Conversational Multi-document Question Answering on Multi-doc QA open-ended
53.94ROUGE-1Oracle
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| OraclePredictor=LLAMA2-7B, Generator=LLAMA2-13B-chat2024.09 | 53.94 | 26.79 | 49.98 | |
| AdaCompPredictor=LLAMA2-7B, Generator=LLAMA2-13B-chat2024.09 | 48.75 | 22.67 | 45.01 | |
| Top-5Predictor=LLAMA2-7B, Generator=LLAMA2-13B-chat2024.09 | 46.25 | 20.52 | 42.66 | |
| Top-RandomPredictor=LLAMA2-7B, Generator=LLAMA2-13B-chat2024.09 | 46.02 | 20.51 | 42.24 | |
| Top-1Predictor=LLAMA2-7B, Generator=LLAMA2-13B-chat2024.09 | 44.01 | 19.23 | 40.37 | |
| RECOMPPredictor=LLAMA2-7B, Generator=LLAMA2-13B-chat2024.09 | 38.82 | 15.08 | 35.4 |