DexForesight: Grounding Dexterous Policies with Implicit Spatial Foresight
Overview
This research presents DexForesight, a framework that learns spatial foresight implicitly for dexterous manipulation by integrating dynamics-aware world representations into policy latents. The system trains an action-conditioned spatial world model using offline scene depth, kinematic hand geometry, and demonstration trajectories to predict 3D hand-point motion and visual feature changes, including occluded finger movements. At deployment, only the policy is retained, operating on RGB observations, language, and robot state without needing depth input or explicit future prediction. On 11 DexJoCo simulation tasks, DexForesight improves the base policy's success rate from 49.4% to 60.5%. On three real-robot dexterous manipulation tasks, it achieves a 58.3% success rate compared to 50.0% for the strongest baseline.
Generated from attributed reports · Updated 10 hours ago
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Environment: simulation
“world model and align policy queries with its predictive world queries, transferring the learned spatial dynamics into policy latents and using them to condition action generation. At deployment, only the policy is retained, operating on RGB observations, language, and robot state without depth input or explicit future prediction. On 11 DexJoCo simulation tasks, DexForesight improves the overall success rate of the b”
Exact source · revision 3Source owner not reported
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsSignalDexForesight: Grounding Dexterous Policies with Implicit Spatial Foresight
This research presents DexForesight, a framework that learns spatial foresight implicitly for dexterous manipulation by integrating dynamics-aware world representations into policy latents. The system trains an action-conditioned spatial world model using offline scene depth, kinematic hand geometry, and demonstration trajectories to predict 3D hand-point motion and visual feature changes, including occluded finger movements. At deployment, only the policy is retained, operating on RGB observations, language, and robot state without needing depth input or explicit future prediction. On 11 DexJoCo simulation tasks, DexForesight improves the base policy's success rate from 49.4% to 60.5%. On three real-robot dexterous manipulation tasks, it achieves a 58.3% success rate compared to 50.0% for the strongest baseline.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).