AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Seeing What Matters: Lesion-Aware High-Resolution Patch Discovery and Fusion for Chest X-ray Report Generation

Yingshu Li, YUNYI LIU, Zhenghao Chen, Tong Chen, Zailong Chen, Lingqiao Liu, Lei Wang, Luping Zhou

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Attention-based visual token pruning · 1 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 Global-Regional Fusion with randomly sampled high-res patches (ablation)BLIP-3 RRG baseline · MIMIC-CXRBLEU-4: 0.237 -> 0.243Improved
select patches with learned LSRA utility map instead of Grad-CAM (both with GRF) (ablation)LePaX · MIMIC-CXRBLEU-4: 0.251 -> 0.253Improved
add policy supervision (Grad-CAM prior) to LSRA training (ablation)LePaX · MIMIC-CXRBLEU-4: 0.246 -> 0.253Improved
concatenate patch tokens instead of GRF write-back (ablation)LePaX · MIMIC-CXRBLEU-4: 0.253 -> 0.235 (LSRA, no GRF)Hurt
increase uniform input resolution from 384 to 1024 (ablation)BLIP-3 RRG baseline · MIMIC-CXRBLEU-4: 0.237 -> 0.240No clear effect
use LePaX allocation at native resolution 1920 versus 1024 (ablation)LePaX · MIMIC-CXRBLEU-4: 0.245 -> 0.253Improved
increase number of high-res patches Top-K beyond 5 (ablation)LePaX · MIMIC-CXRBLEU-4Hurt
LePaX versus MambaXray-VL-L on CheXpertPlus ROUGELePaX · CheXpertPlus (CXPMRG-Bench)ROUGE: 0.276 -> 0.252Hurt
LePaX versus prior methods on MIMIC-CXR NLGLePaX · MIMIC-CXRBLEU-4: 0.193 -> 0.253Improved

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