Classification on AI-READI Scenario A — Positive-Class Undersampling (train)
21ACCLLaMA*
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LLaMA*Augmentation=CF Added, Prompting=Fine-tuned2026.01 | 21 | 20 | 24.56 | 22.41 | 25.37 | |
| BioMistral*Augmentation=CF Added, Prompting=Fine-tuned2026.01 | 14.93 | 18.33 | 22.81 | 20.69 | 20.9 | |
| GPT-4Augmentation=CF Added, Prompting=Zero-shot2026.01 | 10.29 | 10 | 12.28 | 12.07 | 11.94 | |
| LLaMAAugmentation=CF Added, Prompting=Zero-shot2026.01 | 8.54 | 6.67 | 10.53 | 8.62 | 7.46 | |
| BioMistralAugmentation=CF Added, Prompting=Zero-shot2026.01 | 7.1 | 5 | 8.77 | 6.9 | 7.46 | |
| NNData Condition=Model trained on reduced data2026.01 | -12.87 | -11.76 | -16.18 | -14.71 | -14.1 |