Research map / Vision-language and multimodal / Human and embodied modeling
Med-UniC: Unifying Cross-Lingual Medical Vision-Language Pre-Training by Diminishing Bias
Zhongwei Wan, Che Liu, Mi Zhang, Jie Fu, Benyou Wang, Sibo Cheng, Lei Ma, César Quilodrán-Casas, Rossella Arcucci
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Add CTR to the full Med-UniC objective (ablation) | Med-UniC · CXP500 (En, zero-shot) | F1: 63.5 -> 75.4 | Improved |
| Add self-supervised vision alignment (ablation) | Med-UniC · COVIDx (1% training data) | ACC: 72.8 -> 74.5 | Improved |
| Add CTR while omitting MLM (ablation) | Med-UniC · CXP500 (En, zero-shot) | F1: 70.3 -> 75.4 | Improved |
| Add text-feature alignment to CTR (ablation) | Med-UniC · COVIDx (1% training data) | ACC: 74.9 -> 76.5 | Improved |
| Add text-to-text alignment to CTR (ablation) | Med-UniC · COVIDx (1% training data) | ACC: 75.8 -> 76.5 | Improved |
| Pre-train on English-only MIMIC data rather than English and Spanish data (ablation) | Med-UniC · Object-CXR (1% training data) | mAP: 6.6 -> 3.4 | Hurt |
| Use 512 rather than 1024 CTR embedding dimensions (ablation) | Med-UniC · COVIDx (1% training data) | ACC: 75.5 -> 76.5 | Hurt |
| Use 2048 rather than 1024 CTR embedding dimensions (ablation) | Med-UniC · COVIDx (1% training data) | ACC: 75.9 -> 76.5 | Hurt |
| Freeze all language-model layers instead of updating selected layers (ablation) | Med-UniC · COVIDx (10% training data) | AUC: 89.0 -> 87.1 | Hurt |
| Use ViT-L/32 rather than ResNet-50 as the visual backbone (ablation) | Med-UniC · SIIM segmentation (100% training data) | Dice: 64.4 -> 73.3 | Improved |
Later papers that compared with this method
1 later accepted paper reports Med-Unic† as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Towards All-in-One Medical Image Re-Identification (CVPR 2025 · better on 1 of 1 shared result)
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
- MedUnifier: Unifying Vision-and-Language Pre-training on Medical Data with Vision Generation Task using Discrete Visual Representations (CVPR 2025 · similarity 0.89)
- ExGra-Med: Extended Context Graph Alignment for Medical Vision-Language Models (NeurIPS 2025 · similarity 0.89)
- Interpretable Bilingual Multimodal Large Language Model for Diverse Biomedical Tasks (ICLR 2025 · similarity 0.88)
- MedSIGHT: Towards Grounded Visual Comprehension in Medical Large Vision-Language Models (ICML 2026 · similarity 0.88)
- OmniMedVQA: A New Large-Scale Comprehensive Evaluation Benchmark for Medical LVLM (CVPR 2024 · similarity 0.88)
- MMedPO: Aligning Medical Vision-Language Models with Clinical-Aware Multimodal Preference Optimization (ICML 2025 · similarity 0.87)
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.
