Research map / Vision-language and multimodal / Cross-modal learning
X-PCR: A Benchmark for Cross-modality Progressive Clinical Reasoning in Ophthalmic Diagnosis
Gui Wang, Zehao Zhong, YongSong Zhou, Yudong Li, Ende Wu, Wooi Ping Cheah, Rong Qu, Jianfeng Ren, Linlin Shen
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 |
|---|---|---|---|
| scale InternVL from 8B to 32B | InternVL · X-PCR progressive reasoning | average stage accuracy (%): 64.60 -> 70.14 | Improved |
| go from commercial GPT-5 to open-source InternVL-32B | MLLMs · X-PCR progressive reasoning | average stage accuracy (%): 76.24 -> 70.14 | Hurt |
| increase task difficulty from resident to specialist level | GPT-5 · X-PCR progressive reasoning | expertise-stratified accuracy (%): 80.64 -> 62.92 | Hurt |
| move from IQA to clinical decision-making stage | GPT-5 · X-PCR progressive reasoning | stage accuracy (%): 98.90 -> 54.71 | Hurt |
| use multiple modalities (Group 1 hospital cases) instead of single-modality images | GPT-5 · X-PCR cross-modality | accuracy (%): 72.60 -> 59.60 | Hurt |
| use multiple modalities (Group 1 hospital cases) instead of single-modality images | Qwen3-VL-30B-A3B · X-PCR cross-modality | accuracy (%): 64.33 -> 48.07 | Hurt |
| add more imaging modalities (from CFP/OCT to CFP/OCT/FFA/EP) | GPT-5 · X-PCR hospital cases | accuracy (%): 59.60 -> 47.60 | Hurt |
| keep OCT image (remove it as ablation gives MCS 1.39) (ablation) | GPT-5 · X-PCR hospital cases Group 1 | MCS (accuracy drop %): 1.39 | Improved |
| keep CFP image (removing it gives MCS 0.97) (ablation) | GPT-5 · X-PCR hospital cases Group 1 | MCS (accuracy drop %): 0.97 | Improved |
| keep CFP image (removing it gives MCS -0.59) (ablation) | Qwen3-VL-8B · X-PCR hospital cases Group 1 | MCS (accuracy drop %): -0.59 | Hurt |
| keep CFP in Group 3 (removing it gives MCS -5.08) (ablation) | GPT-5 · X-PCR hospital cases Group 3 | MCS (accuracy drop %): -5.08 | Hurt |
| keep ICGA in Group 3 (removing it gives MCS -4.99) (ablation) | GPT-5 · X-PCR hospital cases Group 3 | MCS (accuracy drop %): -4.99 | Hurt |
Closest papers by meaning
- DrVD-Bench: Do Vision-Language Models Reason Like Human Doctors in Medical Image Diagnosis? (NeurIPS 2025 · similarity 0.86)
- Multi-OphthaLingua: A Multilingual Benchmark for Assessing and Debiasing LLM Ophthalmological QA in LMICs (AAAI 2025 · similarity 0.86)
- M3CoTBench: Benchmark Chain-of-Thought of MLLMs in Medical Image Understanding (ICLR 2026 · similarity 0.85)
- CXReasonBench: A Benchmark for Evaluating Structured Diagnostic Reasoning in Chest X-rays (NeurIPS 2025 · similarity 0.85)
- Med-CMR: A Fine-Grained Benchmark Integrating Visual Evidence and Clinical Logic for Medical Complex Multimodal Reasoning (CVPR 2026 · similarity 0.84)
- Beyond Classification Accuracy: Neural-MedBench and the Need for Deeper Reasoning Benchmarks (ICLR 2026 · similarity 0.84)
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.
