可玩视频生成
TLDR
无监督学习可玩视频生成，用户通过离散动作控制视频，采用自监督编码器-解码器与动作瓶颈。
评分理由
The paper introduces a novel unsupervised problem and a self-supervised framework that learns discrete action labels from unlabelled videos. Strengths include a clear problem formulation and demonstration on diverse datasets, but the abstract lacks explicit comparison to world models or reinforcement learning, and the evaluation details are not fully described.
Read-first 评分解释
综合优先阅读分 30.6，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 11。
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：37。