AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Autoregressive-based Progressive Coding for Ultra-Low Bitrate Image Compression

Ziyuan Zhang, Yichong Xia, Bin Chen, Tianwei Zhang, Hao Wang, Han Qiu

ICLR 2026 · Image generation › Autoregressive and latent generation · approach: Next-scale and token autoregressive images

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 lossless re-encoding with VAR probabilities (ablation)ARPC · DIV2Kbitrate: approximately 30% bitrate reductionImproved
add lossless re-encoding (ablation)ARPC · CLIC2020encoding time (s): 0.20 -> 1.8~6.2Hurt
add scale random dropout (ablation)ARPC · DIV2Kperceptual metricsImproved
add group mask to bitwise quantizer (ablation)ARPC · DIV2KbitrateImproved
use group sizes (4,5,4) rather than (5,6,2) (ablation)ARPC · DIV2KBD-rate FID: 64.19 -> 0Improved
use group sizes (4,5,4) rather than (2,2,9) (ablation)ARPC · DIV2KBD-rate FID: 29.87 -> 0Improved
use modified VAR tokenizer instead of vanilla VAR (ablation)ARPC · DIV2KBD-rate FID: 79.51 -> 0Improved
add GM-BMSRQ to tokenizer (ablation)tokenizer · ImageNet 512 valPSNR: 22.6 -> 22.12Hurt
add scale random dropout to tokenizer (ablation)tokenizer · ImageNet 512 valrFID: 0.56 -> 0.67Hurt
use BLIP-2 instead of lightweight BLIP captioner (ablation)ARPC · DIV2KBD-rate FID: 0.08 -> 0No clear effect
add lossless re-encoding (vs BLIP w/o LRE) (ablation)ARPC · DIV2KBD-rate FID: 25.56 -> 0.08Improved
use InternVL 2.0 detailed captions instead of BLIP2 (ablation)ARPC · DIV2KCLIP ScoreImproved

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