VARD: Efficient and Dense Fine-Tuning for Diffusion Models with Value-based RL

TL;DR

VARD introduces a value-based reinforcement learning method that provides dense, differentiable supervision for fine-tuning diffusion models, resulting in more stable training and improved generation quality for complex tasks.

Contribution

It proposes a novel value function-based approach with KL regularization to enable efficient, stable, and reward-aligned fine-tuning of diffusion models.

Findings

Enhanced trajectory guidance in diffusion models

Improved training efficiency and stability

Extended RL applicability to complex, non-differentiable rewards

Abstract

Diffusion models have emerged as powerful generative tools across various domains, yet tailoring pre-trained models to exhibit specific desirable properties remains challenging. While reinforcement learning (RL) offers a promising solution,current methods struggle to simultaneously achieve stable, efficient fine-tuning and support non-differentiable rewards. Furthermore, their reliance on sparse rewards provides inadequate supervision during intermediate steps, often resulting in suboptimal generation quality. To address these limitations, dense and differentiable signals are required throughout the diffusion process. Hence, we propose VAlue-based Reinforced Diffusion (VARD): a novel approach that first learns a value function predicting expection of rewards from intermediate states, and subsequently uses this value function with KL regularization to provide dense supervision throughout the generation process. Our method maintains proximity to the pretrained model while enabling effective and stable training via backpropagation. Experimental results demonstrate that our approach facilitates better trajectory guidance, improves training efficiency and extends the applicability of RL to diffusion models optimized for complex, non-differentiable reward functions.