Research map / Image generation / Text-to-image diffusion
Designing Instance-Level Sampling Schedules via REINFORCE with James-Stein Shrinkage
Peiyu Yu, Suraj Kothawade, Sirui Xie, Ying Nian Wu, Hongliang Fei
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 |
|---|---|---|---|
| learn instance-level schedule with JS baseline | Flux-Dev · HPDv2 (5 steps) | HPSv2: 23.73 -> 29.21 | Improved |
| use James-Stein baseline instead of RLOO (ablation) | Flux-Dev scheduler · HPDv2 (5 steps) | HPSv2: 26.48 -> 29.21 | Improved |
| use James-Stein baseline instead of cross-context RLOO (ablation) | Flux-Dev scheduler · HPDv2 (5 steps) | HPSv2: 26.92 -> 29.21 | Improved |
| use single-pass JS policy instead of TPDM-style PPO (ablation) | Flux-Dev scheduler · HPDv2 (5 steps) | HPSv2: 15.73 -> 29.21 | Improved |
| add learned JS schedule | SD3.5-M · Text rendering (40 steps) | OCR-R: 53.34 -> 58.19 | Improved |
| add learned JS schedule | SD3.5-M · Text rendering (40 steps) | HPSv2: 27.23 -> 27.07 | Hurt |
| add learned JS schedule | Flux-Dev · Text rendering (40 steps) | OCR-R: 49.77 -> 58.58 | Improved |
| add learned JS schedule | Flux-Dev · Text rendering (40 steps) | OCR-A: 20.06 -> 21.30 | Improved |
| add learned JS schedule | SD3.5-M · GenEval | Overall: 0.62 -> 0.68 | Improved |
| add learned JS schedule | Flux-Dev · GenEval | Counting: 0.58 -> 0.77 | Improved |
| add learned JS schedule | Flux-Dev · GenEval | Overall: 0.64 -> 0.70 | Improved |
| use JS baseline instead of RLOO in synthetic variance study (K=8,B=... L=4) (ablation) | REINFORCE baseline · Synthetic gradient variance | gradient variance: 1.055 -> 0.612 | Improved |
Closest papers by meaning
- Towards Better Alignment: Training Diffusion Models with Reinforcement Learning Against Sparse Rewards (CVPR 2025 · similarity 0.89)
- GARDO: Reinforcing Diffusion Models without Reward Hacking (NeurIPS 2026 · similarity 0.87)
- Fast constrained sampling in pre-trained diffusion models (NeurIPS 2025 · similarity 0.87)
- David and Goliath: Small One-step Model Beats Large Diffusion with Score Post-training (ICML 2025 · similarity 0.87)
- Diffusion Tree Sampling: Scalable inference‑time alignment of diffusion models (NeurIPS 2025 · similarity 0.87)
- $\Psi$-Sampler: Initial Particle Sampling for SMC-Based Inference-Time Reward Alignment in Score Models (NeurIPS 2025 · similarity 0.87)
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.
