arXiv Artificial Intelligence· Yuanpeng He, Ceyao Zhang, Fangjing Li, Kefei Zhu, Lijian Li, Yueheng Li, Tianjia He, Wenpin Jiao, Zhi Jin, Yaodong Yang, Yuanpei Chen·· 19 小时前AI 评分30
RAP-WAM：通过预测排练与动作扰动学习世界-动作模型
RAP-WAM: Learning World-Action Models through Predictive Rehearsal and Action Perturbations
AI 导读
RAP-WAM 将预测排练（Predictive Rehearsal）与动作扰动（Action Perturbations）结合未来监督查询和 Coupled Action Projection 进行策略学习，无需独立评分网络或部署时探测，基座直接从视频预训练骨干学习。
来源：arXiv Artificial Intelligence · arxiv.org