AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Your VAR Model is Secretly an Efficient and Explainable Generative Classifier

Yi-Chung Chen, David I. Inouye, Jing Gao

ICLR 2026 · Image generation › Fast diffusion sampling · approach: Distributional control of diffusion · 2 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
replace diffusion classifier with VAR-based A-VARC+DiT-based diffusion classifier · ImageNet-100accuracy / GFLOPs: 90.30 -> 89.32; 415056.0 -> 4649.4 GFLOPs (89x less)Improved
add likelihood smoothing (S=10) (ablation)VARC · ImageNet-100top-1 accuracy: 83.30 -> 88.26Improved
add CCA finetuning (ablation)VARC · ImageNet-100top-1 accuracy: 83.30 -> 88.68Improved
add CCA finetuning (ablation)VARC · ImageNet-Sketch / ImageNet-Atop-1 accuracy: IN-Sketch 36.0 -> 34.8; IN-A 10.3 -> 9.9Hurt
add smoothing and CCA together (ablation)VARC · ImageNet-100top-1 accuracy: 83.30 -> 89.72Improved
increase smoothing samples beyond 10-16 (ablation)A-VARC · ImageNet-100top-1 accuracy: 88.26 -> 88.42 (S=32)No clear effect
increase smoothing noise sigma to 1.0 (ablation)A-VARC · ImageNet-100top-1 accuracy: 88.26 -> 39.54Hurt
use VAR classifier for robustness under distribution shift (vs diffusion classifier)A-VARC+ · ImageNet-Sketchtop-1 accuracy: DC 53.7 vs A-VARC+ 34.0Hurt
use rectified-flow MeanFlow classifier instead of DiT diffusion classifierdiffusion classifier · ImageNet-100top-1 accuracy: 86.32 -> 50.30Hurt
increase VAR model size from d16 to d30VAR classifier · ImageNet-100top-1 accuracy: 83.30 -> 64.96Hurt
use generative classifiers for class-incremental learning (no rehearsal)A-VARC+ · first 10 ImageNet classesaverage accuracy: 41.2 -> 77.4Improved
apply CCA finetuning before token-wise mutual information explanation (ablation)A-VARC+ · ImageNet-100 insertion/deletion AUCinsertion AUC: 0.845 -> 0.944Improved

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