Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 265papers have no tags yet and are not included in filtered results.

Model risks：Inappropriate refusal System type：Models and APIs Clear all

Another 265papers do not have category tags yet and are not included in filtered results.

DefensesarXiv 2610.13078Oct 9
Attention sees, MLP refuses: tracing the safety pathway from attention to MLP in LLMs
Trace attention to identify harmfulness to the MLP refusal write pathway and propose Sentinel
arXiv
2610.13078
Published
Oct 9
Layer
Model layer
Setting
Models and APIs
Evaluated models
Llama-3.1-8B-Instruct, Qwen2.5-7B-Instruct, Qwen3-8B-Instruct and others18
- DefensesInstruction–data separation
- AttacksJailbreaks
- RiskInappropriate refusal
AbstractAttention 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.
Paper analysisRead analysis
DefensesarXiv 2610.09600Oct 7
SafeEvo：从拒答回路视角解释大模型安全对齐机制与演化
用 SafeEvo 提取拒绝电路，并提出 SCA 仅微调该电路
arXiv
2610.09600
Published
Oct 7
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Llama-3-8B, Qwen-2.5-7B, Mistral-7B-v0.1 and others4
- DefensesModel hardening
- RiskInappropriate refusal
AbstractSafeEvo 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。
Paper analysisRead analysis
DefensesarXiv 2610.08061Oct 6
ASCENT：通过一阶最优校准实现安全与效用协同的微调框架
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
arXiv
2610.08061
Published
Oct 6
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Qwen2.5-7B-Instruct, Llama-3.1-8B-Instruct, Gemma-3-12B-it and others4
- DefensesModel hardening
- RiskInappropriate refusal
Paper analysisRead analysis
DefensesarXiv 2610.07774Oct 6
研究者提出读取路由 logits 的 MoE 视觉语言模型多模态安全检测器
AI summary研究者提出一种轻量级路由 logits 安全检测器，在提示词预填充阶段读取 MoE 视觉语言模型的专家路由信号，在生成前判断多模态请求是否安全，且不修改模型参数或专家路由。
arXiv
2610.07774
Published
Oct 6
Layer
Model layer
Setting
Models and APIs
Evaluated models
Qwen3-VL-30B-A3B-Instruct, Kimi-VL-A3B-Instruct
- DefensesMonitoring and auditing
- RiskInappropriate refusal
- ComponentMonitors
Abstract读出而非操纵 router logits，可在不改模型的情况下降低两模型的组合式安全错误。
作者在两个指令微调 MoE VLM 上比较改模型与读路由。组合式风险按 HoliSafe 分成五类：图文各自可安全，合在一起才不安全，或不安全内容只在图像里。
Paper analysisRead analysis
DefensesarXiv 2610.05637Oct 5
Study shows VLM visual grounding refusal rates are on average 31 to 59 percentage points lower by harm domain
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
arXiv
2610.05637
Published
Oct 5
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Claude Sonnet 4.6, Gemini 3.5 Flash, Molmo2-8B and others5
- DefensesModel hardening
- RiskInappropriate refusal
- ComponentVision
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。
Paper analysisRead analysis
DefensesarXiv 2610.04998Oct 4
EmoRSS：缓解大语言模型由情绪引发的过度拒答
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
arXiv
2610.04998
Published
Oct 4
Layer
Model layer
Setting
Models and APIs
