Research map / Vision-language and multimodal / Multimodal LLMs
VisionUnite: A Vision-Language Foundation Model for Ophthalmology Enhanced With Clinical Knowledge
Zihan Li, Diping Song, Zefeng Yang, Deming Wang, Fei Li, Xiulan Zhang, Paul E. Kinahan, Yu Qiao
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 Vision Projector to CLIP loss plus Vision Adapter (ablation) | VisionUnite · Fundus multi-round VQA (SBERT) component ablation | SBERT similarity (%): 79.06 -> 79.94 | Improved |
| add Vision Adapter with classification loss to CLIP plus Projector (ablation) | VisionUnite · Fundus multi-round VQA (SBERT) component ablation | SBERT similarity (%): 77.35 -> 79.94 | Improved |
| add CLIP contrastive loss to Adapter plus Projector (ablation) | VisionUnite · Fundus multi-round VQA (SBERT) component ablation | SBERT similarity (%): 78.30 -> 79.94 | Improved |
| add Vision Adapter with classification loss to CLIP loss alone (ablation) | VisionUnite · Fundus multi-round VQA (SBERT) component ablation | SBERT similarity (%): 75.14 -> 79.06 | Improved |
| add MS COCO pretraining (ablation) | VisionUnite · Fundus multi-round VQA (SBERT) pretraining ablation | SBERT similarity (%): 74.69 -> 77.65 | Improved |
| add PMC-OA pretraining (ablation) | VisionUnite · Fundus multi-round VQA (SBERT) pretraining ablation | SBERT similarity (%): 74.69 -> 76.76 | Improved |
| add combined pretraining datasets (ablation) | VisionUnite · Fundus multi-round VQA (SBERT) pretraining ablation | SBERT similarity (%): 74.69 -> 79.38 | Improved |
| VisionUnite versus junior ophthalmologist | VisionUnite · Fundus multi-round VQA (180 images, expert-rated) | diagnostic accuracy: .735 -> .780 | No clear effect |
| VisionUnite versus GPT-4V | VisionUnite · Fundus multi-round VQA (180 images, expert-rated) | diagnostic accuracy: .498 -> .780 | Improved |
| fine-tuned VisionUnite versus fine-tuned LLaVA | VisionUnite · Fundus multi-round VQA (SBERT similarity) | SBERT similarity (%): 74.26 -> 79.94 | Improved |
| VisionUnite versus best open-source baseline on multiple-choice | VisionUnite · Fundus multiple-choice (25 conditions) | accuracy (%): 85.22 -> 94.36 | Improved |
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.
- Mini-Gemini — Mini-Gemini: Mining the Potential of Multi-Modality Vision Language Models (TPAMI 2025) · better on 2 of 2 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
- BIOMEDICA: An Open Biomedical Image-Caption Archive, Dataset, and Vision-Language Models Derived from Scientific Literature (CVPR 2025 · similarity 0.87)
- MedSIGHT: Towards Grounded Visual Comprehension in Medical Large Vision-Language Models (ICML 2026 · similarity 0.86)
- Multi-OphthaLingua: A Multilingual Benchmark for Assessing and Debiasing LLM Ophthalmological QA in LMICs (AAAI 2025 · similarity 0.85)
- MMed-RAG: Versatile Multimodal RAG System for Medical Vision Language Models (ICLR 2025 · similarity 0.85)
- MIMO: A Medical Vision Language Model with Visual Referring Multimodal Input and Pixel Grounding Multimodal Output (CVPR 2025 · similarity 0.85)
- OpenMedReason: Scientific Reasoning Supervision for Medical Vision–Language Models (NeurIPS 2026 · similarity 0.85)
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.
