Research map / Reinforcement learning / Visual and representation RL
Structural Action Transformer for 3D Dexterous Manipulation
Xiaohan Lei, Min Wang, Bohong Weng, Wengang Zhou, Houqiang Li
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 structural-centric action representation instead of temporal-centric (ablation) | SAT · 11 simulation tasks | success rate: 0.64 -> 0.71 | Improved |
| add the joint embedding (Embodied Joint Codebook) (ablation) | SAT · 11 simulation tasks | success rate: 0.01 -> 0.71 | Improved |
| add the Functional Category embedding (ablation) | SAT · 11 simulation tasks | success rate: 0.02 -> 0.71 | Improved |
| add the Embodiment ID embedding (ablation) | SAT · 11 simulation tasks | success rate: 0.57 -> 0.71 | Improved |
| add the Rotation Axis embedding (ablation) | SAT · 11 simulation tasks | success rate: 0.62 -> 0.71 | Improved |
| add the global point cloud token (ablation) | SAT · 11 simulation tasks | success rate: 0.68 -> 0.71 | Improved |
| add the local point cloud tokens (ablation) | SAT · 11 simulation tasks | success rate: 0.69 -> 0.71 | Improved |
| add the causal mask (ablation) | SAT · 11 simulation tasks | success rate: 0.68 -> 0.71 | Improved |
| increase token dimension from 16 to 32 (ablation) | SAT · 11 simulation tasks | success rate: 0.66 -> 0.71 | Improved |
| increase token dimension from 64 to 256 (ablation) | SAT · 11 simulation tasks | success rate: 0.71 -> 0.70 | No clear effect |
| pre-train on human, robot and simulation data instead of human only (ablation) | SAT · 11 simulation tasks | success rate: 0.68 -> 0.71 | Improved |
| pre-train on human data instead of robot data (ablation) | SAT · 11 simulation tasks | success rate: 0.66 -> 0.68 | Improved |
Closest papers by meaning
- Dexterous Manipulation Transfer via Progressive Kinematic-Dynamic Alignment (AAAI 2026 · similarity 0.91)
- ManipTrans: Efficient Dexterous Bimanual Manipulation Transfer via Residual Learning (CVPR 2025 · similarity 0.90)
- Scaffolding Dexterous Manipulation with Vision-Language Models (NeurIPS 2025 · similarity 0.90)
- DexFlyWheel: A Scalable and Self-improving Data Generation Framework for Dexterous Manipulation (NeurIPS 2025 · similarity 0.90)
- Vision-Language-Action Pretraining from Large-Scale Human Videos (ICML 2026 · similarity 0.90)
- Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · similarity 0.89)
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.
