Zero-shot Information Retrieval on BEIR
39.1NFCorpus NDCG@10 (Zero-shot)SSR-CLS
Evaluation Results
| Method | Links | ||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SSR-CLSRetrieval Type=Late-interaction sparse retrieval2026.05 | 39.1 | — | — | — | 61.2 | 69.9 | 40 | 45.5 | — | — | 46.6 | 33.7 | 88.7 | 48.4 | 17.9 | 82.8 | 23.5 | 73.3 | 76.8 | 53.4 | — | — | |
| SSR-tokRetrieval Type=Late-interaction sparse retrieval2026.05 | 38.8 | — | — | — | 60.7 | 69.5 | 39.4 | 45.2 | — | — | 46.1 | 33.1 | 88.3 | 48.2 | 17.5 | 81.9 | 22.8 | 72.9 | 76.4 | 52.9 | — | — | |
| Splade-v3Retrieval Type=Learned sparse retrieval2026.05 | 35.8 | — | — | — | 58.6 | 69.2 | 37.5 | 43.9 | — | — | 50.8 | 29.6 | 81.4 | 45.2 | 15.8 | 79.8 | 23.3 | 71 | 74.8 | 51.2 | — | — | |
| XTRRetriever Deployment=One Retriever for All, Model Size=xxl2023.04 | 35.3 | — | 78.9 | — | 60.9 | 66.2 | 43.8 | — | — | — | 44.2 | — | 88.1 | 44.3 | 17.1 | 77 | 24.5 | 74.3 | — | 52.7 | — | 30.9 | |
| SPARevaluation_protocol=zero-shot, dense_backbone=GTR, lexical_component=UniCOIL A, training_dataset=MS MARCO2021.10 | 34 | — | 73.7 | — | 57 | 66.2 | 39.8 | — | — | — | 52.6 | 30.3 | 89.6 | 41.9 | 17.2 | 66.5 | 21.4 | 62.8 | 39.1 | 49.4 | — | — | |
| XTRRetriever Deployment=One Retriever for All, Model Size=base2023.04 | 34 | — | 73.6 | — | 53 | 64.7 | 34.7 | — | — | — | 40.7 | — | 86.1 | 40.9 | 14.5 | 73.7 | 20.7 | 71 | — | 49.1 | — | 31.3 | |
| AligneRRetrieval Type=Late-interaction dense retrieval2026.05 | 34 | — | — | — | 52.4 | 61.5 | 33.5 | 38.8 | — | — | 28.9 | 34.8 | 82.3 | 41.6 | 14.3 | 72.4 | 18.2 | 70.3 | 68.8 | 46.6 | — | — | |
| XTRRetrieval Type=Late-interaction dense retrieval2026.05 | 34 | — | — | — | 52.8 | 64.7 | 34.8 | 45 | — | — | 40.7 | 31.3 | 85.9 | 40.8 | 14.6 | 73.7 | 20.8 | 71.1 | 73.6 | 48.8 | — | — | |
| SPARevaluation_protocol=zero-shot, dense_backbone=Contriever, lexical_component=UniCOIL A, training_dataset=MS MARCO2021.10 | 33.8 | — | 73.5 | — | 53.1 | 67.6 | 33.7 | — | — | — | 48.8 | 27.5 | 87 | 42.2 | 17.1 | 81.2 | 24.5 | 68.3 | 36.5 | 49.6 | — | — | |
| PLAIDRetrieval Type=Late-interaction dense retrieval2026.05 | 33.8 | — | — | — | 56.4 | 66.7 | 36.3 | 39.7 | — | — | 46.1 | 26.3 | 85 | 44.7 | 15.4 | 78.8 | 17.4 | 69.2 | 73.7 | 49.3 | — | — | |
| SPARevaluation_protocol=zero-shot, dense_backbone=Contriever, lexical_component=BM25 A, training_dataset=MS MARCO2021.10 | 33.7 | — | 63 | — | 51.3 | 66.4 | 34.1 | — | — | — | 45.9 | 24.9 | 87.5 | 42.8 | 16.9 | 76.9 | 24.8 | 69.5 | 35.8 | 48.1 | — | — | |
| CITADELRetrieval Type=Late-interaction dense retrieval2026.05 | 33.7 | — | — | — | 54 | 66.3 | 33 | 39.9 | — | — | 51.1 | 34.2 | 85.3 | 42.2 | 15.9 | 76.6 | 18.3 | 69 | 71.5 | 49.4 | — | — | |
