MedRepBench - Abnormal Flag Recall (OCR Text Only): leaderboard
Structured extraction from photographed, screenshotted and electronic Chinese laboratory reports, by text-only language models given the recognized OCR text; MedRepBench (Shang et al., 2025). Recall in percent of the reference items' abnormality flags, with items matched by normalized name within a budget of as many items as the reference lists; higher is better. Abnormality must be inferred by comparing the value with the range.
Metric: Field-level recall (%; abnormality flag of each ground-truth lab item, matched by exact normalized item name within the model's top-K items, K = number of reference items; text LLM on the OCR text). Source: arxiv.org. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V3 | 89.06 | #705 |
| 2 | Qwen 3 235B A22B 2507 | 88.03 | #613 |
| 3 | Qwen 3 32B | 85.77 | #960 |
| 4 | Qwen 3 14B | 85.43 | #1235 |
| 5 | Qwen 3 235B A22B | 84.54 | #622 |
| 6 | DeepSeek R1 Distill Qwen 14B | 83.01 | #2594 |
| 7 | Qwen 3 30B A3B | 82.67 | #1124 |
| 8 | DeepSeek R1 Distill Qwen 32B | 82.29 | #1780 |
| 9 | Qwen 3 1.7B | 69.04 | #5551 |
| 10 | Qwen 3 8B | 68.5 | #1838 |
| 11 | DeepSeek-R1-Distill-Qwen-7B | 67.13 | #8045 |
| 12 | Qwen 3 4B | 64.29 | #2297 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=medrepbench-abnormal-flag-recall-ocr-text-only · How It Works · Data refreshed daily, snapshot 2026-11-12.