EPAG - Top-k Diagnosis Accuracy: leaderboard
EPAG pre-consultation (Seo et al., 2026): the model acts as a doctor agent asking five turns of multiple-choice history questions of a GPT-4o-mini patient agent, in Korean, over 520 patient profiles covering 26 diseases. A separate GPT-4o-mini diagnostician then predicts a ranked list of diseases from the collected history. Share of cases where the gold disease appears anywhere in that list, in percent; higher is better.
Metric: Disease Diagnosis Accuracy, Gold Anywhere in Ranked List (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 83.81 | #492 |
| 2 | Claude 3.7 Sonnet | 82.31 | #455 |
| 3 | MedGemma-4B | 82.31 | #2460 |
| 4 | GPT-4.1 Mini | 81.36 | #775 |
| 5 | GPT-4o | 81.35 | #741 |
| 6 | Claude 3.5 Sonnet | 81.35 | #747 |
| 7 | Claude 3.5 Haiku | 80.77 | #1563 |
| 8 | GPT-4o Mini | 79.62 | #1627 |
| 9 | Phi-3.5-mini-instruct | 78.84 | #5757 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=epag-top-k-diagnosis-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-12.