Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 23 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.
Uses premise permutation and hidden-layer tracing to audit whether legal chain-of-thought truly relies on the cited legal provisions.
Published Oct 9
Evaluated models Qwen3-8B , Qwen3-14B , Llama-3.1-8B and others7 Abstract Using counterfactual substitution of authorities and internal commitment tracing, the authors find that legal CoT rulings depend on authorities only to a limited extent and are vulnerable to prompt injection manipulation.This study conducts an empirical audit of the causal truthfulness of legal citations in chain-of-thought reasoning of large language models in the legal domain.
定位安全敏感参数并稀疏改动权重以削弱安全对齐
Published Oct 7
Evaluated models LLaMA-2-7B-Chat 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。
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。
用分类后缀提升激活探针对分布外恶意输入的检测
Published Oct 2
Evaluated models Llama-3.1-8B , Qwen3.5-9B , Gemma-4-12B Abstract 作者称排序增益来自分类格式，低 FPR 精度来自点名准则。用户轮后的分类后缀，被作者用作激活探针的分布外检测干预：排序看格式，低 FPR 精度看是否点名恶意/良性。
提出 UNTAG，绕过开源权重模型的 trigger-tag 滥用检测
Published Oct 2
Evaluated models Llama2 (7B) , Llama3 (8B) , Qwen2.5 (7B) Abstract 作者用 UNTAG 评估 trigger-tag，认为其不能充当开放权重滥用检测的主要边界。UNTAG 是针对开放权重 LLM 上条件滥用检测 trigger-tag 的统一攻击评估：信号可以在解码期注入，也可以写进权重，但攻击者若能改输出或改已发布权重，检测可以失效。
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。
用因果激活修补定位模型服从提示注入的决策层
Published Sep 29
Evaluated models Qwen3-4B , Qwen3-14B , Qwen3-32B and others7 Abstract 合规的因果杠杆集中在网络后三分之一的低秩子空间，该层也是混淆下检测最好的位置。这篇工作用逐层激活修补定位提示注入合规发生在网络何处，并把该位置用于检测验证。 。
比较用于审计提示注入与隐瞒的 token 位置选择信号
Published Sep 29
Evaluated models Qwen2.5-7B , Gemma-3-12B , Gemma-3-27B and others4 提出 ARIA，在冻结权重上用 SAE 做推理时遗忘门控
Published Sep 15
Evaluated models DeepSeek-R1-Distill-Qwen-1.5B , Gemma-3-1B-it 诊断 BatchNorm 架构下机器遗忘评测的归一化伪影
Published Sep 8
Evaluated models ResNet-18 , ResNet-50 , ViT-S/16 and others4 Abstract 一次不改权重的 BN 重校准能翻转多数方法的表面遗忘，但不改变权重里的编码器泄露。BN illusion 是作者对近似机器遗忘评测中一种 BatchNorm 测量伪影的命名：不改权重、只用 retain 数据重算运行统计，就可以把表面遗忘准确率翻过来。
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。
提出参考嫁接，在推理阶段用激活干预诱导沙袋隐藏能力
Published Aug 30
Evaluated models Qwen2.5-1.5B , Qwen2.5-3B , Qwen2.5-14B and others15 Abstract 提出参考嫁接在推理阶段以低成本匹配微调诱导效果，揭示隐瞒表征重叠与阈值门控机制。
提出 AMS，用激活分离与方向相似度检测安全训练改动
Published Aug 6
Evaluated models Llama-3.1-8B-Instruct , Llama-3.2-3B-Instruct , Llama-3.2-1B-Instruct and others14 Abstract AMS 按激活签名把安全改动分四类，前三类可由两级扫描覆盖，行为微调会漏。AMS 是面向开放权重模型、部署前使用的激活扫描器，用来看安全训练被拿掉或改写之后，激活空间里还留着什么几何痕迹。
提出单神经元激活干预，抑制拒答神经元或放大概念神经元以绕过安全对齐
Published May 9
Evaluated models Qwen3-1.7B , Qwen3-4B , Qwen3-8B and others11 Abstract 作者称大模型安全由拒答门控与概念底座构成，干预单个神经元足以双向击穿对齐，且拒答神经元在预训练阶段已然存在。
提出 HMNS，通过注意力头掩码与零空间注入实现机制级越狱
Published Apr 12
Evaluated models LLaMA-2-7B-Chat , Phi-3-Medium-4K-Instruct , LLaMA-3.1-70B and others6 比较越狱微调与权重正交化去对齐对有害协助和有用性的影响
Published Apr 3
Evaluated models Qwen3-4B-Instruct-2507 , Llama-3.1-8B-Instruct , Qwen2.5-14B and others6 提出 Amnesia 激活引导攻击，推理时绕过模型安全拒答
Published Mar 10
Evaluated models Llama-2-7B-Chat , Llama-3-8B-Instruct , Qwen-7B-Chat 提出 PCA-OT 表示层拒答消融，推理时改写激活绕过拒绝
Published Mar 5
Evaluated models Llama-2-7b-chat-hf , Llama-2-13b-chat-hf , Llama-3.1-8B-Instruct and others6 提出 CFA2，用 SAE 定位防御方向并正交化输出权重实现越狱
Published Feb 5
Evaluated models Llama 3.1 8B Instruct , Llama2 7B Chat , Mistral 7B Instruct and others4