AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

TaskTok: Delving into Task Tokens for Task-driven Image Restoration

Hongjae Lee, Sojung Kang, Jaeseong Yu, Seung-Won Jung

ECCV 2026 · Image generation › Autoregressive and latent generation · approach: Generative latent space tooling · 2 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 pre-restoration module before tokenizer (ablation)TiTok-64 (Mix-B) · ImageNet classificationTop-1 accuracy (%): 47.4 -> 48.5Improved
add pre-restoration module before tokenizer (ablation)TiTok-64 (Mix-B) · PASCAL VOC2012 segmentationmIoU: 55.5 -> 56.9Improved
add pre-restoration module before tokenizer (ablation)TiTok-64 (Mix-B) · PASCAL VOC2012 detectionmAP: 22.3 -> 23.1Improved
add token refiner updating all tokens (ablation)TiTok-64 (Mix-B) with pre-restoration · ImageNet classificationTop-1 accuracy (%): 48.5 -> 46.3Hurt
add token refiner updating all tokens (ablation)TiTok-64 (Mix-B) with pre-restoration · PASCAL VOC2012 detectionmAP: 23.1 -> 21.8Hurt
add token-level loss L_tok (ablation)TiTok-64 with refiner · ImageNet classificationTop-1 accuracy (%): 46.3 -> 46.7No clear effect
add learnable token switch (ablation)TiTok-64 with refiner and L_tok · ImageNet classificationTop-1 accuracy (%): 46.7 -> 46.9Improved
add mask regularizer L_mask (ablation)TaskTok-64 with token switch · ImageNet classificationTop-1 accuracy (%): 46.9 -> 50.9Improved
add importance-order initialization pi (ablation)TaskTok-64 with L_mask · ImageNet classificationTop-1 accuracy (%): 50.9 -> 53.0Improved
remove codebook quantization (continuous tokens) (ablation)TaskTok-64 · ImageNet classificationTop-1 accuracy (%): 53.0 -> 55.3Improved
select tokens for another task (Seg/Det) when evaluating CLS (ablation)TaskTok-64 · ImageNet classificationTop-1 accuracy (%): -1.4% / -1.8% from 55.3Hurt
select tokens across Seg and Det tasks (ablation)TaskTok-64 · PASCAL VOC2012mIoU/mAP: -0.6% / -0.6%Hurt

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