AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Continuous Speculative Decoding for Autoregressive Image Generation

Zili Wang, Zheng Zhang, Kun Ding, Qi Yang, Fei Li, SHIMING XIANG

ECCV 2026 · Image generation › Fast diffusion sampling · approach: Discrete diffusion solvers · 18 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 denoising trajectory alignment (ablation)continuous speculative decoding (MAR-H/MAR-B) · ImageNet 256, draft length 32acceptance rate: 0.07 -> 0.30Improved
add denoising trajectory alignment (ablation)continuous speculative decoding (MAR-H/MAR-B) · ImageNet 256, draft length 32draft-target distance: 2.56 -> 1.13Improved
add denoising trajectory alignment (ablation)continuous speculative decoding (MAR-H/MAR-B) · ImageNet 256, draft length 4acceptance rate: 0.14 -> 0.32Improved
add denoising trajectory alignment to the approximated ratio (ablation)p(Yp)/q(Yq) criterion · probability ratio analysisacceptance rate: 14% -> 32%Improved
use approximated two-path ratio instead of exact path ratio p(Y)/q(Y) (ablation)acceptance criterion · probability ratio analysisacceptance rate: 0.0% -> 14%Improved
add token pre-filling of 5% target tokens (ablation)continuous speculative decoding (MAR-H/MAR-B) · ImageNet 256, draft length 32acceptance rate: 0.25 -> 0.30Improved
increase pre-filling from 5% to 15% (ablation)continuous speculative decoding (MAR-H/MAR-B) · ImageNet 256, draft length 32speedup: 1.63x -> 1.61xNo clear effect
increase pre-fill ratio from 5% to 15% (ablation)continuous speculative decoding (MAR-H/MAR-B) · ImageNet 256, draft length 32acceptance rate: 0.30 -> 0.33Improved
increase draft length from 4 to 32continuous speculative decoding (MAR-H/MAR-B) · ImageNet 256, bs=256speedup ratio: 1.62x -> 2.33xImproved
increase draft length from 4 to 32continuous speculative decoding (MAR-H/MAR-B) · ImageNet 256acceptance rate: 0.30 -> 0.19Hurt
increase batch size from 1 to 256continuous speculative decoding (MAR-H/MAR-B, draft length 32) · ImageNet 256speedup ratio: 1.44x -> 2.33xImproved
apply continuous speculative decoding with MAR-B draftMAR-H · ImageNet 256 w/ CFGFID: 1.55 -> 1.60No clear effect

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