Research map / Probabilistic and scientific modeling / Variational and generative sampling
Flow Matching with Semidiscrete Couplings
Alireza Mousavi-Hosseini, Stephen Y. Zhang, Michal Klein, marco cuturi
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 |
|---|---|---|---|
| replace independent coupling with semidiscrete OT coupling | flow matching · ImageNet-64 unconditional, Euler 4 | FID: 79.95 -> 45.62 | Improved |
| replace independent coupling with semidiscrete OT coupling | flow matching · PetFace-64 unconditional, Euler 4 | FID: 56.53 -> 20.54 | Improved |
| replace independent coupling with semidiscrete OT coupling | flow matching · ImageNet-64 class-conditional, Dopri5 | FID: 3.91 -> 3.63 | Improved |
| replace independent coupling with semidiscrete OT coupling | flow matching · PetFace-64 unconditional, Dopri5 | FID: 1.26 -> 1.26 | No clear effect |
| use entropic regularization eps>0 in SD-OT (ablation) | SD-FM · ImageNet-64 / PetFace | FID: 45.62 vs 45.90 (Euler 4, eps 0 vs 0.1) | No clear effect |
| use a more converged SD-OT potential (lower chi2) (ablation) | SD-FM · ImageNet-32 | FID and curvature | Improved |
| apply SD-OT coupling to MeanFlow training | MeanFlow · ImageNet-256 latent | FID | Improved |
| increase guidance correction weight r (ablation) | SD-FM with CFG correction · ImageNet-32 | precision/recall | Mixed |
Closest papers by meaning
- On Fitting Flow Models with Large Sinkhorn Couplings (NeurIPS 2026 · similarity 0.90)
- Optimal Flow Matching: Learning Straight Trajectories in Just One Step (NeurIPS 2024 · similarity 0.87)
- Elucidating Flow Matching ODE Dynamics via Data Geometry and Denoisers (ICML 2025 · similarity 0.85)
- PrismFlow: Residual Dynamics for Flow Matching in Time-Series Generation (NeurIPS 2026 · similarity 0.85)
- Switched Flow Matching: Eliminating Singularities via Switching ODEs (ICML 2024 · similarity 0.85)
- Coreset-Induced Conditional Velocity Flow Matching (NeurIPS 2026 · similarity 0.85)
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.
