Medical Report Classification on NIH-CXR
97.2Micro PrecisionBioClinicalBERT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| BioClinicalBERTTraining Mode=Baseline2026.04 | 97.2 | 57.5 | 72.3 | |
| MedalpacaTraining Mode=SFT + GRPO2026.04 | 94 | 75.8 | 83.9 | |
| LlamaTraining Mode=SFT2026.04 | 92.9 | 77.8 | 84.7 | |
| MedalpacaTraining Mode=SFT2026.04 | 92.9 | 74.2 | 82.4 | |
| Qwen-3BTraining Mode=SFT2026.04 | 90.7 | 75.2 | 82.2 | |
| Qwen-3BTraining Mode=SFT + GRPO2026.04 | 87.5 | 77.8 | 82.4 | |
| LlamaTraining Mode=SFT + GRPO2026.04 | 87 | 85.5 | 86.2 | |
| Qwen-7BTraining Mode=SFT + GRPO2026.04 | 86.8 | 81.7 | 84.1 | |
| Phi3Training Mode=SFT + GRPO2026.04 | 82.9 | 59 | 68.9 | |
| Qwen-7BTraining Mode=SFT2026.04 | 82.7 | 83.5 | 83 | |
| Phi3Training Mode=SFT2026.04 | 79.5 | 49.6 | 61.1 | |
| GPT4-oTraining Mode=Baseline2026.04 | 77.6 | 94.9 | 85.4 | |
| Qwen-7BTraining Mode=Baseline2026.04 | 60.2 | 34.2 | 43.5 | |
| Phi3Training Mode=Baseline2026.04 | 42.9 | 35.9 | 39.1 | |
| Qwen-3BTraining Mode=Baseline2026.04 | 36.2 | 17.9 | 24 | |
| LlamaTraining Mode=Baseline2026.04 | 28.6 | 13.7 | 18.5 | |
| MedalpacaTraining Mode=Baseline2026.04 | 12.2 | 8.1 | 9.7 |