README.md

June 5, 2026 · View on GitHub

Project Page Hugging Face Dataset
GitHub Stars arXiv ICML 2026 License

🎉 Accepted to ICML 2026

🌟 Github | 📥 Model Download | 📁 Dataset (UniMed-5M, partial) | 📄 Paper Link | 🌐 Project Page

UniMedVL: Unifying Medical Multimodal Understanding and Generation through Observation-Knowledge-Analysis

A unified medical foundation model enabling both understanding and generation capabilities within a single architecture

🔥 News

📚 Introduction

We introduce UniMedVL, the unified medical foundation model for seamless multimodal understanding and generation. Four key innovations distinguish UniMedVL:

  • Unified Observation-Knowledge-Analysis Architecture: UniMedVL sets itself apart from prior medical AI models by following a clinically-inspired three-level framework that mirrors how physicians process medical information, enabling both understanding and generation within a single architecture.

  • Versatile Medical Multimodal Capabilities: UniMedVL supports a broad spectrum of medical tasks, including visual question answering, medical report generation, text-to-medical-image synthesis, cross-modal translation, and virtual staining across 9 imaging modalities.

  • Large-Scale Medical Dataset: We present UniMed-5M, a comprehensive medical multimodal dataset containing 5.6M+ high-quality samples with three-stage quality verification and expert validation, covering understanding, generation, and interleaved tasks.

  • Superior Performance: UniMedVL achieves state-of-the-art performance on multiple evaluation datasets, with 75.4% accuracy on SLAKE VQA, 53.5% on PathVQA, and competitive generation quality (96.29 average gFID), setting a new standard in unified medical AI.

UniMedVL Architecture

📝 Open-Source Plan

  • 🖼️ Visualizations - Result figures and model demonstrations
  • 💾 Model Checkpoints - Pre-trained UniMedVL weights (14B parameters)
  • 🔧 Inference Code - Model loading and inference examples
  • 📁 UniMedVL-5M Dataset - Training dataset and Evaluation dataset
  • 🏋️ Training Code - Full training pipeline and configuration files

🔬 Methodology

📋 OKA Framework: Observation-Knowledge-Analysis

UniMedVL follows a workflow-guided three-level framework that mirrors how physicians process medical information:

OKA Framework Workflow

🎯 Training Strategy

Three-Stage Progressive Curriculum Learning:

  1. 🔧 Stage 1 - Foundation Training (85K steps)

    • Basic medical pattern recognition
    • Visual-language alignment
    • Data ratio: 75% I2T, 25% T2I

  2. 📚 Stage 2 - Instruction Tuning (120K steps)

    • Cross-modal understanding enhancement
    • Medical expertise development
    • Data ratio: 40% I2T, 45% T2I, 10% Interleaved

  3. 🚀 Stage 3 - Unified Training (70K steps)

    • Advanced multimodal synthesis
    • Interleaved task mastery
    • Data ratio: 37% I2T, 35% T2I, 25% Interleaved

💬 Qualitative Results

Here we present some comprehensive visualization results demonstrating UniMedVL's capabilities. For additional visualization results and comparisons, please see our Project Page.

Performance Across Training Stages
Performance Comparison

Comprehensive performance comparison across training stages and modalities

Multimodal Tasks Demonstration
Multimodal Task Results

Comprehensive visualization of UniMedVL's multimodal capabilities across diverse medical tasks

Medical Visual Question Answering
Medical VQA Examples

Medical Visual Question Answering examples showing model's diagnostic reasoning capabilities

Medical Report Generation
Medical Report Generation

Automated medical report generation examples across different imaging modalities

Text-to-Medical-Image Generation
Text-to-Image Generation Examples 1

Text-to-medical-image generation results showing high-quality synthesis

Text-to-Image Generation Examples 2

Additional text-to-medical-image generation examples across modalities

Medical-Image Generation across 8 modalities

Chest X-Ray (CXR)

Chest X-Ray

Computed Tomography (CT)

CT Scan

Magnetic Resonance Imaging (MRI)

MRI Scan

Ultrasound

Ultrasound

Histopathology (HIS)

Histopathology

Retinal Fundus Photography (CFP)

Retinal Fundus

Optical Coherence Tomography (OCT)

