Research map / Multi-agent systems and games / Human-AI and goal-conditioned
MyoDex: A Generalizable Prior for Dexterous Manipulation
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 |
|---|---|---|---|
| Use a multi-task MyoDex prior for fine-tuning instead of training task-specific experts from scratch | MyoHand policies · MyoDM out-of-domain tasks (43) | solved tasks: 51% (22/43) -> 86% (37/43) | Helped |
| Use a multi-task MyoDex prior for fine-tuning instead of task-specific experts | MyoHand policies · MyoDM out-of-domain tasks (43) | success: 0.69 ± 0.30 -> 0.89 ± 0.25 | Helped |
| Use a multi-task MyoDex prior for fine-tuning instead of a distilled student | MyoHand policies · MyoDM out-of-domain tasks (43) | solved tasks: 30% (13/43) -> 86% (37/43) | Helped |
| Use a multi-task MyoDex prior for fine-tuning instead of a distilled student | MyoHand policies · MyoDM out-of-domain tasks (43) | iterations to solve: 7.7k ± 1.9k -> 1.9k ± 2.1k | Helped |
| Use AdroitDex instead of PGDM | Adroit Hand · TCDM unseen tasks (34) | iteration steps to reach 74.5% success: 50M -> 10M | Helped |
| Train the MyoDex prior on 6 environments rather than 14 (ablation) | MyoDex · MyoDM transfer tasks (39) | success: 0.92 ± 0.21 -> 0.78 ± 0.32 | Hurt |
| Train the MyoDex prior on 18 environments rather than 14 (ablation) | MyoDex · MyoDM transfer tasks (39) | solved tasks: 95% (37/39) -> 87% (34/39) | Hurt |
| Extract the prior earlier, at 2.5k iterations rather than 12.5k (ablation) | MyoDex · MyoDM tasks (57) | solved tasks: 98% (56 / 57) | Mixed |
| Extract the prior later, at 37.5k iterations rather than 12.5k (ablation) | MyoDex · MyoDM tasks (57) | solved tasks: 61% (35 / 57) | Hurt |
| Pre-train on diverse manipulation tasks rather than homogeneous tasks (ablation) | MyoDex Alt prior · Unseen transfer tasks | transfer performance | Helped |
Closest papers by meaning
- ManipTrans: Efficient Dexterous Bimanual Manipulation Transfer via Residual Learning (CVPR 2025 · similarity 0.89)
- DexFlyWheel: A Scalable and Self-improving Data Generation Framework for Dexterous Manipulation (NeurIPS 2025 · similarity 0.87)
- DexHandDiff: Interaction-aware Diffusion Planning for Adaptive Dexterous Manipulation (CVPR 2025 · similarity 0.87)
- HandEdit: A Unified Benchmark for Egocentric Human-to-Robot Dexterous Hand Image Editing (NeurIPS 2026 · similarity 0.87)
- Scaffolding Dexterous Manipulation with Vision-Language Models (NeurIPS 2025 · similarity 0.86)
- DexTrack: Towards Generalizable Neural Tracking Control for Dexterous Manipulation from Human References (ICLR 2025 · similarity 0.86)
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.