mlsanityData to 7 Oct 2026

Refining Compositional Diffusion for Reliable Long-Horizon Planning

arXiv 2605.03075 ↗First posted 4 May 2026cs.RO cs.AI cs.LGcited by 2

Not on the map. Why?

Abstract (arXiv)

Compositional diffusion planning generates long-horizon trajectories by stitching together overlapping short-horizon segments through score composition. However, when local plan distributions are multimodal, existing compositional methods suffer from mode-averaging, where averaging incompatible local modes leads to plans that are neither locally feasible nor globally coherent. We propose Refining Compositional Diffusion (RCD), a training-free guidance method that steers compositional sampling toward high-density, globally coherent plans. RCD leverages the self-reconstruction error of a pretrained diffusion model as a proxy for the log-density of composed plans, combined with an overlap consistency term that enforces consistency at segment boundaries. We show that the combined guidance concentrates sampling on high-density plans that mitigate mode-averaging. Experiments on challenging long-horizon tasks from OGBench, including locomotion, object manipulation, and pixel-based observations, demonstrate that RCD consistently outperforms existing methods.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 71 of its 86 references that are in the map. First posted on arXiv 4 May 2026.

Builds on

71 of 86 references.

References · 71

  1. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  2. 2020Denoising Diffusion Implicit Modelscited by 9,655
  3. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 8,964
  4. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,185
  5. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,730
  6. 2022Diffusion Posterior Sampling for General Noisy Inverse Problemscited by 1,375
  7. 2021Offline Reinforcement Learning with Implicit Q-Learningcited by 1,245
  8. 2022Planning with Diffusion for Flexible Behavior Synthesiscited by 1,023
  9. 2024Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusioncited by 634
  10. 2022Diffusion Policies as an Expressive Policy Class for Offline Reinforcement Learningcited by 543
  11. 2022Is Conditional Generative Modeling all you need for Decision-Making?cited by 504
  12. 2019Learning Latent Plans from Playcited by 310
38 more
  1. 2022Contrastive Learning as Goal-Conditioned Reinforcement Learningcited by 241
  2. 2023Reduce, Reuse, Recycle: Compositional Generation with Energy-Based Diffusion Models and MCMCcited by 216
  3. 2024OGBench: Benchmarking Offline Goal-Conditioned RLcited by 216
  4. 2025Flow Q-Learningcited by 162
  5. 2023Manifold Preserving Guided Diffusioncited by 153
  6. 2023AdaptDiffuser: Diffusion Models as Adaptive Self-evolving Plannerscited by 147
  7. 2025Diffusion-Based Planning for Autonomous Driving with Flexible Guidancecited by 138
  8. 2023HIQL: Offline Goal-Conditioned RL with Latent States as Actionscited by 133
  9. 2023Optimal Goal-Reaching Reinforcement Learning via Quasimetric Learningcited by 116
  10. 2023DiffCollage: Parallel Generation of Large Content with Diffusion Modelscited by 100
  11. 2024Foundation Policies with Hilbert Representationscited by 89
  12. 2023Generative Skill Chaining: Long-Horizon Skill Planning with Diffusion Modelscited by 87
  13. 2023MADiff: Offline Multi-agent Learning with Diffusion Modelscited by 73
  14. 2024Simple Hierarchical Planning with Diffusioncited by 71
  15. 2024Guidance with Spherical Gaussian Constraint for Conditional Diffusioncited by 69
  16. 2025Horizon Reduction Makes RL Scalablecited by 55
  17. 2025Diffusion Guidance Is a Controllable Policy Improvement Operatorcited by 51
  18. 2024DiffuserLite: Towards Real-time Diffusion Planningcited by 40
  19. 2024Policy-Guided Diffusioncited by 40
  20. 2024Potential Based Diffusion Motion Planningcited by 40
  21. 2024CleanDiffuser: An Easy-to-use Modularized Library for Diffusion Models in Decision Makingcited by 38
  22. 2025Generative Trajectory Stitching through Diffusion Compositioncited by 38
  23. 2024Multi-Robot Motion Planning with Diffusion Modelscited by 37
  24. 2025Simultaneous Multi-Robot Motion Planning with Projected Diffusion Modelscited by 37
  25. 2025What Makes a Good Diffusion Planner for Decision Making?cited by 37
  26. 2024DiffStitch: Boosting Offline Reinforcement Learning with Diffusion-based Trajectory Stitchingcited by 35
  27. 2024Inference-Time Policy Steering through Human Interactionscited by 35
  28. 2025Monte Carlo Tree Diffusion for System 2 Planningcited by 33
  29. 2024Understanding and Improving Training-free Loss-based Diffusion Guidancecited by 31
  30. 2023Refining Diffusion Planner for Reliable Behavior Synthesis by Automatic Detection of Infeasible Planscited by 27
  31. 2024Closing the Gap between TD Learning and Supervised Learning -- A Generalisation Point of Viewcited by 26
  32. 2025Latent Diffusion Planning for Imitation Learningcited by 26
  33. 2024GTA: Generative Trajectory Augmentation with Guidance for Offline Reinforcement Learningcited by 22
  34. 2025Option-aware Temporally Abstracted Value for Offline Goal-Conditioned Reinforcement Learningcited by 22
  35. 2025DriftLite: Lightweight Drift Control for Inference-Time Scaling of Diffusion Modelscited by 21
  36. 2025Temporal Difference Flowscited by 20
  37. 2025floq: Training Critics via Flow-Matching for Scaling Compute in Value-Based RLcited by 20
  38. 2025Decoupled Q-Chunkingcited by 18

50 most cited of 71 All 71 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Spectral Integrated Gradients for Coarse-to-Fine Feature Attributioncited by 2
  2. 2026Diffusion Integrated Gradients: Controllable Path Generation for Flexible Feature Attributioncited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2026Coarse-to-Fine Compositional Diffusion for Long-Horizon Planning
  2. 2025Compositional Monte Carlo Tree Diffusion for Extendable Planning
  3. 2026Plan First, Diffuse Later: Extrinsic Graph Guidance for Long-Horizon Diffusion Planning
  4. 2025What Do You Need for Compositional Generalization in Diffusion Planning?
  5. 2025Model-Based Diffusion Sampling for Predictive Control in Offline Decision Making
  6. 2026Compositional Planning with Jumpy World Models
  7. 2025Cascaded Diffusion Models for Neural Motion Planning
  8. 2026Trajectory-Level Mode Guidance for Controllable Diffusion-Based Multi-Robot Motion Planning