EPPCMinerBen (2-Shot) - Subcode Classification: leaderboard
Metric: Micro-averaged F1 (%) of EPPC subcode classification (choosing the code-specific subcode for each predicted code, compared with the expert code-subcode pairs), two in-context examples, mean over seeded runs, on 1,933 expert-annotated sentences from 752 Yale New Haven Health patient-portal secure messages; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 Distill Qwen 32B | 47.95 | #640 |
| 2 | DeepSeek R1 Distill Llama 70B | 46.15 | #582 |
| 3 | Llama 3.3 70B Instruct | 44.46 | #520 |
| 4 | Llama 3.1 70B Instruct | 44.15 | #548 |
| 5 | Qwen2.5-14B-Instruct-1M | 41.56 | #566 |
| 6 | Gemma 2 27B (IT) | 30.88 | #633 |
| 7 | Llama 3.1 8B Instruct | 24.66 | #1018 |
| 8 | Llama 3.2 3B Instruct | 15.65 | #1321 |
| 9 | Qwen 2.5 1.5B Instruct | 14.47 | #1388 |
| 10 | Llama 3.2 1B Instruct | 8.26 | #1518 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=eppcminerben-2-shot-subcode-classification · How It Works · Data refreshed daily, snapshot 2026-10-11.