Skip to Main Content
Table 2: 

Test set first-pass retrieval results on the passage and document TREC 2019 DL evaluation as well as ms marco eval MRR@10 (passage) and MRR@100 (document) under MRR(MS).

ModelMRR(MS)RRNDCG@10Holes@10
Passage Retrieval 
BM25-Anserini 0.186 0.825 0.506 0.000 
de-bert 0.295 0.936 0.639 0.165 
me-bert 0.323 0.968 0.687 0.109 
de-hybrid-e 0.306 0.951 0.659 0.105 
me-hybrid-e 0.336 0.977 0.706 0.051 
Document Retrieval 
Base-Indri 0.192 0.785 0.517 0.002 
de-bert 0.841 0.510 0.188 
me-bert 0.877 0.588 0.109 
de-hybrid-e 0.287 0.890 0.595 0.084 
me-hybrid-e 0.310 0.914 0.610 0.063 
ModelMRR(MS)RRNDCG@10Holes@10
Passage Retrieval 
BM25-Anserini 0.186 0.825 0.506 0.000 
de-bert 0.295 0.936 0.639 0.165 
me-bert 0.323 0.968 0.687 0.109 
de-hybrid-e 0.306 0.951 0.659 0.105 
me-hybrid-e 0.336 0.977 0.706 0.051 
Document Retrieval 
Base-Indri 0.192 0.785 0.517 0.002 
de-bert 0.841 0.510 0.188 
me-bert 0.877 0.588 0.109 
de-hybrid-e 0.287 0.890 0.595 0.084 
me-hybrid-e 0.310 0.914 0.610 0.063 
Close Modal

or Create an Account

Close Modal
Close Modal