Search papers, labs, and topics across Lattice.
The paper addresses the challenge of aligning denoising generative models with human preferences using online reinforcement learning (RL). It introduces Variational GRPO (V-GRPO), which leverages ELBO-based likelihood surrogates within the Group Relative Policy Optimization (GRPO) framework, enhanced with variance reduction and controlled gradient steps. V-GRPO achieves state-of-the-art performance in text-to-image synthesis with significant speedups compared to MDP-based methods, demonstrating that ELBO-based approaches can be both stable and efficient for RL fine-tuning of generative models.
ELBO-based reinforcement learning, previously dismissed for visual generation, can actually outperform MDP-based methods for aligning denoising generative models with human preferences.
Aligning denoising generative models with human preferences or verifiable rewards remains a key challenge. While policy-gradient online reinforcement learning (RL) offers a principled post-training framework, its direct application is hindered by the intractable likelihoods of these models. Prior work therefore either optimizes an induced Markov decision process (MDP) over sampling trajectories, which is stable but inefficient, or uses likelihood surrogates based on the diffusion evidence lower bound (ELBO), which have so far underperformed on visual generation. Our key insight is that the ELBO-based approach can, in fact, be made both stable and efficient. By reducing surrogate variance and controlling gradient steps, we show that this approach can beat MDP-based methods. To this end, we introduce Variational GRPO (V-GRPO), a method that integrates ELBO-based surrogates with the Group Relative Policy Optimization (GRPO) algorithm, alongside a set of simple yet essential techniques. Our method is easy to implement, aligns with pretraining objectives, and avoids the limitations of MDP-based methods. V-GRPO achieves state-of-the-art performance in text-to-image synthesis, while delivering a $2\times$ speedup over MixGRPO and a $3\times$ speedup over DiffusionNFT.
Stanford HAI