| ColBERTv2Retrieval Type=Late-interaction dense retrieval2026.05 | 33.5 | — | — | — | 56.3 | 66.6 | 36.1 | 39.8 | — | — | 46.3 | 26.1 | 85.3 | 44.9 | 15.2 | 78.7 | 17.3 | 69.1 | 73.5 | 49.2 | — | — | |
| Splade-v2Retrieval Type=Learned sparse retrieval2026.05 | 33.5 | — | — | — | 52.1 | 68.4 | 33.8 | 43.3 | — | — | 47.6 | 36.3 | 83.8 | 43.4 | 15.9 | 78.7 | 23.5 | 69.4 | 71.5 | 50.1 | — | — | |
| COILRetrieval Type=Late-interaction dense retrieval2026.05 | 33.2 | — | — | — | 52.1 | 71.3 | 31.3 | 35.3 | — | — | 29.5 | 28.1 | 83.8 | 39.9 | 15.5 | 83.8 | 21.7 | 70.7 | 66.8 | 47.4 | — | — | |
| GTR-largeevaluation_protocol=zero-shot, training_dataset=MS MARCO2021.10 | 32.9 | — | 55.7 | — | 54.7 | 57.9 | 42.4 | — | — | — | 52.5 | 21.9 | 89 | 39.1 | 15.8 | 71.2 | 26.2 | 63.9 | 38.4 | 47.2 | — | — | |
| docT5queryevaluation_protocol=zero-shot2021.10 | 32.8 | — | 71.3 | — | 39.9 | 58 | 29.1 | — | — | — | 34.9 | 34.7 | 80.2 | 33.1 | 16.2 | 71.4 | 20.1 | 67.5 | 32.5 | 44.4 | — | — | |
| Contrieverevaluation_protocol=zero-shot, training_dataset=MS MARCO2021.10 | 32.8 | — | 59.6 | — | 49.8 | 63.8 | 32.9 | — | — | — | 44.6 | 23 | 86.5 | 41.3 | 16.5 | 75.8 | 23.7 | 67.7 | 34.5 | 46.6 | — | — | |
| BM25evaluation_protocol=zero-shot2021.10 | 32.5 | — | 65.6 | — | 32.9 | 60.3 | 23.6 | — | — | — | 31.5 | 36.7 | 78.9 | 31.3 | 15.8 | 75.3 | 21.3 | 66.5 | 29.9 | 43 | — | — | |
| SPARevaluation_protocol=zero-shot, dense_backbone=GTR, lexical_component=BM25 A, training_dataset=MS MARCO2021.10 | 32.3 | — | 60.7 | — | 55.7 | 60.9 | 40.9 | — | — | — | 51.4 | 23.5 | 89.3 | 41.1 | 16.3 | 58 | 20.5 | 57.3 | 37.4 | 46.1 | — | — | |
| GenQRetriever Deployment=One Retriever per Domain2023.04 | 31.9 | — | 61.9 | — | 35.8 | 53.4 | 30.8 | — | — | — | 49.3 | — | 83 | 32.8 | 14.3 | 66.9 | 17.5 | 64.4 | — | 43.1 | — | 18.2 | |
| SPARevaluation_protocol=zero-shot, dense_backbone=ANCE, lexical_component=UniCOIL A, training_dataset=MS MARCO2021.10 | 31.8 | — | 76.4 | — | 51.1 | 63 | 32.3 | — | — | — | 47.2 | 30.3 | 86.9 | 36.1 | 15.5 | 80.2 | 23.5 | 62.6 | 35.8 | 48.1 | — | — | |
| GTR-largeevaluation_protocol=zero-shot, training_dataset=MS MARCO, repro=our reproduction2021.10 | 31.4 | — | 56.3 | — | 55.1 | 57.8 | 41.1 | — | — | — | 50.9 | 22 | 88.5 | 39.5 | 15.5 | 56.6 | 19.8 | 54 | 36.2 | 44.6 | — | — | |
| ColBERTRetrieval Type=Late-interaction dense retrieval2026.05 | 30.7 | — | — | — | 51.8 | 59.2 | 31.7 | 36 | — | — | 23.6 | 20.5 | 85.7 | 39.2 | 14.6 | 76.9 | 18.1 | 66.8 | 67.9 | 41.5 | — | — | |
| ColBERTevaluation_protocol=zero-shot, training_dataset=MS MARCO2021.10 | 30.5 | — | 67.7 | — | 52.4 | 59.3 | 31.7 | — | — | — | 23.3 | 20.2 | 85.4 | 39.2 | 14.5 | 77.1 | 18.4 | 67.1 | 35 | 44.4 | — | — | |
