Document Reranking on Robust04 Title
38.07MAPPARADE-CNN
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PARADE-CNNBackbone=ELECTRA, Aggregation strategy=CNN, Evaluation protocol=Fine-tuned2020.08 | 38.07 | 48.21 | 56.25 | |
| PARADE-TransformerBackbone=ELECTRA, Aggregation strategy=Transformer, Evaluation protocol=Fine-tuned2020.08 | 38.03 | 49.2 | 56.59 | |
| BirchBackbone=BERT, Evaluation protocol=Fine-tuned2020.08 | 37.63 | 47.49 | 54.54 | |
| PARADE-MaxBackbone=ELECTRA, Aggregation strategy=Max, Evaluation protocol=Fine-tuned2020.08 | 37.11 | 47.23 | 54.42 | |
| CEDR-KNRM (Max)Backbone=ELECTRA, Aggregation strategy=Max-pooling2020.08 | 37.01 | 47.69 | 54.75 | |
| ELECTRA-KNRMBackbone=ELECTRA, Kernel size=112020.08 | 36.73 | 47.55 | 54.7 | |
| PARADE-SumBackbone=ELECTRA, Aggregation strategy=Sum, Evaluation protocol=Fine-tuned2020.08 | 35.26 | 47.11 | 53.85 | |
| PARADE-AttnBackbone=ELECTRA, Aggregation strategy=Attention, Evaluation protocol=Fine-tuned2020.08 | 34.62 | 45.76 | 52.66 | |
| PARADE-AvgBackbone=ELECTRA, Aggregation strategy=Average, Evaluation protocol=Fine-tuned2020.08 | 33.52 | 44.64 | 51.24 | |
| ELECTRA-MaxPBackbone=ELECTRA, Pre-training dataset=MSMARCO, Evaluation protocol=Fine-tuned2020.08 | 31.83 | 43.37 | 49.59 | |
| BM25+RM3Model Type=Query Expansion2020.08 | 30.33 | 39.74 | 45.14 | |
| BM25Model Type=Unsupervised2020.08 | 25.31 | 36.31 | 42.4 |