Research map / Video / Video generation
Spatial-Temporal Aware Visuomotor Diffusion Policy Learning
Zhenyang Liu, Yikai Wang, Kuanning Wang, Longfei Liang, Xiangyang Xue, Yanwei Fu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add 3D/4D awareness supervision (DP4) over DP3 | DP3 · Adroit (overall) | success rate (%): 68.3 -> 84.7 | Improved |
| add DP4 supervision | DP3 · DexArt (overall) | success rate (%): 68.5 -> 82.5 | Improved |
| DP4 vs ManiGaussian | ManiGaussian · RLBench first five tasks (avg) | success rate (%): 57.1 -> 63.3 | Improved |
| DP4 vs ManiGaussian | ManiGaussian · RLBench last five tasks (avg) | success rate (%): 33.2 -> 39.9 | Improved |
| DP4 vs DP3 | DP3 · Real-robot (overall) | success rate (%): 46.0 -> 54.6 | Improved |
| add RGB supervision via Gaussian world model (ablation) | DP4 base · Adroit Pen | success rate (%): 45.0 -> 48.0 | Improved |
| add depth supervision alone (ablation) | DP4 base · Adroit Pen | success rate (%): 45.0 -> 47.0 | No clear effect |
| add depth to RGB supervision (ablation) | DP4 base · Adroit Pen | success rate (%): 48.0 -> 72.0 | Improved |
| add 4D dynamics supervision (ablation) | DP4 (RGB+depth) · Adroit Pen | success rate (%): 72.0 -> 75.0 | Improved |
| add 4D dynamics supervision (ablation) | DP4 (RGB+depth) · Adroit Door | success rate (%): 75.0 -> 80.0 | Improved |
| add 4D dynamics supervision (ablation) | DP4 (RGB+depth) · Adroit Hammer | success rate (%): 98.0 -> 100.0 | Improved |
| add RGB+depth supervision to base (ablation) | DP4 base · Adroit Door | success rate (%): 64.0 -> 75.0 | Improved |
Closest papers by meaning
- Latent Diffusion Planning for Imitation Learning (ICML 2025 · similarity 0.90)
- DynaMo: In-Domain Dynamics Pretraining for Visuo-Motor Control (NeurIPS 2024 · similarity 0.89)
- 3D Equivariant Visuomotor Policy Learning via Spherical Projection (NeurIPS 2025 · similarity 0.89)
- Prediction with Action: Visual Policy Learning via Joint Denoising Process (NeurIPS 2024 · similarity 0.89)
- VidMan: Exploiting Implicit Dynamics from Video Diffusion Model for Effective Robot Manipulation (NeurIPS 2024 · similarity 0.88)
- GravMAD: Grounded Spatial Value Maps Guided Action Diffusion for Generalized 3D Manipulation (ICLR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
