Diffusion-Sharpening: Fine-tuning Diffusion Models with Denoising Trajectory Sharpening

TL;DR

Diffusion-Sharpening is a novel fine-tuning method for diffusion models that optimizes sampling trajectories using a path integral framework, improving efficiency and alignment without increasing inference costs.

Contribution

It introduces a trajectory-level fine-tuning approach that leverages reward feedback and amortizes inference costs, outperforming existing RL-based and trajectory optimization methods.

Findings

Faster convergence in training.

Superior inference efficiency without extra NFEs.

Outperforms existing methods in alignment and human preference metrics.

Abstract

We propose Diffusion-Sharpening, a fine-tuning approach that enhances downstream alignment by optimizing sampling trajectories. Existing RL-based fine-tuning methods focus on single training timesteps and neglect trajectory-level alignment, while recent sampling trajectory optimization methods incur significant inference NFE costs. Diffusion-Sharpening overcomes this by using a path integral framework to select optimal trajectories during training, leveraging reward feedback, and amortizing inference costs. Our method demonstrates superior training efficiency with faster convergence, and best inference efficiency without requiring additional NFEs. Extensive experiments show that Diffusion-Sharpening outperforms RL-based fine-tuning methods (e.g., Diffusion-DPO) and sampling trajectory optimization methods (e.g., Inference Scaling) across diverse metrics including text alignment, compositional capabilities, and human preferences, offering a scalable and efficient solution for future diffusion model fine-tuning. Code: https://github.com/Gen-Verse/Diffusion-Sharpening