AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

MARché: Fast Masked Autoregressive Image Generation with Cache-Aware Attention

Chaoyi Jiang, Sungwoo Kim, Lei Gao, Hossein Zarch, Won Woo Ro, Murali Annavaram

ECCV 2026 · Image generation › Autoregressive and latent generation · approach: Next-scale and token autoregressive images · 1 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 caching tokens to active set (ablation)MARché · ImageNet 256 ablationFID: 2.70 -> 2.56Improved
add generating tokens to active set (ablation)MARché · ImageNet 256 ablationFID: 504.82 -> 2.56Improved
use attention-guided selection instead of random token recomputation (ablation)MARché · ImageNet 256 ablationFID: 564.61 -> 2.56Improved
select refreshing tokens by high attention instead of low attention (ablation)MARché · ImageNet 256 ablationFID: 3.80 -> 2.56Improved
select refreshing tokens by high attention instead of random (ablation)MARché · ImageNet 256 ablationFID: 3.01 -> 2.56Improved
add refreshing tokens (64) to caching (ablation)MARché · ImageNet 256 ablationFID: 3.17 -> 2.56Improved
use layer 4 instead of layer 1 for refresh token selection (ablation)MARché · ImageNet 256 ablationFID: 2.62 -> 2.49; latency 0.154s -> 0.1613sMixed
full refresh at layers 1-6 instead of layer 1 only (ablation)MARché · ImageNet 256 ablationFID: 2.62 -> 2.49; latency 0.155 -> 0.164Mixed
full refresh at layers 7-12 instead of layer 1 only (ablation)MARché · ImageNet 256 ablationFID: 2.62 -> 25.35Hurt
full refresh on even layers instead of layer 1 only (ablation)MARché · ImageNet 256 ablationFID: 2.62 -> 4.65Hurt
replace concatenated attention with cache-aware attention kernel (ablation)MARché · kernel latency (lengths 32 and 256)latency: 904.2 us -> 730.3 usImproved
apply cache-aware inference to MAR-BMAR-B · ImageNet 256x256FID: 2.35 -> 2.56; latency 0.104 -> 0.064Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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