Research map / Image generation / Fast diffusion sampling
Discrete Guidance Matching: Exact Guidance for Discrete Flow Matching
Zhengyan Wan, Yidong Ouyang, Liyan Xie, Fang Fang, Hongyuan Zha, Guang Cheng
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 |
|---|---|---|---|
| add posterior-based reward guidance | FUDOKI · GenEval overall | score: 0.77 -> 0.78 | Improved |
| add posterior-based reward guidance | FUDOKI · GenEval Color Attri. | score: 0.67 -> 0.77 | Improved |
| add posterior-based reward guidance | FUDOKI · GenEval Single Obj. | score: 0.96 -> 0.94 | Hurt |
| add posterior-based reward guidance | FUDOKI · GenEval Counting | score: 0.56 -> 0.53 | Hurt |
| add posterior-based reward guidance | FUDOKI · GenEval sub-tasks | score: improves 4 of 6 sub-tasks | Mixed |
| add posterior-based guidance | FUDOKI 1.5B · POPE | accuracy: 86.1 -> 86.8 | Improved |
| add posterior-based guidance | FUDOKI 1.5B · MME-P | score: 1485.4 -> 1492.7 | Improved |
| add posterior-based guidance | FUDOKI 1.5B · MMMU | accuracy: 34.3 -> 35.4 | Improved |
| add posterior-based guidance | FUDOKI 1.5B · MMBench | score: 73.9 -> 74.2 | Improved |
| use posterior-based instead of rate-based guidance | energy-guided sampling · 2-D simulation | sampling speed: 1.6x faster | Improved |
| use exact guidance instead of first-order predictor guidance | energy-guided sampling · 2-D simulation (gamma=10,20) | distribution match | Improved |
Closest papers by meaning
- Discrete Flow Matching (NeurIPS 2024 · similarity 0.91)
- On the Guidance of Flow Matching (ICML 2025 · similarity 0.89)
- Flow Matching with General Discrete Paths: A Kinetic-Optimal Perspective (ICLR 2025 · similarity 0.88)
- Transition Matching: Scalable and Flexible Generative Modeling (NeurIPS 2025 · similarity 0.87)
- Continuous-Time Distribution Matching for Few-Step Diffusion Distillation (NeurIPS 2026 · similarity 0.87)
- Trajectory as the Teacher: Few-Step Discrete Flow Matching via Energy-Navigated Distillation (NeurIPS 2026 · 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.
