Designing Reinforcement Learning for Diffusion Models: A Unified Path-Space View

• URL: http://arxiv.org/abs/2608.14430v1
• Date: Fri, 14 Aug 2026 16:11:55 GMT
• Title: Designing Reinforcement Learning for Diffusion Models: A Unified Path-Space View
• Abstract summary: Reinforcement learning post-training provides a way to align diffusion models with human preferences and task-specific rewards.<n>Current RL algorithms for diffusion models remain fragmented.<n>This paper shows that these seemingly different losses arise from a single path-space principle.
• Score: 17.57225063689823
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Reinforcement learning (RL) post-training provides a direct way to align diffusion models with human preferences and task-specific rewards. However, current RL algorithms for diffusion models remain fragmented: reverse-trajectory methods rely on discretized likelihood ratios, whereas forward-matching methods train on reward-labeled noising versions of the rollout samples. This paper shows that these seemingly different losses arise from a single path-space principle. Starting from the regularized diffusion-RL objective, we use importance sampling between sampling SDEs to obtain an explicit policy-gradient estimator on trajectory space. The estimator contains the stochastic Itô integral underlying Flow-GRPO-type updates; we derive an equivalent variance-reduced value-gradient form that recovers the forward-matching structure of AWM and DiffusionNFT. This identifies the empirical gap between these method families as a variance-reduction effect rather than a difference in RL principle. The derivation yields a unified design space organized by value-gradient estimation, weight functions, and sampling choices. Within this space, we propose a multi-sample KDE value-gradient estimator that reuses rollout groups, together with scale-bounded weight families that retain stable existing recipes while excluding singular ones. Experiments on SD3.5-M and Qwen-Image models validate the variance-reduction explanation and show that the resulting recipe improves over prior diffusion-RL baselines.

Related papers

• Inference-Time Alignment of Diffusion Models via Trust-Region Iterative Twisted Sequential Monte Carlo [15.105260031413698]
  We study inference-time alignment for diffusion-based generative models.<n>We propose a trust-region framework for learning twisting functions in SMC-based inference-time alignment.
  arXiv  Detail & Related papers  (2026-05-24T15:08:52Z)
• DiffusionOPD: A Unified Perspective of On-Policy Distillation in Diffusion Models [55.01951088768769]
  DiffusionOPD is a new multi-task training paradigm for diffusion models based on Online Policy Distillation (OPD)<n>We show that DiffusionOPD consistently surpasses both multi-reward RL and cascade RL baselines in training efficiency and final performance.
  arXiv  Detail & Related papers  (2026-05-14T16:49:09Z)
• Entropy-Based Dimension-Free Convergence and Loss-Adaptive Schedules for Diffusion Models [3.2091923314854416]
  Diffusion generative models synthesize samples by discretizing reverse-time dynamics driven by a learned score (or denoiser)<n>We develop an information-theoretic approach to dimension-free convergence that avoids geometric assumptions.<n>We also propose a Loss-Adaptive Schedule (LAS) for efficient discretization of reverse SDE which is lightweight and relies only on the training loss.
  arXiv  Detail & Related papers  (2026-01-29T16:28:21Z)
• G$^2$RPO: Granular GRPO for Precise Reward in Flow Models [74.21206048155669]
  We propose a novel Granular-GRPO (G$2$RPO) framework that achieves precise and comprehensive reward assessments of sampling directions.<n>We introduce a Multi-Granularity Advantage Integration module that aggregates advantages computed at multiple diffusion scales.<n>Our G$2$RPO significantly outperforms existing flow-based GRPO baselines.
  arXiv  Detail & Related papers  (2025-10-02T12:57:12Z)
• Training-Free Stein Diffusion Guidance: Posterior Correction for Sampling Beyond High-Density Regions [46.59494117137471]
  Training free diffusion guidance provides a flexible way to leverage off-the-shelf classifiers without additional training.<n>We introduce Stein Diffusion Guidance (SDG), a novel training-free framework grounded in a surrogate SOC objective.<n>Experiments on molecular low-density sampling tasks suggest that SDG consistently surpasses standard training-free guidance methods.
  arXiv  Detail & Related papers  (2025-07-07T21:14:27Z)
• DIVE: Inverting Conditional Diffusion Models for Discriminative Tasks [79.50756148780928]
  This paper studies the problem of leveraging pretrained diffusion models for performing discriminative tasks.<n>We extend the discriminative capability of pretrained frozen generative diffusion models from the classification task to the more complex object detection task, by "inverting" a pretrained layout-to-image diffusion model.
  arXiv  Detail & Related papers  (2025-04-24T05:13:27Z)
• Amortized Posterior Sampling with Diffusion Prior Distillation [55.03585818289934]
  Amortized Posterior Sampling is a novel variational inference approach for efficient posterior sampling in inverse problems.<n>Our method trains a conditional flow model to minimize the divergence between the variational distribution and the posterior distribution implicitly defined by the diffusion model.<n>Unlike existing methods, our approach is unsupervised, requires no paired training data, and is applicable to both Euclidean and non-Euclidean domains.
  arXiv  Detail & Related papers  (2024-07-25T09:53:12Z)
• Understanding Reinforcement Learning-Based Fine-Tuning of Diffusion Models: A Tutorial and Review [63.31328039424469]
  This tutorial provides a comprehensive survey of methods for fine-tuning diffusion models to optimize downstream reward functions. We explain the application of various RL algorithms, including PPO, differentiable optimization, reward-weighted MLE, value-weighted sampling, and path consistency learning.
  arXiv  Detail & Related papers  (2024-07-18T17:35:32Z)
• Improved off-policy training of diffusion samplers [93.66433483772055]
  We study the problem of training diffusion models to sample from a distribution with an unnormalized density or energy function.<n>We benchmark several diffusion-structured inference methods, including simulation-based variational approaches and off-policy methods.<n>Our results shed light on the relative advantages of existing algorithms while bringing into question some claims from past work.
  arXiv  Detail & Related papers  (2024-02-07T18:51:49Z)
• Towards Controllable Diffusion Models via Reward-Guided Exploration [15.857464051475294]
  We propose a novel framework that guides the training-phase of diffusion models via reinforcement learning (RL) RL enables calculating policy gradients via samples from a pay-off distribution proportional to exponential scaled rewards, rather than from policies themselves. Experiments on 3D shape and molecule generation tasks show significant improvements over existing conditional diffusion models.
  arXiv  Detail & Related papers  (2023-04-14T13:51:26Z)
• Reflected Diffusion Models [93.26107023470979]
  We present Reflected Diffusion Models, which reverse a reflected differential equation evolving on the support of the data. Our approach learns the score function through a generalized score matching loss and extends key components of standard diffusion models.
  arXiv  Detail & Related papers  (2023-04-10T17:54:38Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.