Conversational Retrieval on QReCC
53.37Top-1 RecallE5-unsupervised + Fine-tune
Evaluation results
| Method | Links | ||
|---|---|---|---|
| E5-unsupervised + Fine-tuneBackbone=E5, Query processing=History concatenation, Training protocol=Fine-tune, Training dataset=HumanAnnotatedConvQA2024.01 | 53.37 | 89.91 | |
| Dragon + RewriteBackbone=Dragon, Query processing=Rewrite, Training protocol=Unsupervised2024.01 | 51.73 | 85.78 | |
| Dragon + Fine-tuneBackbone=Dragon, Query processing=History concatenation, Training protocol=Fine-tune, Training dataset=HumanAnnotatedConvQA2024.01 | 50.73 | 87.17 | |
| Dragon + Fine-tune (SyntheticConvQA)Backbone=Dragon, Query processing=History concatenation, Training protocol=Fine-tune, Training dataset=SyntheticConvQA2024.01 | 49.63 | 86.7 | |
| DragonBackbone=Dragon, Training protocol=Unsupervised2024.01 | 46.17 | 81.96 | |
| E5-unsupervised + RewriteBackbone=E5, Query processing=Rewrite, Training protocol=Unsupervised2024.01 | 45.5 | 85.89 | |
| E5-unsupervisedBackbone=E5, Training protocol=Unsupervised2024.01 | 44.71 | 84.99 |