| SPARevaluation_protocol=zero-shot, dense_backbone=ANCE, lexical_component=BM25 Λ, training_dataset=MS MARCO2021.10 | 27.7 | — | 68.4 | — | 47.3 | 53.6 | 32.1 | — | — | — | 45 | 28.1 | 86.7 | 32.1 | 14.1 | 72.6 | 23.2 | 59.6 | 33.2 | 44.5 | — | — | |
| ANCEevaluation_protocol=zero-shot, training_dataset=MS MARCO2021.10 | 23.7 | — | 65.4 | — | 44.6 | 45.6 | 29.5 | — | — | — | 41.5 | 24 | 85.2 | 28.1 | 12.2 | 66.9 | 19.8 | 50.7 | 29.6 | 40.5 | — | — | |
| GTR-XLRetrieval Category=Dense, Model Scale=XL2021.12 | 0.343 | — | 0.584 | 0.317 | 0.559 | 0.591 | 0.444 | 0.268 | 0.35 | 0.479 | 0.531 | 0.23 | 0.89 | 0.396 | 0.159 | 0.717 | 0.27 | 0.635 | 0.388 | 0.452 | 0.453 | — | |
| GTR-XXLRetrieval Category=Dense, Model Scale=XXL2021.12 | 0.342 | — | 0.501 | 0.324 | 0.568 | 0.599 | 0.467 | 0.273 | 0.346 | 0.506 | 0.54 | 0.256 | 0.892 | 0.408 | 0.161 | 0.74 | 0.267 | 0.662 | 0.399 | 0.457 | 0.458 | — | |
| GTR-LargeRetrieval Category=Dense, Model Scale=Large2021.12 | 0.329 | — | 0.557 | 0.32 | 0.547 | 0.579 | 0.424 | 0.265 | 0.343 | 0.47 | 0.525 | 0.219 | 0.89 | 0.391 | 0.158 | 0.712 | 0.262 | 0.639 | 0.384 | 0.444 | 0.445 | — | |
| docT5queryRetrieval Category=Lexical / Sparse2021.12 | 0.328 | — | 0.713 | 0.431 | 0.399 | 0.58 | 0.291 | 0.307 | 0.42 | 0.437 | 0.349 | 0.347 | 0.802 | 0.331 | 0.162 | 0.714 | 0.201 | 0.675 | 0.325 | 0.429 | 0.434 | — | |
| BM25Retrieval Category=Lexical / Sparse2021.12 | 0.325 | — | 0.656 | 0.465 | 0.329 | 0.603 | 0.236 | 0.33 | 0.398 | 0.408 | 0.315 | 0.367 | 0.789 | 0.313 | 0.158 | 0.753 | 0.213 | 0.665 | 0.299 | 0.413 | 0.423 | — | |
| TAS-BRetrieval Category=Dense2021.12 | 0.319 | — | 0.481 | 0.383 | 0.463 | 0.584 | 0.3 | 0.289 | 0.377 | 0.427 | 0.429 | 0.162 | 0.835 | 0.384 | 0.149 | 0.7 | 0.228 | 0.643 | 0.314 | 0.414 | 0.415 | — | |
| GenQRetrieval Category=Dense2021.12 | 0.319 | — | 0.619 | 0.398 | 0.358 | 0.534 | 0.308 | 0.281 | 0.396 | 0.362 | 0.493 | 0.182 | 0.83 | 0.328 | 0.143 | 0.669 | 0.175 | 0.644 | 0.347 | 0.41 | 0.41 | — | |
| GTR-BaseRetrieval Category=Dense, Model Scale=Base2021.12 | 0.308 | — | 0.539 | 0.271 | 0.495 | 0.535 | 0.349 | 0.261 | 0.337 | 0.437 | 0.511 | 0.205 | 0.881 | 0.347 | 0.149 | 0.66 | 0.241 | 0.6 | 0.357 | 0.416 | 0.416 | — | |
| ColBERTRetrieval Category=Dense2021.12 | 0.305 | — | 0.677 | 0.474 | 0.524 | 0.593 | 0.317 | 0.274 | 0.393 | 0.391 | 0.233 | 0.202 | 0.854 | 0.392 | 0.145 | 0.771 | 0.184 | 0.671 | 0.35 | 0.429 | 0.431 | — | |
| ANCERetrieval Category=Dense2021.12 | 0.237 | — | 0.654 | 0.306 | 0.446 | 0.456 | 0.295 | 0.249 | 0.382 | 0.392 | 0.415 | 0.24 | 0.852 | 0.281 | 0.122 | 0.669 | 0.198 | 0.507 | 0.296 | 0.389 | 0.389 | — | |
| DPRRetrieval Category=Dense2021.12 | 0.189 | — | 0.332 | 0.127 | 0.474 | 0.391 | 0.112 | 0.155 | 0.161 | 0.252 | 0.175 | 0.131 | 0.248 | 0.263 | 0.077 | 0.562 | 0.148 | 0.318 | 0.153 | 0.234 | 0.237 | — |