Research map / Probabilistic and scientific modeling / Variational and generative sampling
Stable Mean Flow: Lyapunov-Inspired One-Step Flow Matching
Guangxun Zhang, Mason Haberle, Davi Geiger
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 Lyapunov-inspired non-expansivity stability loss to Mean Flow | Mean Flow · CIFAR-10 (NFE=1) | FID: 2.92 -> 2.86 | Improved |
| add stability loss on ImageNet XL/2 | MeanFlow-XL/2 · ImageNet (epoch 240) | FID: 3.43 -> 3.37 | No clear effect |
| add stability loss and evaluate with 5 sampling steps | Mean Flow · CIFAR-10 (NFE=5) | FID | No clear effect |
| add stability loss for faster early training convergence | Mean Flow · CIFAR-10 and checkerboard toy, early training | convergence speed (qualitative) | Improved |
| add stability loss to avoid JVP blow-up | Mean Flow · CIFAR-10 training | JVP stability: Mean Flow collapses around 53,000 iterations | Improved |
| set stability weight mu to 0.1 versus none at radius 0.01 (ablation) | SMFM · CIFAR-10 sweep (500k iters) | FID: 86.73 -> 79.86 | Improved |
| increase stability weight mu from 0.1 to 1 at radius 0.01 (ablation) | SMFM · CIFAR-10 sweep (500k iters) | FID: 79.86 -> 253.67 | Hurt |
| increase perturbation radius from 0.01 to 0.02 at mu 0.1 (ablation) | SMFM · CIFAR-10 sweep (500k iters) | FID: 79.86 -> 95.17 | Hurt |
| decrease perturbation radius from 0.01 to 0.005 at mu 0.1 (ablation) | SMFM · CIFAR-10 sweep (500k iters) | FID: 79.86 -> 85.41 | Hurt |
Closest papers by meaning
- Mean Flows for One-step Generative Modeling (NeurIPS 2025 · similarity 0.88)
- Optimal Flow Matching: Learning Straight Trajectories in Just One Step (NeurIPS 2024 · similarity 0.88)
- Drift Flow Matching (NeurIPS 2026 · similarity 0.87)
- Align Your Flow: Scaling Continuous-Time Flow Map Distillation (NeurIPS 2025 · similarity 0.87)
- Beckmann Transport Models: From Autonomous Flows to One-Step Maps (NeurIPS 2026 · similarity 0.87)
- How to build a consistency model: Learning flow maps via self-distillation (NeurIPS 2025 · 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.
