mlsanityData to 14 Oct 2026

RL with diffusion and flow models

346 papers6 this weekLow confidence in this grouping: the papers may belong to more than one topic, or the group may changeLow confidence in this grouping: the papers may belong to more than one topic, or the group may change · About

Keywords: diffusion, flow, policy, reinforcement, reinforcement learning

Key facts. RL with diffusion and flow models: 346 papers on arXiv since 2024 (data through 14 Oct 2026). 67 of its 346 mapped papers (19%) cite “Diffusion Policy Policy Optimization” (2024, arXiv 2409.00588). 16 new papers 30 Sep–14 Oct.

What’s new

30 Sep – 14 Oct 2026

16 new papers, 30 Sep – 14 Oct; 6 this week.

  1. This week · 6
  2. SAM3RL: Memory Control via Reinforcement Learning for Visual Object Tracking
  3. PrT-Flow: Probability-Tractable Flow Matching for Reinforcement Learning also fits: RL for robot locomotion and manipulation
  4. DR-APO: Decoupled-Representation Advantage-Guided Policy Optimization from Continuous Control to Embodied Navigation
  5. DEFlow: Deep Equilibrium Flow Maximum-Entropy Reinforcement Learning also fits: RL for robot locomotion and manipulation
  6. XAI-Guided Fourier Domain Generalisation for Pure-Simulation Sim-to-Real Deep Reinforcement Learning also fits: Deep RL algorithms and representations
  7. A Survey of Flow Matching in Reinforcement Learning
  8. Earlier · 10
  9. Drift, Then Spread: Learning One-Step Generative Policies for Online Reinforcement Learning also fits: RL theory and sample complexity
  10. Single-step Retrosynthesis via Reinforcement Learning from Forward Feedback also fits: RL training of language and agent models
  11. MORAD: Reinforcement Learning with Multi-Objective Rewards for RNA Inverse Design also fits: RL training of language and agent models
  12. Reinforcement Learning via Fixed-point Flow Matching
All 16
  1. Q-Learning with Scalar Adjoint Matching
  2. Safe Image Generation via Reinforcement Learning
  3. PWM: Personalized World Models with Online Reinforcement Learning
  4. Score-Calibrated Flow for Sampling from Unnormalized Densities with Applications to Generative Online Reinforcement Learning
  5. Reinforcement Learning on the Discrete Composition Channel of a Crystal Generator: Validated Gains and Reward Hacking
  6. VIGOR: Zero-Shot Visual Generalization via Latent-Space Consistency in Model-Based Reinforcement Learning also fits: Deep RL algorithms and representations

Foundations

Most cited by papers here

  1. Diffusion Policy Policy Optimization2024cited by 67 here
  2. Diffusion-based Reinforcement Learning via Q-weighted Variational Policy Optimization2024cited by 51 here
  3. Flow Q-Learning2025cited by 48 here
  4. Diffusion Actor-Critic with Entropy Regulator2024cited by 47 here
  5. Flow Matching Policy Gradients2025cited by 35 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2024Diffusion Policy Policy Optimizationcited by 67 here
  2. 2025Flow Q-Learningcited by 48 here
  3. 2025Flow-Based Policy for Online Reinforcement Learningcited by 19 here
  4. 2025One-Step Flow Policy Mirror Descentcited by 13 here
  5. 2026Q-learning with Adjoint Matchingcited by 15 here
19 more it builds on, not in this chain
  1. 2024Diffusion Actor-Critic with Entropy Regulatorcited by 47 here
  2. 2024Bigger, Regularized, Optimistic: scaling for compute and sample-efficient continuous controlcited by 12 here
  3. 2024Diffusion-based Reinforcement Learning via Q-weighted Variational Policy Optimizationcited by 51 here
  4. 2024Learning Multimodal Behaviors from Scratch with Diffusion Policy Gradientcited by 21 here
  5. 2024Understanding Reinforcement Learning-Based Fine-Tuning of Diffusion Models: A Tutorial and Reviewcited by 17 here
  6. 2024Gymnasium: A Standard Interface for Reinforcement Learning Environmentscited by 11 here
  7. 2025Efficient Online Reinforcement Learning for Diffusion Policycited by 30 here
  8. 2025DIME:Diffusion-Based Maximum Entropy Reinforcement Learningcited by 28 here
  9. 2025Online Reward-Weighted Fine-Tuning of Flow Matching with Wasserstein Regularizationcited by 15 here
  10. 2025Maximum Entropy Reinforcement Learning with Diffusion Policycited by 19 here
  11. 2025GenPO: Generative Diffusion Models Meet On-Policy Reinforcement Learningcited by 14 here
  12. 2025ReinFlow: Fine-tuning Flow Matching Policy with Online Reinforcement Learningcited by 30 here
  13. 2025Diffusion Guidance Is a Controllable Policy Improvement Operatorcited by 11 here
  14. 2025Steering Your Diffusion Policy with Latent Space Reinforcement Learningcited by 17 here
  15. 2025Reinforcement Learning with Action Chunkingcited by 15 here
  16. 2025EXPO: Stable Reinforcement Learning with Expressive Policiescited by 11 here
  17. 2025Flow Matching Policy Gradientscited by 35 here
  18. 2025Coefficients-Preserving Sampling for Reinforcement Learning with Flow Matchingcited by 11 here
  19. 2025SAC Flow: Sample-Efficient Reinforcement Learning of Flow-Based Policies via Velocity-Reparameterized Sequential Modelingcited by 17 here

