Research map / Reinforcement learning / Visual and representation RL
DyWA: Dynamics-adaptive World Action Model for Generalizable Non-prehensile Manipulation
Jiangran Lyu, Ziming Li, Xuesong Shi, Chaoyi Xu, Yizhou Wang, He 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 world modeling to DAgger (ablation) | DAgger · Unknown State (1 view), Seen | success rate: 59.9 -> 61.6 | Improved |
| add dynamics adaptation (RMA) to DAgger (ablation) | DAgger · Unknown State (1 view), Seen | success rate: 59.9 -> 65.6 | Improved |
| add both world model and dynamics adaptation (no FiLM) (ablation) | DAgger · Unknown State (1 view), Seen | success rate: 59.9 -> 73.3 | Improved |
| add FiLM conditioning to RMA (ablation) | RMA · Unknown State (1 view), Seen | success rate: 65.6 -> 70.0 | Improved |
| add FiLM to world model plus adaptation (ablation) | DyWA · Unknown State (1 view), Seen | success rate: 73.3 -> 82.2 | Improved |
| add world model to adaptation plus FiLM (ablation) | DyWA · Unknown State (1 view), Seen | success rate: 70.0 -> 82.2 | Improved |
| use FiLM instead of cross attention (ablation) | DyWA · Unknown State (1 view), Seen | success rate: 70.1 -> 82.2 | Improved |
| use FiLM instead of MLP concatenation (ablation) | DyWA · Unknown State (1 view), Seen | success rate: 73.3 -> 82.2 | Improved |
| add FiLM to world model and adaptation (ablation) | DyWA · Unknown State (1 view), Unseen | success rate: 59.4 -> 75.0 | Improved |
| add dynamics adaptation to world model | World Model · Unknown State (1 view), Unseen | success rate: 59.4 -> 57.9 | No clear effect |
| add world model and FiLM to RL teacher (ablation) | Teacher policy · sim | success rate: 94.1 -> 93.5 | No clear effect |
| add dynamics adaptation (ablation) | DyWA · real-world friction mu4 | avg time (s): 124 -> 51 | Improved |
Later papers that compared with this method
1 later accepted paper reports DyWA as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- DexMove: Learning Tactile-Guided Non-Prehensile Manipulation with Dexterous Hands (ICLR 2026 · better on 9 of 9 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
- PointWorld: Scaling 3D World Models for In-The-Wild Robotic Manipulation (CVPR 2026 · similarity 0.88)
- PhysMani: Physics-principled 3D World Model for Dynamic Object Manipulation (ECCV 2026 · similarity 0.88)
- Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation (ICLR 2025 · similarity 0.87)
- DynaRend: Learning 3D Dynamics via Masked Future Rendering for Robotic Manipulation (NeurIPS 2025 · similarity 0.87)
- Generalizable Domain Adaptation for Sim-and-Real Policy Co-Training (NeurIPS 2025 · similarity 0.86)
- AffordDP: Generalizable Diffusion Policy with Transferable Affordance (CVPR 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.
