Closed-set Question Answering on PubHealth
74.5AccuracySELF-RAG
Evaluation results
| Method | Links | |
|---|---|---|
| SELF-RAGScale=13B, Retrieval=Yes2023.10 | 74.5 | |
| SELF-RAGScale=7B, Retrieval=Yes2023.10 | 72.4 | |
| ChatGPTRetrieval=No, Proprietary=true2023.10 | 70.1 | |
| Llama2-FTScale=7B, Retrieval=Yes, Fine-tuned=true2023.10 | 64.3 | |
| AlpacaScale=13B, Retrieval=No2023.10 | 55.5 | |
| Ret-ChatGPTRetrieval=Yes, Proprietary=true2023.10 | 54.7 | |
| Ret-Llama2-ChatScale=13B, Retrieval=Yes, Proprietary=true2023.10 | 52.1 | |
| AlpacaScale=13B, Retrieval=Yes2023.10 | 51.1 | |
| AlpacaScale=7B, Retrieval=No2023.10 | 49.8 | |
| Llama2-ChatScale=13B, Retrieval=No, Proprietary=true2023.10 | 49.4 | |
| AlpacaScale=7B, Retrieval=Yes2023.10 | 40.2 | |
| Llama2Scale=7B, Retrieval=No2023.10 | 34.2 | |
| Llama2Scale=13B, Retrieval=Yes2023.10 | 30.2 | |
| Llama2Scale=7B, Retrieval=Yes2023.10 | 30 | |
| Llama2Scale=13B, Retrieval=No2023.10 | 29.4 |