Dense Retrieval on MS MARCO (dev)
0.416MRR@10RetroMAE
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| RetroMAEFine-tuning=Knowledge distillation2022.05 | 0.416 | 0.709 | 0.927 | 0.988 | |
| ERNIE-SearchFine-tuning=Knowledge distillation2022.05 | 0.401 | — | — | 0.982 | |
| ColBERTv2Fine-tuning=Knowledge distillation, Model type=Late-interaction2022.05 | 0.397 | — | — | 0.984 | |
| AR2Fine-tuning=Knowledge distillation2022.05 | 0.395 | — | — | 0.986 | |
| RocketQAv2Fine-tuning=Knowledge distillation2022.05 | 0.388 | — | — | 0.981 |