dFlowGRPO: Rate-Aware Policy Optimization for Discrete Flow Models

arXiv:2605.09291 · cs.LG, stat.AP · Submitted 2026-05-10 · Read on arXiv

cs.LG, stat.AP

Submitted: 2026-05-10

Updated: 2026-10-01

Code: https://github.com/WanZhengyan/dFlowGRPO

Terminology

• Reinforcement Learning — 3× in this paper · explained in 4 episodes
• GRPO — 2× in this paper · explained in 6 episodes
• Generative Model — 1× in this paper · explained in 2 episodes
• Generative Models — 1× in this paper · explained in 3 episodes

Sources

• LLaDA2.0: Scaling Up Diffusion Language Models to 100B
• LLaDA2.1: Speeding Up Text Diffusion via Token Editing
• Training Diffusion Models with Reinforcement Learning
• Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• Direct Language Model Alignment from Online AI Feedback
• Tulu 3: Pushing Frontiers in Open Language Model Post-Training
• Aligning Text-to-Image Models using Human Feedback
• SEED-Bench: Benchmarking Multimodal LLMs with Generative Comprehension
• LaViDa-R1: Advancing Reasoning for Unified Multimodal Diffusion Language Models
• Discrete Diffusion Models: Novel Analysis and New Sampler Guarantees
• Flow-GRPO: Training Flow Matching Models via Online RL
• Reinforcement Learning Meets Masked Generative Models: Mask-GRPO for Text-to-Image Generation
• FS-DFM: Fast and Accurate Long Text Generation with Few-Step Diffusion Language Models
• Principled RL for Diffusion LLMs Emerges from a Sequence-Level Perspective
• Proximal Policy Optimization Algorithms
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• d2: Improving Reasoning in Diffusion Language Models via Trajectory Likelihood Estimation
• FUDOKI: Discrete Flow-based Unified Understanding and Generation via Kinetic-Optimal Velocities
• Unified Reward Model for Multimodal Understanding and Generation

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks