Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 52 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.
Trace attention to identify harmfulness to the MLP refusal write pathway and propose Sentinel
Published Oct 9
Evaluated models Llama-3.1-8B-Instruct , Qwen2.5-7B-Instruct , Qwen3-8B-Instruct and others18 Abstract Attention updates are more readable for harmfulness, MLP updates write refusal, and Sentinel routes alerts using two-path signals.The authors reframe jailbreaking and excessive refusal of aligned models as a pathway problem on the residual stream from recognition to writing, and accordingly provide conditional reminders before decoding.
Proposes ReSI, which recursively improves model safety alignment using automatic red teaming and recipe search.
Published Oct 9
Evaluated models Qwen3.6-27B , Ministral-3-14B-Reasoning-2512 , DeepSeek-V4-Flash-0731 and others4 提出 GUISE 基准与 AXIS 训练,防御叙事包装越狱
Published Oct 8
Evaluated models Qwen3-1.7B , Qwen3-4B , GLM-4-9B Abstract GUISE 分开测包装与语域。作者测量叙事包装下的拒答失效,并训练一个在未见包装上拒绝、同时仍回答同风格良性请求的适配器。问题是:安全对齐模型拒绝直说的有害请求,却回答嵌进角色或叙事里的同一请求。文言文被用来把语域和包装拆开。
提出 SLDR,用选择性层恢复与动态路由防御恶意微调
Published Oct 8
Evaluated models Llama3.1-8B-Instruct , Llama3-8B-Instruct , Qwen2.5-7B-Instruct and others4 Abstract 作者称 SLDR 能压低有害输出并保持下游效用。SLDR 是针对恶意微调的微调后防御:在已对齐模型上做有符号层探测,只在敏感分数最大与最小的两层训练 LoRA,再用 lmax 的最后 token 表示决定是否启用适配器。
提出 SafeBridge,对齐循环语言模型不同循环深度的安全性
Published Oct 7
Evaluated models Ouro-1.4B , Ouro-2.6B 用 SafeEvo 提取拒绝电路,并提出 SCA 仅微调该电路
Published Oct 7
Evaluated models Llama-3-8B , Qwen-2.5-7B , Mistral-7B-v0.1 and others4 Abstract SafeEvo 识别出基座模型中存在的弱安全电路及其演化动态,据此提出的 SCA 算法实现了安全对齐、能力保留与低过度拒绝的平衡。
提出 ASCENT,在效用微调中做一阶安全校准并只压制负贡献更新
Published Oct 6
Evaluated models Qwen2.5-7B-Instruct , Llama-3.1-8B-Instruct , Gemma-3-12B-it and others4 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。
提出审计感知遗忘框架 AVCE,从文本与视觉通路擦除扩散模型概念
Published Oct 5
Evaluated models Stable Diffusion XL (SDXL) , Stable Diffusion v1.5 (SD v1.5) , Flux 1.0 Abstract AVCE 在审计锚点上改两条注意力通路。AVCE 是面向扩散模型的概念擦除方法,要让擦除经得起绕过文本条件的潜空间审计,而不只是挡住对抗提示词。
形式化 watchman 防御以阻止序贯马赛克攻击
Published Oct 5
Evaluated models Qwen3.6-35B-A3B , Gemma-4-26B-A4B-it , DeepSeek-V2-Lite (16B-A2.4B) Abstract watchman 理论给出零失败且有用的条件。这篇工作同时给出序贯马赛克防御的博弈理论,以及冻结 LLM 上的分角色自博弈训练。
提出 SSRFT,用角色监督内化安全原则以替代拒答式对齐
Published Oct 4
Evaluated models Qwen2.5-3B , Gemma2-2B , Llama3.1-8B and others6 Abstract SSRFT 用角色监督替代拒答中心对齐。SSRFT 是一种用监督微调把预定义安全角色写进参数的安全对齐方法。作者称它是首个以安全角色内化本身为对齐目标的框架。
提出 Target-free LAT,用无目标多方向潜干预做对抗安全对齐以抵御越狱
Published Oct 3
Evaluated models Qwen2.5-7B-Instruct , Llama-3-8B-Instruct , Mistral-7B-Instruct-v0.2 Abstract Target-free LAT 用无目标多方向潜干预扩大行为覆盖,作者称能提高越狱鲁棒性并保住效用。Target-free LAT 是一种不监督有害输出的潜空间对抗对齐。它要处理的问题是:现有 LAT 每步大多只造一种有害行为,对不上真实越狱里的直接遵从、角色扮演、编码和多语言等模式。
提出 FADE,对文生视频模型做帧感知多概念擦除
Published Oct 3
Evaluated models Wan2.1-T2V-1.3B , Wan2.1-T2V-14B , CogVideoX-2B and others4 Abstract FADE 用闭式编辑、帧感知适配器和软路由,在单骨干上擦除多概念并压低逐帧残差。FADE 是面向文生视频扩散 Transformer 的多概念遗忘方法,用来抑制指定物体、画风、裸体、暴力、仇恨和名人身份,并尽量保持其他生成。作者要处理两件事。与帧无关的抑制会让概念在少数帧重现,片段均值会掩盖这一失败。
提出 PlurPO,用自模拟利益相关者偏好后训练缓解社交谄媚
Published Oct 2
Evaluated models Qwen3-8B , Phi-4 , Llama 3.1 8B and others5 Abstract PlurPO 用自身多方否决信号做后训练,在社会谄媚指标上降低过度认可。PlurPO 是一种后训练方法,用来降低个人建议等无真值场景中的社会谄媚。
提出 FDCU 双约束子空间投影遗忘,抵御再训练恢复有害知识
Published Sep 30
Evaluated models Qwen2.5-3B , Llama-3-8B-Instruct , Qwen-3-8B Abstract FDCU 用双掩码限制浅层抑制,在知识擦除与安全输出上提高再训练后的稳定性。FDCU 是一种约束参数更新的遗忘方法,用来降低再训练把已抑制恶意行为重新引出的可能。
用持续更新的探针做微调,降低有害性并保持线性可监测性
Published Sep 30
Evaluated models Mistral 7B Instruct v0.1 , Llama 3 8B Instruct Abliterated , Qwen3-14B Abstract 持续重拟合的探针可降低有害性、提高诚实性,训练后概念仍能被线性探针读出。Probe-guided fine-tuning 把探针分数当作唯一对齐信号,用来推动 harmlessness 与 honesty,并检查训练后这些概念是否仍线性可读。
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。
用纠正性监督改写固定有害微调样本,缓解涌现性失准
Published Sep 29
Evaluated models Qwen2.5-14B-Instruct , Gemma-4-12B-it Abstract 在固定毒行上,把坏回答改成正确答案比删掉更能降低 EM,正确内容比多出来的安全指令更关键。Correct, Don’t Delete 比较的是:微调数据里已经固定的有害行,是删掉还是改成同一提示上的正确答案,更能减轻 emergent misalignment。
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 提出 SURE 三阶段框架,用韩语对抗提示与安全打分做 SFT 与 DPO
Published Sep 29
Evaluated models KULLM3-10.7B , Qwen-2-7B , Llama-3-8B