AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

Towards Text-Mask Consistency in Medical Image Segmentation

Jie Gui, HangTu, Wen Sha, Xiuquan Du

ICLR 2026 · Vision-language and multimodal › Cross-modal learning · approach: Contrastive vision-language alignment

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 DualA text fusion to the visual-only baseline (ablation)visual-only baseline · MosMedData+Dice (%): 73.61 -> 75.59Improved
replace DualA with BCAM (ablation)C2Seg · MosMedData+Dice (%): 75.59 -> 76.50Improved
replace DualA with BCAM (ablation)C2Seg · CVC-ClinicDBDice (%): 89.56 -> 90.31Improved
add K-Gate on top of BCAM (ablation)C2Seg · MosMedData+Dice (%): 76.50 -> 77.03Improved
add K-Gate on top of BCAM (ablation)C2Seg · CVC-ClinicDBDice (%): 90.31 -> 90.68Improved
add hard-label contrastive learning in Stage I (ablation)C2Seg · MosMedData+Dice (%): 77.03 -> 77.32Improved
use CaCL soft-label contrastive learning instead of hard-label contrastive learning (ablation)C2Seg · MosMedData+Dice (%): 77.32 -> 77.81Improved
use CaCL soft-label contrastive learning instead of hard-label contrastive learning (ablation)C2Seg · CVC-ClinicDBDice (%): 91.26 -> 91.82Improved
use CLIP text encoder instead of BioBERT-Base (ablation)C2Seg · MosMedData+Dice (%): 76.22 -> 77.81Improved
use CLIP text encoder instead of PubMedBERT (ablation)C2Seg · MosMedData+Dice (%): 76.69 -> 77.81Improved
use hybrid CNN+KAN visual encoder instead of pure CNN (ablation)C2Seg · MosMedData+Dice (%): 77.29 -> 77.81Improved
use hybrid CNN+KAN visual encoder instead of pure KAN (ablation)C2Seg · MosMedData+Dice (%): 76.21 -> 77.81Improved

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