AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Annealed Relaxation of Speculative Decoding for Faster Autoregressive Image Generation

Xingyao Li, Fengzhuo Zhang, Cunxiao Du, Hui Ji

AAAI 2026 · Image generation › Autoregressive and latent generation · approach: Next-scale and token autoregressive images · 3 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 annealed per-step relaxation (Cool-SD) instead of uniform relaxation parameter (ablation)relaxed speculative decoding · LlamaGen-XL and Lumina-mGPT, MS-COCO valFID vs accepted length/latencyImproved
replace LANTERN++ resampling distribution with the derived optimal G* (ablation)LANTERN++ · LlamaGen-XL and Lumina-mGPT, MS-COCO valFID vs speed trade-offImproved
larger k in LANTERN++ with its original resampling (ablation)LANTERN++ · MS-COCO valFIDHurt
add Cool-SD relaxation (delta=1.1) over lossless Eagle-1Eagle-1 · LlamaGen-XL, MS-COCO valaccepted length: 2.42 -> 2.73Improved
raise relaxation budget to delta=2Cool-SD · LlamaGen-XL, MS-COCO valaccepted length: 2.42 -> 3.34Improved
increase delta from 1.1 to 2 (ablation)Cool-SD · LlamaGen-XL, MS-COCO valFID: 21.02 -> 21.20Mixed
add Cool-SD relaxation over Eagle-1Eagle-1 · Lumina-mGPT, MS-COCO valaccepted length: 2.75 -> 3.15Improved
Cool-SD instead of LANTERN++ at similar qualityLANTERN++ · Lumina-mGPT, MS-COCO valspeed-up: 2.48 -> 2.69Improved
Cool-SD instead of Speculative Jacobi DecodingSJD · Lumina-mGPT, MS-COCO valspeed-up: 2.06 -> 2.69Improved
Cool-SD instead of Speculative Jacobi DecodingSJD · LlamaGen-XL, MS-COCO valspeed-up: 1.84 -> 2.27Improved

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