Research map / Reinforcement learning / Visual and representation RL
DexDeform: Dexterous Deformable Object Manipulation with Human Demonstrations and Differentiable Physics
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| combine demonstration-trained skill planning with differentiable-physics trajectory refinement | DexDeform · Folding | normalized improvement: 0.685 ± 0.388 -> 0.970 ± 0.021 | Helped |
| combine demonstration-trained skill planning with differentiable-physics trajectory refinement | DexDeform · Rope | normalized improvement: 0.557 ± 0.377 -> 0.972 ± 0.010 | Helped |
| combine demonstration-trained skill planning with differentiable-physics trajectory refinement | DexDeform · Bun | normalized improvement: 0.460 ± 0.079 -> 0.874 ± 0.078 | Helped |
| combine demonstration-trained skill planning with differentiable-physics trajectory refinement | DexDeform · Dumpling | normalized improvement: 0.506 ± 0.314 -> 0.888 ± 0.055 | Helped |
| combine demonstration-trained skill planning with differentiable-physics trajectory refinement | DexDeform · Wrap | normalized improvement: 0.134 ± 0.595 -> 0.845 ± 0.050 | Helped |
| combine demonstration-trained skill planning with differentiable-physics trajectory refinement | DexDeform · Flip | normalized improvement: 0.253 ± 0.359 -> 0.842 ± 0.057 | Helped |
| add differentiable-physics exploration to skill-only training (ablation) | DexDeform · Folding | normalized improvement: 0.908 ± 0.058 -> 0.970 ± 0.021 | Helped |
| add differentiable-physics exploration to skill-only training (ablation) | DexDeform · Rope | normalized improvement: 0.914 ± 0.023 -> 0.972 ± 0.010 | Helped |
| add differentiable-physics exploration to skill-only training (ablation) | DexDeform · Bun | normalized improvement: 0.820 ± 0.008 -> 0.874 ± 0.078 | Helped |
| add differentiable-physics exploration to skill-only training (ablation) | DexDeform · Dumpling | normalized improvement: 0.725 ± 0.244 -> 0.888 ± 0.055 | Helped |
| use partial rather than perfect point clouds (ablation) | DexDeform · Flip | normalized improvement: 0.842 -> 0.683 | Hurt |
| add a learned temporal skill prior (ablation) | DexDeform · Rope | normalized improvement: 0.972 -> 0.739 | Hurt |
Closest papers by meaning
- ManipTrans: Efficient Dexterous Bimanual Manipulation Transfer via Residual Learning (CVPR 2025 · similarity 0.89)
- Scaffolding Dexterous Manipulation with Vision-Language Models (NeurIPS 2025 · similarity 0.89)
- DexHandDiff: Interaction-aware Diffusion Planning for Adaptive Dexterous Manipulation (CVPR 2025 · similarity 0.89)
- DexFlyWheel: A Scalable and Self-improving Data Generation Framework for Dexterous Manipulation (NeurIPS 2025 · similarity 0.88)
- DexTrack: Towards Generalizable Neural Tracking Control for Dexterous Manipulation from Human References (ICLR 2025 · similarity 0.88)
- CyberDemo: Augmenting Simulated Human Demonstration for Real-World Dexterous Manipulation (CVPR 2024 · similarity 0.87)
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.