AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

From Prediction to Perfection: Introducing Refinement to Autoregressive Image Generation

Cheng Cheng, Lin Song, Di An, Yicheng Xiao, Xuchong Zhang, Hongbin Sun, Ying Shan

ICLR 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
add TensorAR refinement to LlamaGen-BLlamaGen-B · ImageNet 256 class-conditionalFID: 5.46 -> 4.71Improved
add TensorAR to LlamaGen-XXLLlamaGen-XXL · ImageNet 384 class-conditionalFID: 2.34 -> 2.03Improved
add TensorAR to Open-MAGVIT2-BOpen-MAGVIT2-B · ImageNet 256 class-conditionalFID: 3.08 -> 2.91Improved
fine-tune baseline for same number of steps without TensorAR (ablation)LlamaGen-B/L · ImageNet 256 class-conditionalFIDNo clear effect
add TensorAR to LlamaGen for text-to-imageLlamaGen · GenEval overalloverall score: 0.32 -> 0.61Improved
add TensorAR to Janus-Pro-7BJanus-Pro-7B · GenEvaloverall score: 0.80 -> 0.83 overall; Counting 0.59 -> 0.53Mixed
add TensorAR to LlamaGenLlamaGen · DPG-Benchoverall score: 43.13 -> 73.33Improved
add TensorAR to Janus-Pro-7BJanus-Pro-7B · DPG-Benchoverall score: 84.19 -> 85.57Improved
use exponential noise schedule instead of linear/sine/square root (ablation)TensorAR on LlamaGen-B · ImageNet 256FID: 4.79/4.75/4.84 -> 4.71Improved
increase window size k from 2 to 8 (ablation)TensorAR on LlamaGen-B · ImageNet 256FID: 4.78 -> 4.68Improved
increase Q_in/Q_out depth from 1 to 4 (ablation)TensorAR on LlamaGen-B · ImageNet 256FID: 4.71 -> 4.90Hurt

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