AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Grouped Speculative Decoding for Autoregressive Image Generation

Junhyuk So, Juncheol Shin, Hyunho Kook, Eunhyeok Park

ICCV 2025 · Image generation › Autoregressive and latent generation · approach: Accelerating autoregressive image generation · 6 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
use GSD with group size G=3 instead of SJDLumina-mGPT SJD · Parti-Prompt (Lumina-mGPT 7B)NFE: 1035.3 -> 898.97Improved
use GSD with G=3 instead of SJDLumina-mGPT SJD · MS-COCO 2017 val (Lumina-mGPT 7B)FID: 30.78 -> 31.50Hurt
increase GSD group size to G=50GSD · Parti-Prompt (Lumina-mGPT 7B)latency speedup: 2.38x -> 4.65xImproved
increase GSD group size from G=50 to G=100GSD · Parti-Prompt (Lumina-mGPT 7B)CLIP score: 32.075 -> 31.975Hurt
increase GSD group size from G=3 to G=25GSD · MS-COCO 2017 val (Lumina-mGPT 7B)FID: 31.50 -> 33.55Hurt
use GSD (G=50) instead of Addition (eps=3e-1) naive lossy SDSJD with lossy acceptance · Parti-Prompt (Lumina-mGPT 7B)CLIP score: 31.776 -> 32.075Improved
use GSD (G=25) instead of Amplify (k=4) naive lossy SDSJD with lossy acceptance · MS-COCO 2017 val (Lumina-mGPT 7B)FID: 40.05 -> 33.55Improved
add context-aware clustering by expert p(x) instead of static embedding distance (ablation)GSD · Parti-Prompt clustering ablation (Lumina-mGPT 7B)NFE: 913.80 -> 636.75Improved
cluster by expert p(x) rather than draft q(x) (ablation)GSD · Parti-Prompt clustering ablation (Lumina-mGPT 7B)CLIP score: 31.791 -> 32.075Improved
use dynamic clustering (ours) rather than plain expert p(x) clustering (ablation)GSD · Parti-Prompt clustering ablation (Lumina-mGPT 7B)CLIP score: 32.056 -> 32.075No clear effect
use Jacobi Decoding instead of vanilla ARLumina-mGPT · Parti-Prompt (Lumina-mGPT 7B)NFE: 2392 -> 2300.0No clear effect
use SJD instead of vanilla ARLumina-mGPT · Parti-Prompt (Lumina-mGPT 7B)latency speedup: 1.00x -> 2.15xImproved

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