Long-form Question Answering on Fava
61VeriScore F1EWE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EWEBase Model=Llama-3.1 70B2024.12 | 61 | 50.1 | |
| DRAGINBase Model=Llama-3.1 70B2024.12 | 57.2 | 33.9 | |
| RABase Model=Llama-3.1 70B2024.12 | 56.8 | 37.1 | |
| EWEBase Model=Llama-3.1 8B2024.12 | 53.1 | 36.2 | |
| COVE w/ RetrievalBase Model=Llama-3.1 70B2024.12 | 52.6 | 23.1 | |
| Llama-3.1Base Model=Llama-3.1 70B2024.12 | 52 | — | |
| RABase Model=Llama-3.1 8B2024.12 | 51.8 | 16.8 | |
| DRAGINBase Model=Llama-3.1 8B2024.12 | 51.1 | 10 | |
| Llama-3.1Base Model=Llama-3.1 8B2024.12 | 51 | 36.5 | |
| NESTBase Model=Llama-3.1 70B2024.12 | 50.3 | 24.1 | |
| NESTBase Model=Llama-3.1 8B2024.12 | 50.2 | 14.1 | |
| COVEBase Model=Llama-3.1 70B2024.12 | 49.5 | 33.4 | |
| COVE w/ RetrievalBase Model=Llama-3.1 8B2024.12 | 39.5 | 5.3 | |
| COVEBase Model=Llama-3.1 8B2024.12 | 38.7 | 11 |