Research map / Architectures and efficiency / Attention and transformers
DDP-WM: Disentangled Dynamics Prediction for Efficient World Models
Shicheng Yin, Kaixuan Yin, Weixing Chen, Yang Liu, Guanbin Li, Liang Lin
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 Low-Rank Correction Module (ablation) | DDP-WM · Push-T | MPC success rate (%): 62 -> 70 | Improved |
| add Low-Rank Correction Module (with MPC mask) (ablation) | DDP-WM · Push-T | MPC success rate (%): 70 -> 98 | Improved |
| add sparse MPC cost mask (ablation) | DDP-WM · Push-T | MPC success rate (%): 90 -> 98 | Improved |
| add sparse MPC cost mask (without LRM) (ablation) | Naive Sparse · Push-T | MPC success rate (%): 62 -> 70 | Improved |
| add LRM to naive sparse model, open-loop (ablation) | DDP-WM · Push-T 5-step open-loop | pixel error: 427 -> 361 | Improved |
| add LRM to naive sparse model, open-loop (ablation) | DDP-WM · Wall 5-step open-loop | pixel error: 15 -> 9 | Improved |
| add LRM to naive sparse model, open-loop (ablation) | DDP-WM · PointMaze 5-step open-loop | pixel error: 41 -> 36 | Improved |
| add high-precision 2x2 sub-region localization (no LRM) (ablation) | DDP-WM · Push-T 5-step open-loop | pixel error: 788 -> 427 | Improved |
| add high-precision localization (ablation) | DDP-WM · Push-T mask quality | IoU: 0.3446 -> 0.8935 | Improved |
| sparse disentangled prediction vs dense DINO-WM | DDP-WM · Push-T | MPC success rate (%): 90 -> 98 | Improved |
| sparse disentangled prediction vs dense DINO-WM | DDP-WM · Push-T | MPC decision time (s): 120 -> 16 | Improved |
| sparse disentangled prediction vs dense DINO-WM | DDP-WM · Push-T | FLOPs per step (G): 23 -> 2.5 | Improved |
Closest papers by meaning
- HaM-World: Soft-Hamiltonian World Models with Selective Memory for Planning (NeurIPS 2026 · similarity 0.87)
- World Action Verifier: Self-Improving World Models via Forward-Inverse Asymmetry (NeurIPS 2026 · similarity 0.87)
- EA-WM: Event-Aware Generative World Model with Structured Kinematic-to-Visual Action Fields (NeurIPS 2026 · similarity 0.87)
- ODEWorld: A Continuous Predictive Architecture via Physical-Time Flow (NeurIPS 2026 · similarity 0.87)
- DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning (ICML 2025 · similarity 0.87)
- World Motion Models: Flexible Sequence Modeling of SE(3) Trajectories (NeurIPS 2026 · 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.
