TMPO: Trajectory Matching Policy Optimization for Diverse and Efficient Diffusion Alignment

TL;DR

TMPO introduces a trajectory-level reward distribution matching approach with a Softmax-TB objective, enhancing diversity and efficiency in diffusion model alignment tasks compared to reward maximization methods.

Contribution

The paper proposes TMPO, a novel trajectory matching policy optimization method that improves diversity and reduces reward hacking in diffusion model alignment.

Findings

TMPO improves generative diversity by 9.1% over state-of-the-art methods.

TMPO achieves a better trade-off between reward and diversity.

Dynamic Stochastic Tree Sampling reduces training time while maintaining performance.

Abstract

Reinforcement learning (RL) has shown extraordinary potential in aligning diffusion models to downstream tasks, yet most of them still suffer from significant reward hacking, which degrades generative diversity and quality by inducing visual mode collapse and amplifying unreliable rewards. We identify the root cause as the mode-seeking nature of these methods, which maximize expected reward without effectively constraining probability distribution over acceptable trajectories, causing concentration on a few high-reward paths. In contrast, we propose Trajectory Matching Policy Optimization (TMPO), which replaces scalar reward maximization with trajectory-level reward distribution matching. Specifically, TMPO introduces a Softmax Trajectory Balance (Softmax-TB) objective to match the policy probabilities of K trajectories to a reward-induced Boltzmann distribution. We prove that this objective inherits the mode-covering property of forward KL divergence, preserving coverage over all acceptable trajectories while optimizing reward. To further reduce multi-trajectory training time on large-scale flow-matching models, TMPO incorporates Dynamic Stochastic Tree Sampling, where trajectories share denoising prefixes and branch at dynamically scheduled steps, reducing redundant computation while improving training effectiveness. Extensive results across diverse alignment tasks such as human preference, compositional generation and text rendering show that TMPO improves generative diversity over state-of-the-art methods by 9.1%, and achieves competitive performance in all downstream and efficiency metrics, attaining the optimal trade-off between reward and diversity.