Research map / Reinforcement learning / Visual and representation RL
DemoFunGrasp: Universal Dexterous Functional Grasping via Demonstration-Editing Reinforcement Learning
Chuan Mao, Haoqi Yuan, Ziye Huang, Chaoyi Xu, Kai Ma, 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 |
|---|---|---|---|
| add affordance reward (ablation) | DemoFunGrasp · state-based sim | SAD (cm): 4.60 -> 3.02 | Improved |
| add affordance reward (ablation) | DemoFunGrasp · state-based sim | GSR (%): 78.28 -> 77.04 | Hurt |
| add style disturbance (ablation) | DemoFunGrasp · state-based sim | GSR (%): 58.67 -> 77.04 | Improved |
| add object size clipping (ablation) | DemoFunGrasp · state-based sim | SAD (cm): 3.63 -> 3.02 | Improved |
| add object size clipping (ablation) | DemoFunGrasp · state-based sim | GSR (%): 81.67 -> 77.04 | Hurt |
| add close reward (ablation) | DemoFunGrasp · state-based sim | SA (%): 90.58 -> 94.74 | Improved |
| add qpos style reward (ablation) | DemoFunGrasp · state-based sim | GSR (%): 74.38 -> 77.04 | Improved |
| condition on affordance and style vs DemoGrasp | DemoFunGrasp · sim, seen categories | SAD (cm): 6.27 -> 3.02 | Improved |
| style/affordance conditioning vs UniDexGrasp | DemoFunGrasp · sim, unseen categories | Style Diversity: 1.00 -> 1.44 | Improved |
| style/affordance conditioning vs UniDexGrasp | DemoFunGrasp · sim, seen categories | GSR (%): 74.3 -> 76.26 | No clear effect |
| use VLM-predicted instead of human affordance | DemoFunGrasp · real world, 9 daily objects | GSR (%): 71 -> 64 | Hurt |
| select best style per object | DemoFunGrasp · sim, unseen categories | Style Diversity: 1.44 -> 1.36 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DemoGrasp — DemoGrasp: Universal Dexterous Grasping from a Single Demonstration (ICLR 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
- Cross-Embodiment Dexterous Grasping with Reinforcement Learning (ICLR 2025 · similarity 0.90)
- DexFuncGrasp: A Robotic Dexterous Functional Grasp Dataset Constructed from a Cost-Effective Real-Simulation Annotation System (AAAI 2024 · similarity 0.89)
- Scaffolding Dexterous Manipulation with Vision-Language Models (NeurIPS 2025 · similarity 0.89)
- UniGraspTransformer: Simplified Policy Distillation for Scalable Dexterous Robotic Grasping (CVPR 2025 · similarity 0.88)
- Efficient Residual Learning with Mixture-of-Experts for Universal Dexterous Grasping (ICLR 2025 · similarity 0.87)
- DexGrasp Anything: Towards Universal Robotic Dexterous Grasping with Physics Awareness (CVPR 2025 · 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.
