Seizure Frequency Extraction on Real 300 1.0 (Purist)
77.6Micro ScoreOur COT
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Our COTmodel=Qwen2.5-14B, Fine-tuning data=1,797 synthetic letters2026.03 | 77.6 | 0.8 | 49.7 | 3.1 | 77.4 | 1 | |
| Our COTmodel=Ministral-8B, Fine-tuning data=1,797 synthetic letters2026.03 | 76.6 | 0.5 | 47.5 | 1.4 | 76.4 | 0.4 | |
| Our labelmodel=Qwen2.5-7B, Fine-tuning data=1,797 synthetic letters2026.03 | 76.2 | 0.9 | 46.5 | 0.7 | 76.1 | 0.7 | |
| Our labelmodel=Llama-3.1-8B, Fine-tuning data=1,797 synthetic letters2026.03 | 75.8 | 2 | 48.2 | 2.4 | 75.8 | 1.8 | |
| Our COTmodel=Qwen2.5-7B, Fine-tuning data=1,797 synthetic letters2026.03 | 75.6 | 1 | 46.8 | 1.4 | 75.5 | 1.1 | |
| Our COTmodel=Llama-3.1-8B, Fine-tuning data=1,797 synthetic letters2026.03 | 75.5 | 1.4 | 46.6 | 3 | 75.4 | 1.1 | |
| Our labelmodel=lingshu-7B, Fine-tuning data=1,797 synthetic letters2026.03 | 75.5 | 0.6 | 46.4 | 2 | 75.5 | 0.5 | |
| Our COTmodel=Gemma-3-4B, Fine-tuning data=1,797 synthetic letters2026.03 | 75 | 0.9 | 44.9 | 1.4 | 74.4 | 0.9 | |
| Our COTmodel=Medgemma-4b, Fine-tuning data=1,797 synthetic letters2026.03 | 75 | 1.2 | 45.9 | 3.9 | 74.5 | 1.1 | |
| Our labelmodel=Ministral-8B, Fine-tuning data=1,797 synthetic letters2026.03 | 74.7 | 1.8 | 51.8 | 2.7 | 75.4 | 1.4 | |
| Our labelmodel=Qwen2.5-14B, Fine-tuning data=1,797 synthetic letters2026.03 | 74.5 | 1.8 | 49.6 | 2 | 75.1 | 1.4 | |
| Our COTmodel=lingshu-7B, Fine-tuning data=1,797 synthetic letters2026.03 | 73.5 | 0.6 | 43.9 | 0.5 | 73.2 | 0.4 | |
| Our labelmodel=Medgemma-4B, Fine-tuning data=1,797 synthetic letters2026.03 | 72.9 | 1.3 | 45.5 | 2.2 | 72.8 | 1.2 | |
| x per Mmodel=Ministral-8B, Fine-tuning data=1,797 synthetic letters2026.03 | 71.8 | 0.9 | 39.9 | 0.5 | 69.9 | 0.9 | |
| Our labelmodel=Gemma-3-4B, Fine-tuning data=1,797 synthetic letters2026.03 | 71.7 | 2.4 | 44.8 | 3.8 | 72 | 2 | |
| x per Mmodel=Qwen2.5-7B, Fine-tuning data=1,797 synthetic letters2026.03 | 71.5 | 2.3 | 38.8 | 2.1 | 69.5 | 2.2 | |
| x per Mmodel=Qwen2.5-14B, Fine-tuning data=1,797 synthetic letters2026.03 | 71.3 | 1.4 | 39.1 | 0.9 | 68.9 | 1.3 | |
| x per Mmodel=Gemma-3-4B, Fine-tuning data=1,797 synthetic letters2026.03 | 69.4 | 1.9 | 37.9 | 1.3 | 67.8 | 1.6 | |
| x per Mmodel=Llama-3.1-8B, Fine-tuning data=1,797 synthetic letters2026.03 | 69.1 | 0.9 | 39.2 | 1.4 | 67.9 | 0.9 | |
| x per Mmodel=lingshu-7B, Fine-tuning data=1,797 synthetic letters2026.03 | 68.1 | 1.9 | 39.1 | 1.5 | 67 | 1.5 | |
| x per Mmodel=Medgemma-4B, Fine-tuning data=1,797 synthetic letters2026.03 | 67.8 | 2.3 | 36.9 | 1.8 | 66.1 | 1.9 |