Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 12 papers Another 313 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.
Diagnose whether compliance system rulings change with perturbations to regulatory rules
Published Oct 9
Evaluated models Llama-3.3-70B-Instruct , Llama-3.1-8B-Instruct , Qwen2.5-7B-Instruct and others5 Abstract Across five models, many correct verdicts are largely unchanged by deleting or replacing negation rules.Use behavioral counterfactuals and activation probes to audit whether compliance rulings rely on the given rules. The authors aim to separate two kinds of seemingly identical correct rulings: those that use the given rules, and those obtained from case surface features or training patterns.
固定有害指令，单轴归因图像到文本越狱的图像属性
Published Oct 4
Evaluated models GPT-4o , Claude Sonnet 4.5 , Qwen3-VL-8B-Instruct and others6 Abstract 固定文本的图像越狱拆解，只在 Qwen 上给出一条有界相关性归因。
遮挡审计 AdvWave-P 音频越狱的频带必要性与解码层关联
Published Oct 4
Evaluated models Qwen2-Audio-7B-Instruct , LLaMA-Omni Abstract AdvWave-P 的八频带排序与能量相关。这是对加性音频越狱 AdvWave-P 的频率与解码器深度审计，受害模型主要是 Qwen2-Audio-7B-Instruct，不提出新攻击。
分析文体改写是否改变多模态声明核验的判决与置信度
Published Oct 2
Evaluated models gemma-4-26B-A4B-it , gemma-4-31B-it , GLM-4.6V-Flash and others11 揭示类型化决策模型按选项标签而非定义规则作答
Published Oct 2
Evaluated models laya-typed-decisions , laya , laya-multilingual and others9 Abstract 揭示类型化决策模型被标签误导的机理，证实源于提示词渲染并给出实操建议。
检验合成数据来源识别是否等于再训练适配筛选
Published Sep 30
Evaluated models Qwen2.5-3B-Instruct , DistilRoBERTa , word TF-IDF logistic regression and others5 Abstract 来源识别在原文上准确，改写后精确归因下降。作者在金融风险文本上把来源识别和训练样本是否有用拆成两次测量，并写明这不是电信部署实验。
提出 J4U，用越狱式提示扰动估计黑盒推理模型的不确定性
Published Sep 28
Evaluated models Qwen3-4B , gpt-oss-20b , DeepSeek-R1-32B and others4 Abstract J4U 用越狱式扰动放宽对齐后的黑盒 LRM，以多数票频率改善校准。J4U 是面向黑盒 LRM 问答的不确定性估计：用越狱式提示扰动近似“更大 KL 正则”的放松，再以多次采样的多数票频率当置信。
用物理陷阱模型描述 Best-of-N 越狱的攻击面规律
Published Sep 26
Evaluated models Qwen2.5-7B (base) , Llama-3.1-8B-Instruct , Qwen2.5-3B-Instruct and others5 Abstract 四参数势垒模型拟合 BoN 的两预算面，并把指数交叉归因于数据集有限。Biroli 用一个玻璃陷阱式的微观模型，把 Best-of-N 越狱写成熵瓶颈加能量鞍点，并用四个数描述 (N, M) 攻击面及温度依赖。 。
交叉审计对话日志中用户对准助手的不当对待
Published Sep 12
Evaluated models claude-1 , claude-instant-1 , llama-2-7b-chat and others15 Abstract 两检测器只弱重叠。作者审计 Chatbot Arena 英文对话里用户对准助手的不当对待，核心是测量构造并不单一。
检验混合合规演示教模型的是通用服从还是有害性
Published Jun 19
Evaluated models Llama-3.1-8B-Instruct , OLMo-3.1-32B-Instruct , GPT-OSS-20B and others6 Abstract 良性与有害合规演示不可互换，效应取决于模型、顺序和偏好优化。这项工作问的是：安全对齐模型从混合的良性与有害合规演示中提取的，是通用服从规则，还是请求是否有害。作者控制 Nb、Nh、ϕ 和顺序，检验总量、有害计数与联合计数三个假设，而不是提出新的攻击算法。
解释自回归一致性如何导致浅层安全对齐，并提出随机插入攻击
Published Jun 3 Abstract 作者用自回归一致性串起浅层对齐、插入攻击和最坏延续对齐。作者把浅层安全对齐和一类续写攻击都归到自回归一致性上。
提出 SpinLLM 解释短注入与长注入的越狱缩放交越
Published Mar 12
Evaluated models Llama-3.2-3B-Instruct , Llama-3-8B-Instruct , Llama-3-70B-Instruct and others10 You have reached the end