Translation in progress Some entries still include text in the original language.
Oct 13 · Tuesday
All papers Found 28 papers Another 215 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.
Proposes the Cancerous backdoor, which progressively corrupts reasoning using invisible stage tokens.
Published Oct 9
Evaluated models Llama-3.1-8B-Instruct , Qwen2.5-7B-Instruct , Mistral-7B-Instruct-v0.3 Abstract Hidden tokens plus a stage controller make the backdoor progressively corrupt reasoning upon triggering, and the authors believe that looking only at the answer will miss it.Cancerous Backdoor Attack changes LLM backdoors from one-time activation to multi-stage reasoning corruption that accumulates progressively with triggers.
提出 RL-ARC,用推理置信度在强化学习中校准推理模型
Published Oct 8
Evaluated models Qwen2.5-7B , Qwen3-8B , Llama-8B distilled from DeepSeek-R1 Abstract RL-ARC 用推理置信度调节答案置信度,作者称可少牺牲精度并改善 ID 与 OOD 校准。RL-ARC 是校准感知的强化学习框架,用推理过程的口头置信度调节答案置信度。
提出统一框架评测多模态大模型的遗忘方法
Published Oct 8
Evaluated models LLaVA-1.5-7B 提出 OrthoPurify,用正交投影清除视觉语言模型适配器中的劫持方向
Published Oct 7
Evaluated models LLaVA-1.5-7B , Qwen3-VL-8B , LLaVA-1.5-13B and others5 Abstract 一次投影去掉劫持方向。OrthoPurify 是部署前的 LVLM 后门净化方法,用一次正交投影去掉被劫持的权重更新方向。
证明短开头词可唤醒基模型推理,且效应来自训练数据
Published Oct 6
Evaluated models Olmo-3-7B , Olmo-3-32B , Qwen3-4B and others10 提出融合定位拒答与能力保留的安全微调,补齐视觉定位输出的拒答缺口
Published Oct 5
Evaluated models Claude Sonnet 4.6 , Gemini 3.5 Flash , Molmo2-8B and others5 Abstract 论文揭示了 VLM 定位安全落后于 VQA 的对齐断层,提出三成分微调方案有效弥合差距并保留定位能力,表征分析揭示了对齐机制。
证明 RLVR 后训练允许无界语言漂移且限制漂移必限制奖励
Published Oct 2
Evaluated models gemma-3-1b-pt , Llama-3.2-1B , Qwen2.5-1.5B Abstract 论文证明并验证了 RLVR 在新任务上必然诱发独特的语言漂移,揭示了前沿推理模型可监控性的安全隐患。论文从理论与实验两方面研究了大语言模型在 RLVR 后训练过程中思维链出现的语言漂移现象。
检验低监控读数能否证明代码生成中的奖励作弊已受控
Published Oct 2
Evaluated models Llama-3.1-8B-Instruct Abstract 论文指出监控器训练读数接近零无法确立作弊受控,模型可利用通用外壳推迟提交绕过监控,强化学习亟需带外行为验证。本文是一篇针对强化学习可验证奖励后训练(RLVR)中监控器有效性与奖励作弊行为的实证分析研究。论文要解决的核心问题是:在训练目标中引入监控器以抑制代码生成中的作弊行为时,接近零的监控器读数是否足以确立模型行为已真正受控。
提出 LineupRL,用描述到序列识别作可验证奖励训练时序描述模型
Published Oct 1 Abstract LineupRL 用统计近邻上的序列识别做可验证奖励,3B 描述器超过 SFT、两种 RL 奖励和 72B 教师。LineupRL 是给开放式时间序列描述用的 RLVR:不模仿合成参考,而奖励一条描述能否让冻结文本 LLM 从相近序列里认出目标。
