AI Paper Atlas

Research map / Video / Video generation

AnyFlow: Any-Step Video Diffusion Model with On-Policy Flow Map Distillation

Yuchao Gu, Guian Fang, Yuxin Jiang, Weijia Mao, Song Han, Han Cai, Mike Zheng Shou

ECCV 2026 · Video › Video generation · approach: Efficient video diffusion models · 33 citations (OpenAlex)

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 changedWhereMeasuredEffect
flow map training instead of consistency ODE-init (ablation)causal 1.3B Wan · VBench T2V overall, 4 NFEoverall: 73.97 -> 80.48Improved
flow map training instead of flow matching (few-step) (ablation)bidirectional 1.3B Wan · VBench T2V overall, 4 NFEoverall: 74.64 -> 81.75Improved
flow map backward simulation instead of consistency backward simulation (ablation)AnyFlow 1.3B · VBench T2V overall, 32 NFEoverall: 82.96 -> 83.96 (bidirectional); 83.49 -> 83.96 (causal)Improved
add on-policy distillation with flow map backward simulation (ablation)flow map trained model · VBench T2V causal, 4 NFEoverall: 80.48 -> 83.54Improved
consistency distillation with more sampling steps (ablation)consistency ODE-init + consistency backward sim · VBench T2V 4 -> 32 NFEoverall: 82.96 -> 79.80Hurt
increase NFE from 4 to 32AnyFlow-FAR-Wan2.1-14B · VBench T2Voverall: 84.05 -> 84.41Improved
increase NFE from 4 to 32AnyFlow-Wan2.1-T2V-14B · VBench T2Voverall: 84.04 -> 84.10Improved
AnyFlow instead of rCM at 4 NFEWan2.1-T2V-14B · VBench T2Voverall: 83.73 -> 84.04Improved
interpolated timestep conditioning instead of zero-init conditioning (ablation)flow map training · visual quality / embedding normembedding norm, saturationImproved
time sampler Beta(2,1.5) instead of uniform weighting (ablation)flow map training · VBench overall, 32 NFEoverallImproved
flow map backward simulation instead of consistency backward simulation at 16 steps (ablation)on-policy distillation · training cost (causal)s/iter: 93.8 -> 53.1Improved
flow map backward simulation instead of consistency backward simulation at 4 steps (ablation)on-policy distillation · training cost (causal)s/iter: 45.9 -> 53.1Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy