Velocity Scaling in Flow Matching
Scaling a learned flow-matching velocity field $vθ$ by a gain $γ(t)$ was recently shown to greatly improve generation quality.
Key points
- Prior work argued that velocity fields trained with mean-squared error (MSE) systematically underestimate velocity magnitude and that scaling corrects this error.
- We show that MSE training does not create a velocity-magnitude deficit.
- We find instead that velocity scaling reduces population time lag: sampled states at model time $t$ resemble training states from an earlier time.
- Across architectures and model sizes, measuring population time lag and using it to select a gain greatly improves generation quality, reducing FID from 28.0 to 12.2 (estimated by linear interpolation between FID measurements at neighboring gains) on ImageNet-256 at NFE 25 without guidance.
Sources (1)
- [1]Velocity Scaling in Flow MatchingarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 07:25 PM
Scaling a learned flow-matching velocity field $v_θ$ by a gain $γ(t)$ was recently shown to greatly improve generation quality.
Prior work argued that velocity fields trained with mean-squared error (MSE) systematically underestimate velocity magnitude and that scaling corrects this error.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 6, 2026Consistent Distribution Matching for Data-Free Diffusion Distillation
- Oct 6, 2026Co-Evolving Paths and Flows via Path-Flow Alignment
- Oct 6, 2026Test-Time Adaptation of Quantized ViTs via Single-Pass Quantizer-Aligned Recalibration
- Oct 6, 2026Compact Robot Policies Need Fine-Grained Visual Representations
- Oct 6, 2026StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models
- Oct 1, 2026nvidia/PixelDiT2-ImageNet