AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Progressive Supernet Training for Efficient Visual Autoregressive Modeling

Xiaoyue Chen, Yuling Shi, Kaiyuan Li, Huandong Wang, Yong Li, Xiaodong Gu, Xinlei Chen, Mingbao Lin

CVPR 2026 · Image generation › Autoregressive and latent generation · approach: Next-scale and token autoregressive images · 7 citations (OpenAlex)

Read the paper

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
use full-depth network on early scales and shallow subnet only on late scales (r7-r10) (ablation)pretrained VAR-d30, training-free · ImageNet 256x256FID: early subnet 12.91 -> late subnet 5.42Improved
jointly train subnets with shared weights (supernet training) instead of training-free subnet (ablation)VARiant d=2 subnet · ImageNet 256x256FID: 132 -> 2.97Improved
jointly train subnets with shared weights instead of training-free subnet (ablation)VARiant d=16 subnet · ImageNet 256x256FID: 5.42 -> 2.05Improved
joint supernet training with a shallower subnet (ablation)full network VAR-d30 · ImageNet 256x256full-network FID: 1.95 -> 2.14 (d=2); 1.95 -> 1.96 (d=16)Hurt
add bridge zone (full depth on scales r1-r6) so skipped layers still get gradients (ablation)VARiant d16 · ImageNet 256x256subnet FID: 9.44 -> 2.05Improved
add bridge zone (ablation)VARiant d16 · ImageNet 256x256full-network FID: 3.12 -> 1.96Improved
replace fixed-ratio subnet sampling with three-phase progressive sampling schedule (ablation)VARiant supernet training · ImageNet 256x256FID (full and subnet jointly): fixed p=0.1: 1.96/2.68; p=1.0: 2.32/2.15; progressive: 1.96/2.05 (full/subnet d16)Improved
train subnet only (sampling ratio 1.0) throughout (ablation)VARiant supernet training · ImageNet 256x256full-network FID: 1.96 -> 2.32Hurt
increase early-scale count N from 6 to 7 at D=16 (ablation)VARiant d16 · ImageNet 256x256FID: FID improves 0.05 with 14% memory increaseImproved
increase N from 7 to 10 at D=16 (ablation)VARiant d16 · ImageNet 256x256FID: FID improvements of 0.02, 0.01, 0.01No clear effect
reduce subnet depth from 30 to 16 layers (ablation)VARiant · ImageNet 256x256, batch 64FID and KV cache: FID 1.95 -> 2.05; KV cache reduced 40.4%Mixed
use single-model VARiant d4 instead of dual-model CoDeVAR acceleration · ImageNet 256x256FID and memory: FID 2.27 -> 2.28; memory 19943MB -> 19582MBMixed

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