Research map / Vision-language and multimodal / Human and embodied modeling
Medic-AD: Towards Medical Vision-Language Model's Clinical Intelligence
Woohyeon Park, Jaeik Kim, Sunghwan Steve Cho, Pa Hong, Wookyoung Jeong, Yoojin Nam, Namjoon Kim, Ginny Y. Wong, Ka Chun Cheung, Jaeyoung Do
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 anomaly-aware <Ano> tokens and stage-wise training to the Lingshu backbone | Lingshu-7B · Zero-shot anomaly detection (avg of 4 datasets) | Avg. F1: 88.7 -> 91.6 | Improved |
| add <Diff> tokens for temporal difference reasoning | Lingshu-7B · MMXU | overall accuracy: 0.620 -> 0.655 | Improved |
| add <Diff> tokens | Lingshu-7B · MMXU (Improved class) | accuracy: 0.734 -> 0.714 | Hurt |
| use <Ano> and heatmap decoder instead of SAM2 decoder for grounding | Citrus-V comparison · BraTS2021 | mIoU: 32.6 -> 87.6 | Improved |
| add <Ano> tokens to the <Diff> pipeline (ablation) | Medic-AD · MMXU | overall accuracy: 0.635 -> 0.655 | Improved |
| use intermediate 4-layer visual features instead of last layer only (ablation) | Medic-AD · MMXU | overall accuracy: 0.619 -> 0.635 | Improved |
| use intermediate 4-layer visual features instead of last layer only (ablation) | Medic-AD · Zero-shot anomaly detection | Avg. F1: 90.9 -> 91.6 | Improved |
| set query-token pooling size to 4x4 (ablation) | Medic-AD · Anomaly detection and MMXU | Avg. F1: 88.7 -> 90.2 -> 91.8 (Fig. 5a) | Improved |
| use 10 visual soft prompts (ablation) | Medic-AD · Anomaly detection and MMXU | Avg. F1 | Improved |
| apply Medic-AD to real hospital follow-up chest X-ray pairs | Lingshu-7B · Real hospital data (300 patients) | RaTEScore: 0.359 -> 0.430 | Improved |
Closest papers by meaning
- Adapting Visual-Language Models for Generalizable Anomaly Detection in Medical Images (CVPR 2024 · similarity 0.88)
- MedSIGHT: Towards Grounded Visual Comprehension in Medical Large Vision-Language Models (ICML 2026 · similarity 0.87)
- DrVD-Bench: Do Vision-Language Models Reason Like Human Doctors in Medical Image Diagnosis? (NeurIPS 2025 · similarity 0.87)
- How Do Medical MLLMs Fail? A Study on Visual Grounding in Medical Images (ICLR 2026 · similarity 0.86)
- MedMO: Grounding and Understanding Multimodal Large Language Model for Medical Images (CVPR 2026 · similarity 0.86)
- MedAtlas: Evaluating LLMs for Multi-Round, Multi-Task Medical Reasoning Across Diverse Imaging Modalities and Clinical Text (AAAI 2026 · similarity 0.86)
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.
