Zhongxi Chen, Yi Yang, Xinyi Chen, Mingzhe Ye, Yaping Li, Zhihong Liu, Congsheng Xu, Haonan Wang (Sharpa) · Implicit Foresight Policy

DexForesight: Grounding Dexterous Policies with Implicit Spatial Foresight

DexForesight trains a dexterous robot policy with an action-conditioned spatial world model that predicts 3D hand-point motion and visual feature changes. The world model is then frozen and its predictive queries are aligned into the policy's latent space, so deployment needs only RGB, language and robot state. On 11 DexJoCo simulation tasks the success rate rises from 49.4% to 60.5%, and on three real-robot tasks it reaches 58.3% against 50.0% for the strongest baseline.