Question Answering on PubHealth (test)
40.86AccuracySFT (DataFlow-Knowledge)
Evaluation Results
| Method | Links | |
|---|---|---|
| SFT (DataFlow-Knowledge)Training=Supervised Fine-Tuning, Synthetic Data=DataFlow-Knowledge2025.12 | 40.86 | |
| CoTReasoning=Chain-of-Thought2025.12 | 29 | |
| RAGReasoning=Retrieval-Augmented Generation2025.12 | 19.6 |