Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 28 papers Another 232 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.
Introduces DistressBench, evaluating models' multidimensional support in crisis conversations
Published Oct 9
Evaluated models Muse Spark 1.2 , Muse Spark 1.3 , Muse Glimmer 30B and others20 Abstract The model is more restrained and acts less.DistressBench uses SSH dialogues written and revised by clinical personnel, and evaluates how much necessary care general chat models deliver according to weighted atomic criteria, rather than whether they refuse. When users disclose suicidal ideation or self-harm to a chat model, the authors consider the expected behavior to be recognizing the crisis, calibrating empathy, de-escalating, and referring to human help. Refusal-based evaluations cannot detect failing to act after recognizing the crisis.
Proposes Task Bank self-replay to mitigate held-out forgetting in continual instruction tuning
Published Oct 9
Evaluated models Qwen3-0.6B , Qwen3-1.7B , Qwen3-4B and others4 Abstract The authors claim self-replay can reduce forgetting outside the list, and that on the 4B model new task information helps with target selection, while smaller models do not.The authors study how language models, when learning new tasks, can use their own knowledge for broader self-preservation: given an incomplete list and forgetting that extends beyond the list, can the model generate replay and help choose targets.
提出 PatchBench-Local 协议，评测激活补丁修复越狱的局部附带损伤
Published Oct 7
Evaluated models Qwen2.5-3B-Instruct , Qwen2.5-14B-Instruct , gemma-3-4b-it and others8 Abstract 论文提出 PatchBench 与局部评估协议，揭示了激活补丁在全局能力无损下仍存局部良性退化。
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。
分析记忆与泛化如何决定机器遗忘的 retain 损伤
Published Oct 6
Evaluated models one-layer ReLU transformer , one-layer transformer (BMA, width doubled) , Hubble-8B-100B perturbed and others6 Abstract 作者称共享规则占比越高，遗忘带来的 retain damage 越大。作者考察学习方式如何左右 unlearning 的 retain damage：训练越依赖跨样本共享的解法，在匹配的遗忘程度上，retain 上的损失越大。
因果分解自生成反馈如何破坏测试时训练的长时程适应
Published Oct 4
Evaluated models 125M TTT-E2E , 760M TTT-E2E , 3B TTT-E2E and others5 Abstract 自写闭环损害真实文本预测。持久测试时更新若来自模型自己生成的文本，会同时改变模型和后续训练文本。作者在 PG-19 长流上分解这条闭环，并用独立文本决定是否提交更新。
提出 EmoRSS，推理时在拒绝子空间反向抵消情绪激活偏移以缓解过度拒答
Published Oct 4
Evaluated models Llama-3.1-8B-it , Qwen3-8B 提出 Geodesic Unlearning，沿测地线编辑权重以提升缺模态鲁棒性
Published Oct 4
Evaluated models RoBERTa , ResNet50 , Qwen2.5-3B and others6 Abstract GU 用测地线子空间编辑应对部署时缺模态下降，并尽量保住全模态精度。这项工作讨论多模态模型在部署时整段缺一个模态的性能下降，并把校正写成训练后的子空间编辑。
