Document Reranking on BEIR (test)
0.378NDCG@10 (NFCorpus)MonoT5-base
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MonoT5-baseEvaluation Protocol=Supervised Reranking2026.05 | 0.378 | 0.154 | 0.376 | 0.476 | 0.796 | 0.675 | 0.476 | 3.1 | 600 | — | — | |
| DARTEvaluation Protocol=Test-time Adaptation2026.05 | 0.354 | 0.205 | 0.389 | 0.605 | 0.67 | 0.719 | 0.49 | 2.1 | 10 | — | — | |
| InstUPREvaluation Protocol=Training-free Reranking2026.05 | 0.352 | 0.19 | 0.398 | — | 0.73 | 0.713 | 0.477 | 2.6 | 200 | — | — | |
| ICREvaluation Protocol=Training-free Reranking2026.05 | 0.347 | 0.171 | 0.381 | — | 0.728 | 0.761 | 0.478 | 0.8 | 200 | — | — | |
| PRF-VecEvaluation Protocol=Test-time Adaptation, n (Number of pseudo-documents)=32026.05 | 0.347 | 0.203 | 0.371 | 0.602 | 0.663 | 0.71 | 0.483 | 0.3 | 2 | — | — | |
| ASRankEvaluation Protocol=Training-free Reranking2026.05 | 0.346 | 0.184 | 0.352 | 0.478 | 0.737 | 0.71 | 0.468 | 3.8 | 200 | — | — | |
| PRF-VecEvaluation Protocol=Test-time Adaptation, n (Number of pseudo-documents)=52026.05 | 0.341 | 0.201 | 0.362 | 0.585 | 0.671 | 0.704 | 0.477 | 1 | 1 | — | — | |
| ColBERTv2Evaluation Protocol=Supervised Dense Retrieval2026.05 | 0.338 | 0.154 | 0.356 | 0.463 | 0.738 | 0.693 | 0.457 | 7.3 | 80 | — | — | |
| BM252026.06 | 0.338 | 0.149 | 0.236 | — | 0.595 | 0.679 | 0.354 | — | — | 0.165 | 0.318 | |
| Dense Retrieval (BGE-small)Evaluation Protocol=Training-free Reranking2026.05 | 0.337 | 0.197 | 0.385 | 0.595 | 0.665 | 0.72 | 0.483 | 0 | 1 | — | — | |
| SPLADEv2Evaluation Protocol=Supervised Dense Retrieval2026.05 | 0.334 | 0.158 | 0.336 | 0.479 | 0.71 | 0.693 | 0.452 | 8.3 | — | — | — | |
| ICRBackbone=Mistral-7B2026.06 | 0.332 | 0.162 | 0.31 | — | 0.639 | 0.724 | 0.384 | — | — | 0.206 | 0.314 | |
| ICRBackbone=Qwen2.5-7B2026.06 | 0.327 | 0.164 | 0.27 | — | 0.661 | 0.711 | 0.383 | — | — | 0.196 | 0.353 | |
| CoreheadBackbone=Qwen2.5-7B2026.06 | 0.325 | 0.145 | 0.306 | — | 0.677 | 0.7 | 0.381 | — | — | 0.193 | 0.324 | |
| TAS-BEvaluation Protocol=Supervised Dense Retrieval2026.05 | 0.319 | 0.149 | 0.3 | 0.427 | 0.481 | 0.643 | 0.387 | 19.7 | — | — | — | |
| CompRankBackbone=Mistral-7B2026.06 | 0.317 | 0.143 | 0.317 | — | 0.812 | 0.642 | 0.392 | — | — | 0.166 | 0.348 | |
| RankGPTBackbone=Mistral-7B2026.06 | 0.31 | 0.147 | 0.241 | — | 0.606 | 0.655 | 0.351 | — | — | 0.167 | 0.329 | |
| ColBERTEvaluation Protocol=Supervised Dense Retrieval2026.05 | 0.305 | 0.145 | 0.317 | 0.233 | 0.677 | 0.671 | 0.391 | 19.9 | — | — | — | |
| BM25 RerankEvaluation Protocol=Training-free Reranking2026.05 | 0.302 | 0.156 | 0.22 | 0.371 | 0.685 | 0.588 | 0.387 | 21.2 | 2 | — | — | |
| RocketQAv2Evaluation Protocol=Supervised Dense Retrieval2026.05 | 0.293 | 0.131 | 0.302 | 0.451 | 0.675 | 0.568 | 0.403 | 18.7 | — | — | — | |
| MoDIREvaluation Protocol=Supervised Dense Retrieval2026.05 | 0.244 | 0.124 | 0.296 | 0.418 | 0.676 | 0.502 | 0.377 | 24.4 | — | — | — | |
| ANCEEvaluation Protocol=Supervised Dense Retrieval2026.05 | 0.237 | 0.122 | 0.295 | 0.415 | 0.654 | 0.507 | 0.372 | 25.4 | — | — | — | |
| DPR-MSMARCOEvaluation Protocol=Supervised Dense Retrieval2026.05 | 0.208 | 0.108 | 0.275 | 0.414 | 0.561 | 0.478 | 0.341 | 31.9 | — | — | — |