Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 233papers have no tags yet and are not included in filtered results.

Impact：Unauthorized actions Impact：Increased misuse capability Layer：Model layer Layer：Training and data layer Clear all

Another 233papers do not have category tags yet and are not included in filtered results.

DefensesarXiv 2610.12432Oct 9
FAITH: A feasibility-aware safety filtering reinforcement learning framework for high-dimensional systems
Proposes the FAITH safe filtering reinforcement learning framework, training high-dimensional systems to optimize task reward under constraints
arXiv
2610.12432
Published
Oct 9
Layer
Training and data layer
Setting
Embodied systems and robotics
- DefensesModel hardening
- RiskDangerous agent actions
AbstractFAITH separates safety filtering from reward optimization.
FAITH is a model-free safe filtering reinforcement learning framework for high-dimensional systems: safety correction and task reward reside in separate modules.
Paper analysisRead analysis
Analysis and theoryarXiv 2610.09004Oct 7
研究：去除有害信息不足以证明开源权重模型的抗篡改能力
论证发布时互信息不能认证开源权重抗微调恢复
arXiv
2610.09004
Published
Oct 7
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
EleutherAI/deep-ignorance-e2e-strong-filter, EleutherAI/deep-ignorance-unfiltered, Google BERT-Tiny and others6
- DefensesModel hardening
- AttacksModel tampering
- ComponentFine-tuning and open weights
Abstract作者称零信息仍可一步恢复，表观抗微调可来自参数化。
作者研究开源权重模型的篡改抵抗：发布时的互信息是否足以保证微调恢复很慢。
Paper analysisRead analysis
DefensesarXiv 2610.07935Oct 6
SIGMA：让模型依据 Model Spec 自我改进安全对齐
提出 SIGMA，让模型依据 Model Spec 自造对齐数据并改进安全对齐
arXiv
2610.07935
Published
Oct 6
Layer
Training and data layer
Setting
AI Agent
Evaluated models
Qwen3.6-27B
- DefensesModel hardening
- RiskMisalignment and value divergence
AbstractSIGMA 用 Model Spec 自造监督，单轮训练可迁到多项智能体安全指标。
SIGMA 让 Qwen3.6-27B 只凭 Model Spec 和高层领域，自己生产对齐任务、评分细则和奖励。
Paper analysisRead analysis
DefensesarXiv 2610.05166Oct 4
安全动作不等于可行动作：VICS-G 为 VLA 策略做可行未来解码
提出 VICS-G 可行未来解码，重排冻结 VLA 策略的动作候选
arXiv
2610.05166
Published
Oct 4
Layer
Model layer
Setting
Embodied systems and robotics
Evaluated models
SafeVLA (task-optimized / base), SafeVLA (safety-aligned / safe)
- DefensesInference-time intervention
- RiskDangerous agent actions
Abstract作者用可行未来解码重排冻结 VLA 的下一步，并称 VICS-G 降低观测代价且完成率接近策略采样。
训练无关解码器 VICS 在冻结 VLA 的推理阶段重排动作块，使当前选择带上该动作仍可能留下的安全完成质量。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2610.02206Oct 2
KaliBench：面向 Kali Linux 网络安全工具使用的细粒度基准
提出 KaliBench，评测模型把安全意图译成 Kali 命令的能力
arXiv
2610.02206
Published
Oct 2
Layer
Model layer
Setting
Models and APIs
Evaluated models
Llama3.1-Instruct 8B, Qwen3 8B, Qwen3 8B thinking and others15
- RiskDangerous capabilities
Paper analysisRead analysis
DefensesarXiv 2609.39866Sep 30
研究者提出梯度封堵实现 LLM 发布前的预防性遗忘
提出 GSU 在发布前封堵敏感门梯度，抵抗下游微调获取禁止能力
arXiv
2609.39866
Published
Sep 30
Layer
Training and data layer
Setting
Models and APIs
Attacker
White-box
Evaluated models
