Research map / Reinforcement learning / Visual and representation RL
DemoGrasp: Universal Dexterous Grasping from a Single Demonstration
Haoqi Yuan, Ziye Huang, Ye Wang, Chuan Mao, Chaoyi Xu, Zongqing Lu
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 |
|---|---|---|---|
| replace RL with sampling editing parameters and behavior cloning (ablation) | DemoGrasp · 175 training objects | success rate: 96.24 -> 77.56 | Hurt |
| add wrist translation editing to RL action space (vs. replay) (ablation) | DemoGrasp · Training set | success rate: 75.29 -> 81.35 | Improved |
| add wrist rotation editing to RL action space (ablation) | DemoGrasp · Training set | success rate: 81.35 -> 94.22 (with translation) | Improved |
| add hand DoF editing to RL action space (ablation) | DemoGrasp · Training set | success rate: 94.22 -> 96.24 | Improved |
| use two RGB views instead of two depth views (ablation) | DemoGrasp vision policy · YCB / real small objects | success rate: 80.3% -> 87.0% (YCB) | Improved |
| use two cameras instead of monocular (ablation) | DemoGrasp vision policy · YCB (RGB) | success rate: 83.2% -> 87.0% | Improved |
| train directly on test sets instead of 175 objects (ablation) | DemoGrasp · five test datasets | success rate: +2.4% average gain | No clear effect |
| use different successful demonstrations (ablation) | DemoGrasp · demonstration quality | success rate | No clear effect |
| condition policy on language instruction instead of unconditioned | DemoGrasp vision policy · Cluttered grasping (sim) | success rate: 83.66% -> 85.33% | Improved |
| mount Shadow Hand on a robot arm instead of floating wrist | DemoGrasp · Test datasets | success rate: 1.4% lower on average | No clear effect |
Later papers that compared with this method
1 later accepted paper reports DemoGrasp as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- DemoFunGrasp: Universal Dexterous Functional Grasping via Demonstration-Editing Reinforcement Learning (CVPR 2026 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DexGrasp Anything: Towards Universal Robotic Dexterous Grasping with Physics Awareness (CVPR 2025 · similarity 0.93)
- Cross-Embodiment Dexterous Grasping with Reinforcement Learning (ICLR 2025 · similarity 0.92)
- UniGraspTransformer: Simplified Policy Distillation for Scalable Dexterous Robotic Grasping (CVPR 2025 · similarity 0.92)
- Scaffolding Dexterous Manipulation with Vision-Language Models (NeurIPS 2025 · similarity 0.91)
- CADGrasp: Learning Contact and Collision Aware General Dexterous Grasping in Cluttered Scenes (NeurIPS 2025 · similarity 0.90)
- ManipTrans: Efficient Dexterous Bimanual Manipulation Transfer via Residual Learning (CVPR 2025 · similarity 0.90)
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.
