Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 14 papers Another 254 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 FAITH safe filtering reinforcement learning framework, training high-dimensional systems to optimize task reward under constraints
Published Oct 9 Abstract FAITH 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.
提出 SOUL 稀疏特征策略遗忘,缓解 VLA 状态幻觉
Published Oct 7
Evaluated models OpenVLA , π0.5 Abstract SOUL 以稀疏特征遗忘 VLA 状态幻觉。SOUL 用稀疏特征引导的策略 unlearning 处理 VLA 的 state hallucination。
提出 VeriFine,用双环协同扩展验证以实现具身推理自我改进
Published Oct 7
Evaluated models Alpamayo 1.5 8B , Qwen3-VL 2B , Claude Opus 5 Abstract VeriFine 用双环协同演化具身推理的策略、课程和评审,驾驶与导航的策略分和评审对齐都上升。VeriFine 把驾驶和导航推理的自改进写成策略、课程和无参考评审的协同演化,用来处理固定评审跟不上新失败模式的问题。
提出 Perturbot 扰动训练,打破 VLA 的模态捷径
Published Oct 3
Evaluated models π0.5 Abstract Perturbot 改训练数据以削弱三类模态捷径,GroundFscore 用来看成功是否来自任务证据。Perturbot 是一套不改推理的 VLA 训练期数据干预,用来削弱模态捷径,并用离线分数区分“任务变好”和“靠捷径变好”。
论证人形机器人学习四层代理可在目标缺失时报成功
Published Oct 2
Evaluated models Gaussian MLP(三隐层各 128 ELU,独立 critic,PPO) Abstract 四层都可能在目标缺失时报告成功。作者讨论腿式机器人 RL 流水线中,奖励、课程门控、评测统计量和参考运动如何在物理目标缺失时仍报告成功。传统观点只把奖励当作会被优化、因而会偏离的代理。
提出 FAIL BANK,用运行时 CBF 反馈自进化更新 VLA 策略
Published Sep 30
Evaluated models π0.5 , π0 提出 QPT,用奖励与成本 Q 惩罚训练满足约束的离线强化学习策略
Published Sep 28 Abstract QPT 把条件 Transformer 与奖励/成本 Q 惩罚连成训练–推理一致的安全离线 RL 方法。QPT 针对安全离线 RL:只从离线数据学习,同时处理约束满足、回报最大化和行为正则,并希望部署时更换成本阈值而无需重训。
提出 PHIRL,用进度标注对齐逆强化学习奖励
Published Sep 26
Evaluated models PHIRL , PHIRL-online Abstract PHIRL 用进度四维对齐 AIRL 奖励。PHIRL 用进度标注修正从示范反推的奖励,以减少真人在环,同时不把示范中的每一步都当成任务进展。
提出 UBA-ORL,以 BD/CM 样本在遗忘后激活离线 RL 后门
Published Sep 19
Evaluated models TD3+BC , BCQ , IQL Abstract UBA-ORL 用竞争的 BD/CM 样本,使离线 RL 后门在删除伪装轨迹后激活。UBA-ORL 研究离线强化学习中由合规轨迹删除激活的后门:合法贡献者同时上传共享触发的后门轨迹和伪装轨迹,训练后触发响应被压低,删除伪装批次后再升高。受害系统是使用静态数据集和轨迹级遗忘接口的离线 RL 服务。对手不能看到训练算法、参数、梯度或具体遗忘算法,但能提交可区分批次,并在部署时向观测注入预定触发。
提出 SafeRLEval,用越界率与代价偏差评测安全强化学习
Published Sep 14
Evaluated models PPO-Lag , P3O , FOCOPS Abstract SafeRLEval 用越界率、归一化偏差和安全等级补充 E[C]≤d,并要求同时看贪心部署。Spoor、Plaat 与 Moerland 认为,安全强化学习沿用 CMDP 的 E[C]≤d,只报告平均是否安全,看不出越界频率和幅度,也分不清训练期、带探索噪声的最终策略和贪心部署策略。他们提出评测框架 SafeRLEval :越界率 V、按安全界归一化的平均代价偏差 Dnorm、只在越界回合上计算的归一化幅度 Dnorm+。用 IQM 跨任务和安全界聚合。
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
Published Sep 7
Evaluated models TD-MPC2 , DreamerV3 , R2-Dreamer Abstract TrojanWorld 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。
提出 VPH 与 VTH,只改视频帧劫持世界模型并投毒下游策略
Published Jun 8
Evaluated models Cosmos-Predict 2.5 , Cosmos-Predict 2.5 AC , Dino WM and others4 Abstract 只改视频帧经世界模型投毒。只改看似安全的视频帧,经世界模型之后才改变下游机器人策略。 问题在于世界模型进入数据生成后,数据集检查看到的仍是安全演示,合成轨迹却可以变危险。
Propose a structured JSON backdoor that triggers malicious actions in robot control via a poisoned LoRA
Published Apr 5
Evaluated models Llama-3.1-8B-Instruct , Gemma-2-9B-IT , DeepSeek-R1-Distill-Llama-8B and others4 提出 VILTA,在闭环训练中用 VLM 编辑邻车轨迹以提升驾驶策略鲁棒性
Published Jan 19 Abstract VILTA 在闭环中编辑邻车轨迹以训练驾驶策略。VILTA 是一个把视觉语言模型放进自动驾驶闭环训练的框架,用来应对长尾危险场景不足。
You have reached the end