无需动作标签的学习行动
TLDR
提出LAPO方法，从视频中恢复潜在动作，无需动作标签即可训练策略和世界模型，并可在网络视频上预训练。
评分理由
The paper presents a novel method for extracting latent action information from videos, which is a significant step toward pre-training RL agents on unlabeled data. Strengths include the innovative approach and potential for scaling to web-scale video. Weaknesses are that experiments are only on procedurally-generated environments, and no real-world validation is mentioned.
Read-first 评分解释
综合优先阅读分 39.7，由主题、引用、图谱、方法、可复现性和近期性等信号加权得到。 原始总分保留为 29。
研究版图角色
候选论文
排序敏感性
稳定性：volatile；排名波动范围：253。