AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

AttTok: Marrying Attribute Tokens with Generative Pre-trained Vision-Language Models towards Medical Image Understanding

Hualiang Wang, Xinyue Xu, Lehan Wang, Bin Pu, Xiaomeng Li

ICLR 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Medical vision-language models · 0 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 AttTok attribute tokensQwen2.5-VL-7B · 5-modality diagnosis (open-end avg)accuracy: 65.8 -> 69.6Improved
add AttTok attribute tokensLingshu-7B · 5-modality diagnosis (open-end avg)accuracy: 67.2 -> 71.8Improved
add AttTok attribute tokensLingshu-7B · Medical VQA avgaccuracy: 73.2 -> 74.0Improved
add AttTok attribute tokensQwen2.5-VL-7B · Medical VQA avgaccuracy: 71.6 -> 72.5Improved
add ACC adapter and ACM loss jointly rather than either alone (ablation)AttTok · five modalitiesaccuracyImproved
add ACM loss alone vs ACC alone (ablation)AttTok · five modalitiesaccuracyImproved
add visual-only clustering alignment (ablation)Lingshu-7B · Fundus and OCT (open-end)accuracy: 63.7 -> 63.2; 74.7 -> 74.2Hurt
add vision-text alignment (ablation)Lingshu-7B · Dermatology (open-end)accuracy: 72.8 -> 73.6Improved
use AttTok alignment instead of best alternative (ablation)Lingshu-7B · Dermatology, Fundus, OCTaccuracy: +3.3 Dermatoscopy, +3.2 Fundus, +2.7 OCTImproved
add AttTokLingshu-7B · Fitzpatrick17kaccuracy: 0.27 -> 0.41Improved
add AttTokLingshu-7B · RFMiDmAP: 0.25 -> 0.36Improved
add AttTok without pre-balancingLingshu-7B · ISIC-2018 imbalanced, DF classF1: 11.8 -> 31.4Improved

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