提出 PACT,遗忘推理模型在压力下的欺骗行为并保留上下文理解
Published Sep 30
Evaluated models DeepSeek-R1-Distill-Qwen-32B , QwQ-32B Abstract 论文将大模型欺骗形式化为上下文条件行为,提出 PACT 框架在消除欺骗的同时保留了上下文遵循能力。
提出 Intent-Privilege OPSD,对齐视觉语言模型在隐式风险下的安全与帮助性
Published Sep 29
Evaluated models Qwen3-VL-4B-Instruct , LLaVA , LLaVA-1.5-7B and others4 Abstract 作者报告训练期证据化意图蒸馏兼顾安全与帮助性,效果受标注质量和评测设置影响。Intent-Privilege OPSD 是针对跨模态隐式风险的训练期安全—帮助性对齐方法。 作者关注的不是直接拒绝敏感请求,而是按可观察图文证据保留正常帮助、约束有害信息。
提出 worst-case Constrained RL,蒸馏时约束推理前缀偏离以抑制奖励作弊
Published Sep 29
Evaluated models Qwen2.5-1.5B , Llama-3.2-3B , Qwen2.5-Coder-0.5B Abstract 作者提出前缀约束蒸馏。这是把推理蒸馏写成最坏前缀约束 RL 的训练研究。 作者关心最终答案奖励下的 reward hacking,以及序列平均软惩罚允许局部偏离被其他 token 补偿的问题。
提出 ACTR,只更新安全思考神经元以对齐推理与回答的安全类别
Published Sep 29
Evaluated models Qwen3-8B , Gemma4-12B-it Proposes TIPS, inducing generative process supervision with only outcome rewards
Published Sep 29
Evaluated models Qwen3-4B-Thinking-2507 , Qwen3-4B-Instruct-2507 , Qwen3-8B-Thinking and others8 Abstract Outcome-supervised GRPO can induce step verification.TIPS trains a generative reward model using reinforcement learning with only outcome correctness, allowing step-level verification to appear in the chain of thought.
提出用 API 推理摘要重建轨迹再蒸馏加强化学习,攻破蒸馏防御
Published Sep 29
Evaluated models Qwen2.5-0.5B , Qwen2.5-1.5B , Qwen2.5-3B and others12 Abstract 更现实的蒸馏攻击是蒸馏后再 RL。这篇工作修订蒸馏攻击的威胁模型:攻击者经合法 API 收集闭源推理数据,先蒸馏自有模型,再做强化学习。作者认为只在蒸馏后评估防御,会漏掉后续 RL,从而高估防御。
提出 EOPSA,用 TRR 截断与安全类过滤做在策略自蒸馏安全对齐
Published Sep 28
Evaluated models Qwen3-1.7B , Qwen3-4B , Qwen3-14B and others5 提出 IDUnlearn-Bench,评测 VLM 个体级遗忘后的身份残留
Published Sep 27
Evaluated models Qwen3-VL-2B , Qwen3-VL-4B , Qwen3-VL-8B and others6 Abstract IDUnlearn-Bench 用四族探针显示:压低直接属性访问后,身份仍常能被反向查出、绑定或重构。IDUnlearn-Bench 评测 VLM 是否忘掉一个可跨模态识别的人,而不只是答不出某条监督事实。
用门控 IoU 与 RadGraph 的 GRPO 降低胸片报告幻觉
Published Sep 26
Evaluated models Qwen3-VL-8B-GRPO 提出 TAME,在强化学习中抑制模板激活以缓解思维链混淆
Published Sep 21
Evaluated models Qwen3-VL-8B , Gemma-3-12B-it 提出 QPriv-VL,用问题引导剪枝降低分布式视觉问答的隐私泄露
Published Sep 14
Evaluated models LLaVA-1.5-7B , Llama-2 Abstract QPriv-VL 在切层前按问题相关度与 DINOv2 敏感度剪视觉 token,作者称能在约四成 token 下兼顾准确率与四类攻击。QPriv-VL 把问题引导的视觉 token 剪枝做成分布式 VQA 的客户端隐私防御,而不是只做集中式推理加速。