Mathematical Information Retrieval on NTCIR Wiki-Formula 12
0.7BPref (Full)Approach0 + DPR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Approach0 + DPRstrategy=end-to-end fusion2022.03 | 0.7 | 0.62 | — | |
| Approach0 + ColBERTstrategy=end-to-end fusion2022.03 | 0.65 | 0.63 | — | |
| Tangent-Stype=Previous system2022.03 | 0.64 | 0.59 | — | |
| GNNtype=Previous system2022.03 | 0.63 | 0.54 | — | |
| Approach0type=Structure search baseline2022.03 | 0.63 | 0.54 | — | |
| MathBERTEnsemble system=true2022.03 | 0.614 | 0.736 | — | |
| MathBERTtype=Previous system2022.03 | 0.61 | 0.74 | — | |
| Tangent-CFTtype=Previous system2022.03 | 0.6 | 0.71 | — | |
| MCATtype=Previous system2022.03 | 0.57 | 0.57 | — | |
| Approach0 + ColBERTstrategy=reranking2022.03 | 0.56 | 0.52 | — | |
| ColBERTTraining stage=fully-trained2022.03 | 0.545 | 0.483 | 25.1 | |
| Approach0 + DPRstrategy=reranking2022.03 | 0.53 | 0.48 | — | |
| further-pretrained BERTPre-trained backbone=BERT, Fine-tuning epochs=1, Further-pretraining epochs=72022.03 | 0.522 | 0.439 | 23.2 | |
| DPRTraining stage=fully-trained2022.03 | 0.516 | 0.427 | 23.4 | |
| SciBERTPre-trained backbone=SciBERT, Fine-tuning epochs=12022.03 | 0.512 | 0.363 | 21.1 | |
| BERTPre-trained backbone=BERT, Fine-tuning epochs=12022.03 | 0.505 | 0.393 | 21.9 | |
| further-pretrained BERTPre-trained backbone=BERT, Fine-tuning epochs=1, Further-pretraining epochs=32022.03 | 0.486 | 0.392 | 21.8 |