OCT

Endoscopy

Endoscopy

📊 Quantitative Performance

Medical Visual Question Answering Performance
ModelParamsTypeVQA-RADSLAKEPathVQAOmniMedVQAGMAI-MMBench
GMAI-VL7BMedical-specific66.372.939.888.561.74
HuatuoGPT-Vision7BMedical-specific53.049.132.050.050.22
Bagel7BUnified60.0958.9139.0571.1348.11
HealthGPT-L1414BUnified58.364.544.474.443.1
UniMedVL14BUnified61.975.453.585.860.75
Medical Image Generation Performance

Text-to-image generation performance across 8 medical imaging modalities. Metrics: gFID ↓ (lower is better) / BioMedCLIP Score ↑ (higher is better)

ModelCFPCXRCTHISMRIOCTUltrasoundEndoscopyAverage
Bagel (7B)217.19/0.650182.80/0.662163.78/0.652206.18/0.643175.74/0.639307.80/0.719255.78/0.672214.61/0.668215.49/0.660
UniMedVL (14B)53.20/0.70873.04/0.70273.04/0.696149.01/0.70490.36/0.70699.27/0.72195.38/0.706133.11/0.70796.29/0.706
Interleaved Multimodal Tasks Performance

Virtual Immunohistochemistry Staining (H&E → IHC)

MethodTypePSNR ↑SSIM ↑
Pyramid Pix2pixSpecialized21.160.477
HealthGPT-M3Unified15.810.242
UniMedVLUnified20.270.456

MRI Super-Resolution (4× upsampling)

MethodTypePSNR ↑SSIM ↑
AMIRSpecialized31.990.939
HealthGPT-M3Unified18.370.580
UniMedVLUnified27.290.890

Cross-Modal Synthesis (T2 ↔ FLAIR MRI)

MethodTypeAverage PSNR ↑Average SSIM ↑
ResViTSpecialized25.380.889
HealthGPT-M3Unified19.090.748
UniMedVLUnified25.070.882
Counterfactual Medical Image Generation

Performance on counterfactual chest X-ray generation with explanatory text. † indicates unified fine-tuning variant.

MethodgFID ↓AUROC ↑F1 ↑BLEU-3 ↑METEOR ↑ROUGE-L ↑
ProgEmu29.210.7920.8910.1240.4100.261
UniMedVL†27.170.7970.8730.2640.4490.465

🚀 Getting Started

Installation

conda env create -f codes/environment.yaml
conda activate unimedvl

Inference Scripts

Two interactive inference scripts are provided in the codes/ directory:

  1. Medical Visual Question Answering (interactive_vqa_inferencer.py)

  2. Medical Image Generation (interactive_image_generator.py)

Quick Usage

  1. Download the UniMedVL checkpoint: 
    huggingface-cli download General-Medical-AI/UniMedVL --local-dir ./checkpoints/UniMedVL
  2. Set model_path and ROOT in the script configuration
  3. Run the script: python codes/interactive_vqa_inferencer.py or python codes/interactive_image_generator.py

📜 License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.

📚 Citations

If you use this project in your research or work, please cite it as:

 @misc{ning2025unimedvlunifyingmedicalmultimodal,
      title={Unimedvl: Unifying Medical Multimodal Understanding And Generation Through Observation-Knowledge-Analysis}, 
      author={Junzhi Ning and Wei Li and Cheng Tang and Jiashi Lin and Chenglong Ma and Chaoyang Zhang and Jiyao Liu and Ying Chen and Shujian Gao and Lihao Liu and Yuandong Pu and Huihui Xu and Chenhui Gou and Ziyan Huang and Yi Xin and Qi Qin and Zhongying Deng and Diping Song and Bin Fu and Guang Yang and Yuanfeng Ji and Tianbin Li and Yanzhou Su and Jin Ye and Shixiang Tang and Ming Hu and Junjun He},
      year={2025},
      eprint={2510.15710},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2510.15710},
}

🙏 Acknowledgments

We sincerely thank the following projects and their contributors for their invaluable open-source contributions that made this research possible:

  • Bagel - Foundation model architecture and training methodology inspiration
  • HealthGPT - Medical domain adaptation and evaluation framework
  • VLMEvalKit - Comprehensive evaluation toolkit for vision-language models