Research map / Vision-language and multimodal / Robot action and navigation
AnyBimanual: Transferring Unimanual Policy for General Bimanual Manipulation
Guanxing Lu, Tengbo Yu, Haoyuan Deng, Season Si Chen, Yansong Tang, Ziwei Wang
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 skill manager (ablation) | PerAct on bimanual · RLBench2 | average success rate (%): 16.75 -> 25.67; +8.92% | Improved |
| add visual aligner (ablation) | PerAct on bimanual · RLBench2 | average success rate (%): 16.75 -> 19.75; +3.00% | Improved |
| add both skill manager and visual aligner (ablation) | PerAct on bimanual · RLBench2 | average success rate (%): 14.67 -> 32.00 | Improved |
| add skill manager on Sync tasks (ablation) | PerAct on bimanual · RLBench2 Sync tasks | success rate (%): 9.50 -> 17.00 | Improved |
| add visual aligner on Generalized tasks (ablation) | PerAct on bimanual · RLBench2 Generalized tasks | success rate (%): 25.50 -> 44.00 | Improved |
| add AnyBimanual to PerAct-LF | PerAct-LF · RLBench2 | average success rate (%): 4.92 -> 8.50 | Improved |
| add AnyBimanual to RVT-LF | RVT-LF · RLBench2 | average success rate (%): 10.58 -> 14.75 | Improved |
| compare AnyBimanual against PerAct2 with 20 demonstrations | PerAct · RLBench2 | average success rate (%): +10.50% | Improved |
| add AnyBimanual over PerAct2 with more demos | PerAct · RLBench2 | average success rate (%): +17.33% | Improved |
| add AnyBimanual vs plain pretraining on press buttons | PerAct2 + Pretraining · RLBench2 press buttons (100 demos) | success rate (%): 40 -> 73 | Improved |
| add visual aligner on simple short-horizon tasks (ablation) | PerAct · RLBench2 lift ball | success rate (%) | Hurt |
| reduce real-world demonstrations from 30 to 5 per task (ablation) | AnyBimanual · Real robot (9 tasks) | average success rate (%): 84.62 -> 53.33 | Hurt |
Closest papers by meaning
- TwinVLA: Data-Efficient Bimanual Manipulation with Twin Single-Arm Vision-Language-Action Models (ICLR 2026 · similarity 0.90)
- RDT-1B: a Diffusion Foundation Model for Bimanual Manipulation (ICLR 2025 · similarity 0.89)
- ManipTrans: Efficient Dexterous Bimanual Manipulation Transfer via Residual Learning (CVPR 2025 · similarity 0.87)
- Forecasting Bimanual Object Manipulation Sequences from Unimanual Observations (AAAI 2024 · similarity 0.87)
- OAKINK2: A Dataset of Bimanual Hands-Object Manipulation in Complex Task Completion (CVPR 2024 · similarity 0.87)
- Exploring the Limits of Vision-Language-Action Manipulation in Cross-task Generalization (NeurIPS 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.
