AnisoFlow learns anisotropic prediction targets for flow models, boosting FID without touching architecture

StanfordUniversity · hf · 2026-11-19

Stanford's AnisoFlow learns an anisotropic, learnable prediction target for flow-based generative models, improving FID and image quality without altering the architecture or sampler. Promo figures describe a tiny pixel-space model at just 918M parameters, with FP8 blockwise comparisons showing visibly better detail and color than the x-pred baseline.

Original post →

More from Multimodal

Multimodal channel →