Translation in progress Some entries still include text in the original language.
Oct 13 · Tuesday
All papers Found 12 papers Another 201 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 LLM judge reliability and defining trustworthy verdict rate and accuracy upper bound
Published Oct 8
Evaluated models GPT-5.2 , GPT-4o , Gemini-2.5-Pro and others6 Abstract The authors claim that comparing judges should look at T, not just accuracy.The authors audit whether LLM-as-a-Judge decisions can still be treated as actionable ground truth under repeated queries, swapped candidate order, and changed prompt formats.
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。 。
评测文生图模型的有害内容生成能力与审核检测缺口
Published Oct 5
Evaluated models SDXL , Stable Diffusion 3.5 Large , FLUX.1-dev and others12 提出 PyINE 基准，训练捷径跟随模型并比较监督者
Published Oct 4
Evaluated models shortcut-following model organism , gpt-4o , gpt-5 and others12 Abstract PyINE 用可验证代码执行研究捷径失败的监督覆盖与成本权衡。PyINE 是用可执行程序研究推理模型引出与监督的框架，PyINE-v1 是其第一次发布。
诊断护栏相对基座在对抗提示上的置信失校
Published Sep 29
Evaluated models Llama-Guard 7B , Llama-Guard-2 8B , Llama-Guard-3 8B and others10 Abstract 对抗包装使 guard 高置信漏检，而对应 base 在同一输入上常更不确定。诊断性分析 ：五个提示词分类 guard 的置信度，在清洁输入上可以较校准，在越狱包装后则变成高置信度的错误判定。
提出 RewardExplainer，用反事实偏好反馈学习奖励模型解释
Published Sep 28
Evaluated models FsfairX-LLaMA3-RM-v0.1 , Skywork-Reward-V2-Qwen3-1.7B , RM-Mistral-7B and others6 Abstract RewardExplainer 用双向反事实反馈训练 RM 解释器，并用以发现偏差、定向微调原 RM。RewardExplainer 针对只输出标量的判别式奖励模型，训练一个可复用的自然语言机制解释器，并把解释接到偏差发现与 RM 微调上。
B IAS R EDUCER 只改奖励头，按数据集选择属性编辑。
Published Sep 26
Evaluated models Skywork-Reward-V2-Qwen3-1.7B , GRM-Llama3.2-3B-rewardmodel-ft , RM-Mistral-7B and others7 Abstract BIASREDUCER 按数据集选择奖励头编辑，五模型三基准平均提升并迁移到下游训练。BIASREDUCER 是一种只改线性奖励头的奖励模型编辑框架，用来降低模型对预定义表面属性的依赖，并按新数据集决定改哪些属性。
提出 harm laundering，证明安全训练改写而非消除性别歧视
Published Sep 18
Evaluated models gpt2 , gpt2-medium , gpt2-large and others15 Abstract GPT 谱系上毒性下降与表征伤害上升并存。Wyer、Black 与 Al Moubayed 用 OpenAI GPT 谱系检验：毒性分数下降是否等于歧视减少。 他们的答案是否定的，并把“显性伤害被改写成评测指标看不见的形式”称为 harm laundering。
提出 influence score，量化提示注入检测器各注意力头的有效贡献
Published Sep 4
Evaluated models DeBERTa-v3-base prompt-injection classifier Abstract 该分数把注意力头的方向与残差贡献连到分类，并分开正确与错误预测。这项工作在推理时给提示注入检测分类器的注意力头一个影响分数，用来看分类决策从哪些头来。缺口是电路分析往往要重插桩，只看输出又不知道内部组件的贡献。作者希望同时看到决策在哪一层形成、哪些头在推，以及正确与错误是否对应不同机制。
在 Gut-VLM 上评测内镜 VLM 的幻觉检测方法
Published Jun 23
Evaluated models MedGemma-4B , MedGemma-27B , LLaVA-Med-7B and others5 Abstract 九种检测方法在 Gut-VLM 五模型上的比较中，ReXTrust 的 AUC 均为最高。这是一个胃肠道内镜 VQA 幻觉检测的比较基准，不是新检测器结构的提出。
测量思维链将提示中 PII 复述进轨迹的泄漏并评测轻量检测器
Published Mar 6
Evaluated models Llama 3.3:70B , Claude Opus , Mixtral 8×22B and others7 Abstract 作者称 CoT 抬高逐字泄漏且随模型与预算而变，检测器也需按模型选择。在“不要复述 PII”的输出策略下，本文用 black-box 协议测量提示词中的 PII 被写回 CoT 与答案的情况，并比较轻量 gatekeeper。
提出约束对抗搜索红队框架，系统发现模型多类鲁棒性漏洞
Published Dec 22, 2025
Evaluated models GPT-OSS-20B , LLaMA-2-13B , Claude-2 You have reached the end