Research map / Probabilistic and scientific modeling / Molecules and proteins
Energy-based Compositional Diffusion Planning
Tao Sun, Utkarsh Aashu Mishra, Jiaxin Lu, Danfei Xu, Iro Armeni
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 energy-based correction (ECD) to CompDiffuser | CompDiffuser · PointMaze Giant | success rate (%): 77 -> 84 | Improved |
| add energy-based correction (ECD) to CompDiffuser | CompDiffuser · AntMaze Large | success rate (%): 74 -> 82 | Improved |
| add energy-based correction (ECD) to CompDiffuser | CompDiffuser · HumanoidMaze Giant | success rate (%): 42 -> 49 | Improved |
| add energy-based correction (ECD) to CompDiffuser | CompDiffuser · AntMaze Giant (15D) | success rate (%): 41 -> 51 | Improved |
| add energy-based correction (ECD) to CompDiffuser | CompDiffuser · AntMaze Large (15D) | success rate (%): 83 -> 89 | Improved |
| add energy-based correction (ECD) to CompDiffuser | CompDiffuser · PointMaze Medium and Large | success rate (%): 100 -> 100 | No clear effect |
| add energy-based correction (ECD) to CompDiffuser | CompDiffuser · HumanoidMaze Medium | success rate (%): 90 -> 92 | Mixed |
| add energy-based correction (ECD) | CDGS · AntSoccer Medium (17D) | success rate (%): 9 -> 20 | Improved |
| use exact boundary reaction term instead of approximation (ablation) | ECD · PointMaze Giant | success rate (%): 84 -> 86 | Improved |
| use exact boundary reaction term instead of approximation (ablation) | ECD · AntMaze Giant | success rate (%): 82 -> 82 | No clear effect |
| use exact boundary reaction term instead of approximation (ablation) | ECD · PointMaze Giant | runtime (s/episode): 8 -> 17 | Hurt |
| increase CDGS resampling rounds from 4x to 8x (ablation) | CDGS · AntMaze Giant | runtime (s): 46 -> 168 | Hurt |
Closest papers by meaning
- Generative Trajectory Stitching through Diffusion Composition (NeurIPS 2025 · similarity 0.90)
- Coarse-to-Fine Compositional Diffusion for Long-Horizon Planning (NeurIPS 2026 · similarity 0.88)
- Compositional Monte Carlo Tree Diffusion for Extendable Planning (NeurIPS 2025 · similarity 0.88)
- Model-based Diffusion for Trajectory Optimization (NeurIPS 2024 · similarity 0.87)
- DiffuserLite: Towards Real-time Diffusion Planning (NeurIPS 2024 · similarity 0.86)
- State-Covering Trajectory Stitching for Diffusion Planners (NeurIPS 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.
