Research map / Probabilistic and scientific modeling / Variational and generative sampling
Source-Guided Flow Matching
Zifan Wang, Alice Harting, Matthieu Barreau, Michael M. Zavlanos, Karl Henrik Johansson
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 |
|---|---|---|---|
| modify source distribution (SGFM) instead of the vector field, with exact transport map | SGFM · theory | distribution recovery | Improved |
| use chi-square norm regularizer (D-Flow) instead of plain Gaussian prior regularizer in optimization sampling (ablation) | SGFM-OPT · Darcy flow validity | validity: 0.907 -> 0.474 | Improved |
| use HMC sampling instead of plain MAP optimization (ablation) | SGFM · Darcy flow validity | validity: 0.907 -> 0.591 | Improved |
| SGFM-OPT chi^2 vs vector-field guidance g-cov-A | SGFM · Darcy flow validity | validity: 0.993 -> 0.474 | Improved |
| use SGFM-OPT-4 (shell regularizer) vs g-covA/g-covG | SGFM · CelebA denoising | PSNR: 30.35 -> 31.60 | Improved |
| SGFM-OPT-4 vs PnP-flow on denoising | SGFM · CelebA denoising | PSNR: 32.14 -> 31.60 | Hurt |
| SGFM-OPT-4 vs PnP-flow on deblurring | SGFM · CelebA deblurring | PSNR: 38.74 -> 35.27 | Hurt |
| SGFM-OPT-5 vs PnP-flow on super-resolution | SGFM · CelebA super-resolution | PSNR: 31.33 -> 33.33 | Improved |
| use projected gradient shell constraint (SGFM-6) instead of regularizer variants (ablation) | SGFM-OPT · CelebA deblurring | PSNR: 35.27 -> 32.60 | Hurt |
| use optimization-based sampling instead of exact sampling on S-shaped example | SGFM / D-Flow · 2D S-shape | diversity | Hurt |
| increase importance-sampling particle count (ablation) | SGFM-IS · 2D toy | Wasserstein error | Improved |
| reduce NFEs with optimal vector field (ablation) | SGFM-IS · 2D toy | sample quality | No clear effect |
Closest papers by meaning
- On the Guidance of Flow Matching (ICML 2025 · similarity 0.92)
- Optimal Flow Matching: Learning Straight Trajectories in Just One Step (NeurIPS 2024 · similarity 0.90)
- Beckmann Transport Models: From Autonomous Flows to One-Step Maps (NeurIPS 2026 · similarity 0.89)
- Drift Flow Matching (NeurIPS 2026 · similarity 0.89)
- Training Free Guided Flow-Matching with Optimal Control (ICLR 2025 · similarity 0.89)
- Self-Corrected Flow Distillation for Consistent One-Step and Few-Step Image Generation (AAAI 2025 · similarity 0.89)
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.
