AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

EVAR: Edge Visual Autoregressive Models via Principled Pruning

Zefang Wang, Ying Li, Yanyu Li, Mingluo Su, Simin Xu, Guanzhong Tian, Huan Wang

ECCV 2026 · Image generation › Autoregressive and latent generation · approach: Next-scale and token autoregressive images · 0 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
add pyramid-aware Hessian accumulation instead of direct concatenation of multi-scale tokens (ablation)EVAR structured OBS pruning · VAR-d16, 40% sparsity, training-freeFID: 69.21 -> 58.21Improved
calibrate with tokens from all scales instead of a single scale (ablation)EVAR structured OBS pruning · VAR-d16, 40% sparsity, training-freeFID: 86.51 -> 69.21Improved
add post-pruning fine-tuning (ablation)EVAR pruned VAR-d16 (40% sparsity) · ImageNet 256FID: 58.21 -> 4.25Improved
add PSAD instead of vanilla KD (ablation)EVAR pruned VAR-d16 (40% sparsity) · ImageNet 256FID: 4.26 -> 3.91Improved
weight distillation toward later scales with linear increase (ablation)EVAR distillation · VAR-d16, 40% sparsityFID: 4.26 -> 5.45Hurt
weight distillation toward later scales with exponential increase (ablation)EVAR distillation · VAR-d16, 40% sparsityFID: 4.26 -> 5.82Hurt
use naive exponential decay scale weighting instead of vanilla KD (ablation)EVAR distillation · VAR-d16, 40% sparsityFID: 4.26 -> 4.36Hurt
EVAR pyramid-aware OBS versus SlimGPT OBS with flat calibrationVAR-d16 pruning · 20% sparsity, training-freeFID: 56.87 -> 7.85Improved
EVAR versus best baseline (Taylor) after 1 epoch fine-tuningVAR-d16 pruning · 40% sparsity, 1-epochFID: 8.35 -> 4.86Improved
prune VAR-d16 from 20% to 40% sparsity with EVAREVAR VAR-d16 · ImageNet 256FID: 3.67 -> 3.91Mixed
pin Core ML to CPU+GPU instead of All compute units (ablation)EVAR on iPad Pro (M4) · single-image latencylatency (ms): 494 -> 277Improved
use CPU+NPU compute units instead of CPU+GPU (ablation)EVAR on iPad Pro (M4) · single-image latencylatency (ms): 277 -> 1090Hurt

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