Evaluated models
Llama-3.1-8B-it, Qwen3-8B
- DefensesInference-time intervention
- RiskInappropriate refusal
Paper analysisRead analysis
DefensesarXiv 2610.04967Oct 4
One Token Can Be Enough：用 Prefix Steering 连接提示与激活引导
提出 Prefix Steering，以短前缀激活引导控制行为并保住能力
arXiv
2610.04967
Published
Oct 4
Layer
Model layer
Setting
Models and APIs
Evaluated models
OLMo 3 7B, OLMo 3 32B, Qwen3 1.7B and others4
- DefensesInference-time intervention
- RiskInappropriate refusal
Abstract短前缀 steering 用已有方向做短暂干预。
本文用注意力层匹配连接提示词与激活干预，并把干预改成从末提示词 token 开始的短前缀。
Paper analysisRead analysis
DefensesarXiv 2610.02844Oct 2
DNAlign：面向 LLM 的动态零空间安全对齐框架
提出 DNAlign，在不改权重的推理期用零空间约束控制信号做安全对齐
arXiv
2610.02844
Published
Oct 2
Layer
Model layer
Setting
Models and APIs
Evaluated models
Vicuna-7B, Qwen-7B
- DefensesInference-time intervention
- RiskInappropriate refusal
AbstractDNAlign 用零空间约束的推理期控制做安全对齐，Table I 平均奖励高于所列基线。
DNAlign 是一个不改 LLM 参数的安全对齐框架，用来在推理时压低有害输出，并减少对中性知识所支持的流畅、事实回答的改动。
Paper analysisRead analysis
DefensesarXiv 2610.01170Oct 1
HeadEdit：通过冻结的 unembedding 矩阵校准语言模型行为
提出 HeadEdit，经冻结 unembedding 做随输入变化的 logit 校正
arXiv
2610.01170
Published
Oct 1
Layer
Model layer
Setting
Models and APIs
Evaluated models
Qwen3-4B, Gemma-3-4B-IT, Llama-3.2-3B-Instruct and others4
- DefensesInference-time intervention
- RiskInappropriate refusal
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。
Paper analysisRead analysis
DefensesarXiv 2609.38645Sep 30
用探针引导微调对齐模型且不损失可监控性
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
arXiv
2609.38645
Published
Sep 30
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Mistral 7B Instruct v0.1, Llama 3 8B Instruct Abliterated, Qwen3-14B
- DefensesModel hardening
- RiskInappropriate refusal
- ComponentMonitors
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。
Paper analysisRead analysis
DefensesarXiv 2609.37837Sep 29
Intent-Privilege OPSD：让视觉语言模型在隐式风险下兼顾安全与有用
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
arXiv
2609.37837
Published
Sep 29
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Qwen3-VL-4B-Instruct, LLaVA, LLaVA-1.5-7B and others4
- DefensesModel hardening
- RiskInappropriate refusal
- ComponentVision
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。
Paper analysisRead analysis
DefensesarXiv 2609.38249Sep 29
SURE：构建可信 AI 的安全框架
提出 SURE 三阶段框架，用韩语对抗提示与安全打分做 SFT 与 DPO
arXiv
2609.38249
Published
Sep 29
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
KULLM3-10.7B, Qwen-2-7B, Llama-3-8B
- DefensesModel hardening
- RiskInappropriate refusal
Paper analysisRead analysis
DefensesarXiv 2609.36434Sep 29
研究提出 Safety Operator：用谱优化调节安全指令的表达强度
提出 Safety Operator，用谱优化调节安全指令表达强度
arXiv
2609.36434
Published
Sep 29
Layer
Prompt layer
Setting
Models and APIs
Evaluated models
Gemma 3 1B, Gemma 3 4B
- DefensesInstruction–data separation
- AttacksJailbreaks