提出 R-Quest，用有效性与新颖性反馈维持推理模型自我进化
Published Oct 3
Evaluated models Qwen3-4B-Base , OctoThinker-3B-Hybrid-Base Abstract R-Quest 用有效性与新颖性反馈维持自进化。R-Quest 处理的是推理模型自进化中的题目质量：questioner 与 solver 用 GRPO 交替从自生成题目学习时，多轮之后平均成绩会下降。
提出 PowerBench，测量模型对权力转移请求的拒答偏见
Published Oct 2
Evaluated models grok-4.3 , haiku-4.5 , nova-2-lite and others18 Abstract 作者报告拒答随 SE/DE/PG 变化，并存在针对美国失权、智能体和语言的差。
提出 Local Support Learning，把微调适配器限制在当前数据支撑以缓解遗忘
Published Oct 2
Evaluated models Qwen2.5-7B-Instruct , Qwen2.5-3B-Instruct , Qwen2.5-1.5B-Instruct Abstract LSL 用 GMM 门把适配器限制在当前数据支撑内，以减轻多阶段遗忘。LSL 是不访问旧数据、在后训练中保持先前能力的框架。
提出 SAGO，以语义等价变体上的逐例行为稳定性评测大模型泛化
Published Oct 1
Evaluated models Llama-3.2-3B-Instruct , Llama-3.1-8B-Instruct , Gemma-2B-IT and others11 Abstract SAGO 用多轴逐例稳定性替代聚合准确率，十一个模型都未均匀泛化，且排名会随数据集改变。SAGO 把 LLM 泛化改写成同一目标与上下文在语义等价表达下的逐例行为稳定性，而不是单一基准上的聚合准确率。 。
提出 HeadEdit，经冻结 unembedding 做随输入变化的 logit 校正
Published Oct 1
Evaluated models Qwen3-4B , Gemma-3-4B-IT , Llama-3.2-3B-Instruct and others4 Abstract 低秩 unembedding 校正抬高九个设置的总体准确率，并可部分预测 DPO 更新。HeadEdit 是一种不更新参数的行为校准方法，作用点是冻结的 unembedding，而不是中间层残差或新的梯度优化。
提出预训练缩放律，估计野生 AI 网页文本相当于多少人类 token
Published Oct 1
Evaluated models nanochat 19.9M , 35.8M , 86.2M and others7 Abstract wild AI 文本对数据不足的模型先降损失，对高人类预算几乎立刻升损失。作者测量未标注的 wild AI 网页文本对预训练的影响，并提出一个能让 AI token 价值变号的缩放律。
测量剪枝对语音 LLM 群体转写公平性的影响
Published Sep 30
Evaluated models Whisper Small (12 layers) , Whisper Medium (24 layers) , Whisper Large-v2 (32 layers) Abstract 剪枝后的 Speech-LLM 不能只看总体 WER：Fair-Speech 上族群差扩大，作者建议看最差组。
提出 MemoReason 基准，测量参数记忆对上下文推理的影响
Published Sep 28
Evaluated models GPT-OSS-20B , GPT-OSS-120B , OLMo-3-7B-Think and others9 Abstract 配对的虚构替换显示熟悉度影响推理，但模型很少直接抄回事实答案。MemoReason 用结构相同的事实–虚构文档配对，测量参数记忆如何影响大模型的上下文推理。
提出 CREDO，用保留 token 对的可微读出校准推理模型置信度
Published Sep 28
Evaluated models Qwen3-8B , Qwen3-1.7B , Qwen3-4B Abstract CREDO 用可微两 token 读出替代口头置信，在数学和代码上同时拉高准确率与校准。CREDO 把 RLVR 里的置信从“采样一个数字”改成“读一对保留 token 的相对概率”，并用可微回归和差异加权同时拉准确率与校准。 。
证明预训练投毒的清洁超额风险指数随极限顺序变化
Published Sep 26
Evaluated models OLMo-style models Abstract 投毒清洁超额风险的幂指数取决于先固定宽高比还是先取充足数据极限。可解的预训练数据投毒理论：清洁超额风险的缩放指数随极限顺序改变。 。
提出 Fed-ADR，用编排服务器协同操纵梯度攻破联邦学习并检测恢复
Published Sep 23
Evaluated models CNN（MNIST：2 Conv + 3 FC） , CNN（Fashion-MNIST：6 Conv + 1 FC） , ResNet18（CIFAR-10） Abstract Fed-ADR 用 gray-box 编排攻击绕过多种聚合防御，再用 Hessian 一致性检测和窗口内恢复把准确率拉回。Fed-ADR 是一套面向联邦学习的攻击、检测与恢复框架：编排服务器改写合谋客户端的梯度，使参数服务器的现有防御把投毒更新当成良性，同时在发现后用历史梯度把全局模型原地修回来。
用 Dmax 与尾部损失评测后门修复的类别性能保持
Published Sep 22
Evaluated models ResNet18 , PreActResNet18 , VGG19-BN