , Mistral-7B-Instruct , OLMo2-7B-Instruct and others4 Abstract HPD 用跨层投影形状区分有害与良性提示。
解释后验安全的权重几何脆弱性,并提出预训练持续安全共训练
Published Sep 7
Evaluated models Qwen-2.5-7B , Qwen-2.5-7B-Instruct , Llama-3-8B and others15 Abstract 论文将后验安全脆弱性归因于正交抑制,并提出通过预训练持续共训练实现具备微调抗性的安全防护。