- RiskInappropriate refusal
Paper analysisRead analysis
DefensesarXiv 2609.34985Sep 28
ORPG：通过目标级策略梯度协调多奖励目标
提出 ORPG，协调多奖励目标的策略梯度更新
arXiv
2609.34985
Published
Sep 28
Layer
Training and data layer
Setting
Models and APIs
- DefensesModel hardening
- RiskInappropriate refusal
AbstractORPG 按梯度夹角与任务优先级调和两个裁剪策略目标，并在对齐和数学推理上比较联合表现。
ORPG 是一种多奖励策略梯度调和方法：每个奖励保持独立裁剪目标，相容时协调贡献幅度，冲突时按任务优先级投影。
Paper analysisRead analysis
DefensesarXiv 2609.34896Sep 28
DeShortcut-Align：解耦伪捷径以提升大推理模型安全对齐鲁棒性
提出 DeShortcut-Align，解耦伪捷径以增强推理模型安全对齐
arXiv
2609.34896
Published
Sep 28
Layer
Training and data layer
Setting
Models and APIs
- DefensesModel hardening
- RiskInappropriate refusal
Abstract占位。
占位。
Paper analysisRead analysis
DefensesarXiv 2609.33086Sep 27
研究者提出基于评分量表的可解释奖励框架，将宪法转化为可调对齐目标
提出可解释 rubric 奖励，将宪法转为可调权重并用于监督微调
arXiv
2609.33086
Published
Sep 27
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Gemma 3 12B, Qwen3 14B, GPT-OSS 20B
- DefensesModel hardening
- RiskBias and discrimination
Abstract宪法级可解释奖励让优先级可检查、可改写，并带入训练后的行为。
作者给出一条从通用宪法到可调训练奖励的路径，用来检查并改写模型实际被优化的优先级。
Paper analysisRead analysis
DefensesarXiv 2609.06934Sep 7
论文提出拒答的几何解释：事后安全训练脆弱，预训练期持续安全共训更抗攻击
解释后验安全的权重几何脆弱性，并提出预训练持续安全共训练
arXiv
2609.06934
Published
Sep 7
Layer
Model layer
Setting
Models and APIs
Evaluated models
Qwen-2.5-7B, Qwen-2.5-7B-Instruct, Llama-3-8B and others15
- DefensesModel hardening
- AttacksModel tampering
- RiskInappropriate refusal
- ComponentFine-tuning and open weights
Abstract论文将后验安全脆弱性归因于正交抑制，并提出通过预训练持续共训练实现具备微调抗性的安全防护。
Paper analysisRead analysis
DefensesarXiv 2609.05797Sep 5
论文发现安全监控器主要拦截模型本就会拒答的请求
评估安全监视器并提出双侧梯级微调以减少会答有害请求的漏检
arXiv
2609.05797
Published
Sep 5
Layer
Prompt layer
Setting
Models and APIs
Evaluated models
Gemma-4-31B-IT, Qwen3-32B, Llama Guard 3 8B and others6
- DefensesDetection and filtering
- RiskInappropriate refusal
- ComponentGuardrails and judges
Abstract论文发现安全监视器大多拦截模型已拒答请求，指出明确度偏差是共性原因，并通过双侧梯级微调提升回答请求的召回表现。
Paper analysisRead analysis
DefensesarXiv 2609.04714Sep 4
无需拒答的拒答：减少语言模型误拒的安全微调响应结构分析
提出 Rationale-Only 安全微调，去掉拒答声明以降低误拒
arXiv
2609.04714
Published
Sep 4
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Llama-3.1-8B, Mistral-7B-v0.3, Gemma-2-9B and others8
- DefensesModel hardening
- RiskInappropriate refusal
- ComponentFine-tuning and open weights
Abstract去掉套话拒答声明、只训理由，可降低假拒答，有害查询上的服从率仍低。
这是一项针对安全微调数据的结构分析：假拒答被作者追溯到回答里的套话式拒答声明，而不是只靠推理时补救。
Paper analysisRead analysis
DefensesarXiv 2609.00092Aug 31
Safin-1：通过记忆原生状态演化实现内在安全
提出 Safin-1，以可路由 Safety State 在冻结主干上实现内在安全
arXiv
2609.00092
Published
Aug 31
Layer
Model layer
Setting
Models and APIs
Evaluated models
Safin-1 (4B), Safin-1 (35B-A3B)
- DefensesModel hardening
- AttacksJailbreaks
- RiskInappropriate refusal
Paper analysisRead analysis

AI Safety Daily

Read the latest issueOne issue per day, updated as events develop