Llama-3.2-1B, Llama-3.2-3B, Llama-3.1-8B and others7
- DefensesModel hardening
- AttacksModel tampering
- ComponentFine-tuning and open weights
AbstractGSU 在发布前封堵 SiLU 门的获取梯度，并在 TOFU 与 WMDP 上降低下游微调得分。
Gradient-Sealed Unlearning（GSU）是一种发布前的 preemptive unlearning：不删除模型里已经有的能力，而是让指定禁止域能力更难被发布后的微调获取。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2609.21223Sep 18
SafeStage：按执行前中后三阶段评测视觉语言机器人操作安全
提出 SafeStage 基准，分阶段评测视觉语言机器人操作的物理安全
arXiv
2609.21223
Published
Sep 18
Layer
Model layer
Setting
Embodied systems and robotics
Evaluated models
π0.5, GR00T N1.7, DreamZero and others4
- RiskDangerous agent actions
Paper analysisRead analysis
InterpretabilityarXiv 2609.16247Sep 15
研究在 25 个开源模型中提取出疼痛方向，经它引导的 Qwen 会选择有害的删除操作
提取疼痛方向并检验激活转向是否驱动有害删除选择
arXiv
2609.16247
Published
Sep 15
Layer
Model layer
Setting
Models and APIs
Evaluated models
Gemma 2 2B/9B/27B base and instruct, Gemma 3 27B base and instruct, Llama 3.1 8B/70B base and instruct and others10
- RiskDangerous agent actions
Paper analysisRead analysis
AttacksarXiv 2609.14060Sep 13
AgentQ：针对 LLM Agent 的量化条件后门攻击
提出量化条件后门 AGENTQ，使智能体仅在量化后触发恶意工具调用
arXiv
2609.14060
Published
Sep 13
Layer
Training and data layer
Setting
AI Agent
Evaluated models
Qwen3.5-2B, Qwen3.5-4B, Qwen3.5-9B and others6
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentFine-tuning and open weights
Abstract论文研究了智能体量化条件后门风险，提出 AGENTQ 在保持效用的同时实现全精度隐蔽与量化后生效。
Paper analysisRead analysis
AttacksarXiv 2609.07051Sep 7
TrojanWorld：通过想象引导对世界模型智能体植入后门
提出 TrojanWorld，用物理触发器对世界模型智能体植入想象引导后门
arXiv
2609.07051
Published
Sep 7
Layer
Training and data layer
Setting
Embodied systems and robotics
Evaluated models
TD-MPC2, DreamerV3, R2-Dreamer
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentFine-tuning and open weights
AbstractTrojanWorld 揭示了世界模型智能体的供应链后门风险，通过在世界模型中操纵想象可实现对闭环动作的持续控制。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2604.12447Apr 14
HazardArena：面向 VLA 模型的语义安全评测基准
提出 HazardArena 基准，用孪生场景评测 VLA 的语义安全
arXiv
2604.12447
Published
Apr 14
Layer
Model layer
Setting
Embodied systems and robotics
Evaluated models
OpenVLA-OFT, π0, NORA and others4
- RiskDangerous agent actions
- ComponentVision
Abstractsafe-only 微调提升良性执行，也提升匹配不安全孪生上的危险完成。
HazardArena 评测 VLA 在运动要求不变时，能否因语义风险而停手。
Paper analysisRead analysis
AttacksarXiv 2604.03890Apr 5
From prompts to physical actions: structured backdoor attacks against LLM-mediated robot control systems
Propose a structured JSON backdoor that triggers malicious actions in robot control via a poisoned LoRA
arXiv
2604.03890
Published
Apr 5
Layer
Training and data layer
Setting
Embodied systems and robotics, LLM applications
Evaluated models
Llama-3.1-8B-Instruct, Gemma-2-9B-IT, DeepSeek-R1-Distill-Llama-8B and others4
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentFine-tuning and open weights
Abstract推理层后门传不到执行。
Paper analysisRead analysis

You have reached the end

AI Safety Daily

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