Inference-Time Scaling in Diffusion Models through Iterative Partial Refinement

TL;DR

The paper introduces Iterative Partial Refinement (IPR), a novel inference-time scaling method for diffusion models that improves sample consistency without external verifiers by re-noising and regenerating regions.

Contribution

IPR is a new inference-time scaling technique for sequential diffusion models that enhances global consistency without relying on external evaluators.

Findings

IPR increases valid solution rate on MNIST Sudoku from 55.8% to 75.0%.

IPR improves reasoning performance in diffusion models.

IPR operates without external verifiers in mixed-noise, sequential inference settings.

Abstract

Inference-time scaling has emerged as a major approach for improving reasoning capabilities, and has been increasingly applied to diffusion models. However, existing inference-time scaling methods for diffusion models typically rely on external verifiers or reward models to rank and select samples, limiting their scalability to settings where such evaluators are available and reliable. Moreover, while recent diffusion models perform sequential inference with region-wise, mixed-noise conditioning, inference-time scaling tailored to this setting remains relatively underexplored. We propose Iterative Partial Refinement (IPR), an inference-time scaling method for sequential diffusion that requires no external verifier. Starting from an already-generated sample, IPR re-noises a subset of regions and regenerates them conditioned on the remaining regions, enabling the model to revise earlier decisions under a richer context than was available during the initial generation. This iterative partial refinement produces more globally consistent samples without external verification. On reasoning tasks requiring global constraint satisfaction, IPR consistently improves performance: on MNIST Sudoku, the valid solution rate increases from 55.8% to 75.0%. These results show that iterative partial refinement alone can serve as an effective inference-time scaling strategy for diffusion models in sequential, mixed-noise settings. Code is available at: https://github.com/ahn-ml/IPR