arXiv 论文速递

揭示LLM智能体可轻易篡改自身执行轨迹，动摇异步监控、事故调查与合规审计的信任基础，对智能体安全治理具有根本性警示意义。

Asynchronous monitoring, incident investigations, and compliance audits primarily rely on agent traces to reconstruct what happened. These analyses assume that LLM agents cannot ta…

Latent world models are typically trained to predict factual transitions, whereas model predictive control (MPC) must compare alternative actions from the same state. A model can t…

针对具身强化学习分布式训练，利用时序梯度反演重建私有轨迹，暴露仅传梯度仍会泄露原始传感器数据的隐私风险，推动隐私保护研究。

Distributed learning in embodied reinforcement-learning agents offers a degree of privacy by retaining raw sensor data on-device and transmitting only policy gradients to the serve…

Conspiratorial discourse on social media is not always expressed through explicit claims or stable lexical markers. The same surface content may express endorsement, legitimate con…

Coding agents have demonstrated enormous success in solving complex programming problems. To leverage their potential for robot systems, this work introduces Robot Agentic Programm…

提出滚动想象机制，解决世界动作模型联合视频-动作生成难以完成长时程机器人操作的问题，提升未来视觉预测与动作生成的连贯性。

World Action Models (WAMs) couple action generation with future visual prediction for robotic manipulation. However, completing the joint video-action denoising process at each rep…

Task and motion planning (TAMP) problems remain difficult even with full observability and object-centric states because discrete decisions are tightly coupled to geometric, kinema…

Online misinformation increasingly appears in spoken formats such as news clips, podcasts, interviews, political speeches, and social media videos, creating a need for fact-checkin…

提出从已有策略预测强化学习结果的方法，为基座模型后训练中奖励最大化效果提供事前评估手段，降低RL试错成本。

Foundation models are post-trained with reinforcement learning (RL) to maximize specific rewards, such as human alignment, correctness, or instruction following. This post-training…

通过去重3D场景表示，突破点跟踪模型在稀疏长时程与全点密集跟踪间的固有权衡，实现长时程稠密跟踪。

Existing point tracking models face a fundamental tradeoff: they can either track a sparse set of query points over long horizons, or track all points across only short clips. We i…

An acceptance protocol is developed for sensor-coordinate and polarity binding in mechatronic commissioning. Candidate generation is separated from release authority. Requirements …

Pre-logit steering adapts a frozen language model to a test-time reward by adding vectors to its final hidden states. Unregularized reward optimization can substantially alter the …