The 24 most cited of 312.

Browse

30 of 346, ranked 6–35

  1. 6Efficient Online Reinforcement Learning for Diffusion Policy202530
  2. 7ReinFlow: Fine-tuning Flow Matching Policy with Online Reinforcement Learning202530
  3. 8DIME:Diffusion-Based Maximum Entropy Reinforcement Learning202528
  4. 9Learning Multimodal Behaviors from Scratch with Diffusion Policy Gradient202421
  5. 10Maximum Entropy Reinforcement Learning with Diffusion Policy202519
  6. 11Flow-Based Policy for Online Reinforcement Learning202519
  7. 12SAC Flow: Sample-Efficient Reinforcement Learning of Flow-Based Policies via Velocity-Reparameterized Sequential Modeling202517
  8. 13Understanding Reinforcement Learning-Based Fine-Tuning of Diffusion Models: A Tutorial and Review202417
  9. 14Q-learning with Adjoint Matching202615
  10. 15Online Reward-Weighted Fine-Tuning of Flow Matching with Wasserstein Regularization202515
All 30
  1. 16GenPO: Generative Diffusion Models Meet On-Policy Reinforcement Learning202514
  2. 17One-Step Flow Policy Mirror Descent202513
  3. 18Diffusion Guidance Is a Controllable Policy Improvement Operator202511
  4. 19Coefficients-Preserving Sampling for Reinforcement Learning with Flow Matching202511
  5. 20Diffusion for World Modeling: Visual Details Matter in Atari202410
  6. 21Maximum Entropy Reinforcement Learning via Energy-Based Normalizing Flow202410
  7. 22A Simple and Effective Reinforcement Learning Method for Text-to-Image Diffusion Fine-tuning20258
  8. 23Large-scale Reinforcement Learning for Diffusion Models20248
  9. 24Flow Policy Gradients for Robot Control20267
  10. 25Flow-Based Single-Step Completion for Efficient and Expressive Policy Learning20257
  11. 26Sampling from Energy-based Policies using Diffusion20247
  12. 27Score as Action: Fine-Tuning Diffusion Generative Models by Continuous-time Reinforcement Learning20256
  13. 28Feedback Efficient Online Fine-Tuning of Diffusion Models20246
  14. 29Reverse Flow Matching: A Unified Framework for Online Reinforcement Learning with Diffusion and Flow Policies20265
  15. 30Rethinking the Design Space of Reinforcement Learning for Diffusion Models: On the Importance of Likelihood Estimation Beyond Loss Design20265
  16. 31FLAC: Maximum Entropy RL via Kinetic Energy Regularized Bridge Matching20265
  17. 32Reinforcement Learning with Discrete Diffusion Policies for Combinatorial Action Spaces20255
  18. 33Data-regularized Reinforcement Learning for Diffusion Models at Scale20255
  19. 34Fine-Tuning Discrete Diffusion Models with Policy Gradient Methods20254
  20. 35Score-Based Diffusion Policy Compatible with Reinforcement Learning via Optimal Transport20254