Multi-hop Retrieval on Multi-hop 4 datasets aggregate (test)
58.5NDCG@10Ours
Evaluation Results
| Method | Links | |
|---|---|---|
| OursBackbone=RetroMAE, Q/Doc=22026.02 | 58.5 | |
| Ours (8k)Backbone=RetroMAE, Q/Doc=22026.02 | 56.91 | |
| OursBackbone=Contriever, Q/Doc=32026.02 | 55.22 | |
| Ours (8k)Backbone=Contriever, Q/Doc=52026.02 | 54.4 | |
| SupervisedBackbone=RetroMAE2026.02 | 51.15 | |
| SupervisedBackbone=Contriever2026.02 | 50.41 | |
| PretrainedBackbone=Contriever2026.02 | 39.7 | |
| PretrainedBackbone=RetroMAE2026.02 | 18.45 |