Research map / Reinforcement learning / RL from feedback
Principled RL for Flow Matching Emerges from the Chunk-level Policy Optimization
Yifu Luo, Haoyuan Sun, Xinhao Hu, Penghui Du, Keyu Fan, Bo Li, SiNan Du, Xu Wan, Zhiyu Chen, Bo Xia, Yongzhe Chang, Kai Wu and 3 more
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| optimize at chunk level instead of step level (GCPO vs Dance-GRPO) | GRPO on FLUX.1 Dev · HPDv2.1 test | HPSv3: 15.080 -> 15.236 | Improved |
| add weighted sampling of chunks by relative L1 distance (ablation) | GCPO · HPDv2.1 test | HPSv3: 15.236 -> 15.373 | Improved |
| add weighted sampling of chunks (ablation) | GCPO · GenEval | overall: 0.69 -> 0.67 | Hurt |
| add weighted sampling of chunks (ablation) | GCPO · DPG | overall: 86.60 -> 85.14 | Hurt |
| use temporal-dynamics adaptive chunking instead of fixed chunk size (ablation) | GCPO · HPDv2.1 test | HPSv3: 15.173 -> 15.236 | Improved |
| fixed chunking [4, 4, 4, 4] versus step-level GRPO (ablation) | GCPO (fixed) · HPDv2.1 test | HPSv3: 15.080 -> 15.078 | No clear effect |
| replace outcome reward with a process reward (ablation) | Dance-GRPO · HPDv2.1 test | HPSv2: 0.365 -> 0.348 | Hurt |
| switch reward model from HPSv3 to PickScore (ablation) | GCPO · HPDv2.1 test | HPSv3: 14.612 -> 14.810 (vs Dance-GRPO) | Improved |
| use L2 distance instead of relative L1 to define chunks (ablation) | GCPO · HPDv2.1 test | HPSv3: 15.236 -> 15.268 | No clear effect |
| use cosine distance instead of relative L1 to define chunks (ablation) | GCPO · HPDv2.1 test | HPSv3: 15.236 -> 15.220 | No clear effect |
| treat the whole trajectory as a single chunk (no temporal dynamics) (ablation) | GCPO · HPDv2.1 test | HPSv3: 15.142 vs 15.236 (fixed [16] vs adaptive) | Hurt |
| train on high-noise chunks only (ablation) | GCPO · HPDv2.1 test | HPSv3 improvement | Mixed |
Closest papers by meaning
- Flow-GRPO: Training Flow Matching Models via Online RL (NeurIPS 2025 · similarity 0.90)
- Flow-DPPO: Divergence Proximal Policy Optimization for Flow Matching Models (NeurIPS 2026 · similarity 0.88)
- On the Guidance of Flow Matching (ICML 2025 · similarity 0.87)
- Reinforcement Learning Meets Masked Generative Models: Mask-GRPO for Text-to-Image Generation (NeurIPS 2025 · similarity 0.87)
- PRECISE: SDE-Consistent Stochastic Sampling for RL Post-Training of Flow-Matching Models (NeurIPS 2026 · similarity 0.87)
- TMPO: Trajectory Matching Policy Optimization for Diverse and Efficient Diffusion Alignment (NeurIPS 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
