Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Coarse-to-Fine Compositional Diffusion for Long-Horizon Planning

About

Diffusion models provide strong priors for generating structured data, but many tasks require outputs beyond the scale on which these models are typically trained. Compositional generation addresses this by composing overlapping local plans from a pretrained short-horizon prior into a long-horizon output. However, standard composition primarily enforces agreement between neighboring local plans, yielding local consistency without directly specifying the global structure of the full composition. As a result, locally compatible plans may still form an implausible route, task sequence, or temporal evolution. Existing methods improve global coherence by repeatedly propagating local consistency signals or by adding inference-time optimization, but these procedures become expensive as the number or dimensionality of local plans increases. We propose Coarse-to-Fine Compositional Diffusion (CoFi), an inference-time sampler that separates global structure formation from local detail refinement. CoFi first aligns local denoised estimates around a shared coarse structure, producing a global scaffold that captures the long-range task-level arrangement. It then diffuses this scaffold to an intermediate noise level and denoises it with the same pretrained local prior, restoring local fine structure while preserving the scaffold-induced global coherence. Across long-horizon robotic planning, panoramic image generation, and long video generation, CoFi not only improves both global coherence and local sample quality over prior compositional baselines, but also requires 2-8x fewer denoiser evaluations.

Byoungwoo Park, Utkarsh A. Mishra, Jaemoo Choi, Juho Lee, Yongxin Chen• 2026

Related benchmarks

TaskDatasetResultRank
Robotic PlanningOGBench PointMaze Giant 48 (stitch)
Success Rate96
16
Robotic PlanningOGBench AntMaze Giant 48 (stitch)
Success Rate85
16
Robotic PlanningOGBench Scene 48 (play)
Success Rate0.63
16
Goal-oriented planningOGBench PointMaze Medium Stitch v1
Success Rate100
12
Robotic PlanningOGBench AntMaze-Stitch Large
Success Rate88
8
Robotic PlanningOGBench AntMaze-Stitch Medium
Success Rate97
8
Robotic PlanningOGBench Pointmaze Stitch Large
Success Rate100
8
Long Video GenerationVBench 273-frame
Subject Consistency94.11
5
Panoramic Image GenerationPanoramic image generation evaluation 512 x 4608
LPIPS0.48
4
Showing 9 of 9 rows

Other info

Follow for update