MedRepBench - Explanation Acceptability: leaderboard
Patient-facing explanations of de-identified Chinese medical report images, written from the image alone; MedRepBench (Shang et al., 2025). Acceptability rate in percent: the share of explanations a DeepSeek-R1 judge scores 1 or 2 on a 0-2 scale for factual grounding, reasoning and clarity; higher is better.
Metric: Acceptability rate (%; share of patient-facing explanations a DeepSeek-R1 judge scores 1 or 2 on a 0-2 scale for factuality, reasoning and clarity; VLM on the report image only). Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 4 Maverick | 60.78 | #1105 |
| 2 | InternVL3-38B | 51.15 | #905 |
| 3 | InternVL3-8B | 48.24 | #1695 |
| 4 | InternVL3-14B | 47 | #1184 |
| 5 | Llama 4 Scout | 46.84 | #1787 |
| 6 | Qwen 2.5 VL 7B | 26.21 | #1796 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=medrepbench-explanation-acceptability · How It Works · Data refreshed daily, snapshot 2026-11-12.