Open-domain Question Answering on Quasar-T (test)
63.9F1 ScoreCluster-Former
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Cluster-Formernumber of clusters=5122020.09 | 63.9 | 54 | |
| Cluster-Formernumber of clusters=2562020.09 | 63.5 | 53.6 | |
| Cluster-Formernumber of clusters=642020.09 | 63.3 | 53.3 | |
| Locality-Sensitive Hashing2020.09 | 62.9 | 53.2 | |
| Sliding Window2020.09 | 62.8 | 52.9 | |
| Sparse Attention2020.09 | 62 | 52.1 | |
| Human Performance2017.11 | 60.6 | 51.5 | |
| Multi-Passage BERTBackbone=BERT-Large2019.08 | 59.1 | 51.1 | |
| Multi passage BERT2020.09 | 59.1 | 51.1 | |
| Multi-Passage BERTBackbone=BERT-Base2019.08 | 59 | 51.3 | |
| BERT-RC2019.08 | 56.8 | 49.7 | |
| TraCRNet2019.08 | 54 | 43.2 | |
| Full Re-Rankercombination=three different re-rankers2017.11 | 49.6 | 42.3 | |
| OpenQA2019.08 | 49.3 | 42.2 | |
| DS-QA2020.09 | 49.3 | 42.2 | |
| Coverage-Based Re-Ranker2017.11 | 49.1 | 40.6 | |
| HAS-QA2019.08 | 48.9 | 43.2 | |
| DECAPROP2020.09 | 46.9 | 38.6 | |
| Strength-Based Re-Rankervariant=Counting2017.11 | 46.7 | 37.1 | |
| Baseline Re-Ranker (BM25)2017.11 | 45.2 | 33.6 | |
| DrQA2019.08 | 44.5 | 37.7 | |
| DrQA2020.09 | 44.5 | 37.7 | |
| Strength-Based Re-Rankervariant=Probability2017.11 | 42.4 | 36.1 | |
| R³2017.11 | 41.7 | 35.3 | |
| R³2019.08 | 41.7 | 35.3 | |
| R32020.09 | 41.7 | 35.3 | |
| Reinforced Ranker-Reader2017.08 | 40.9 | 34.2 | |
| Simple Ranker-Reader2017.08 | 38.8 | 31.9 | |
| Single Reader2017.08 | 38.5 | 31.5 | |
| BIDAF2017.08 | 28.5 | 25.9 | |
| BiDAF2017.11 | 28.5 | 25.9 | |
| GA2017.08 | 26.4 | 26.4 | |
| GA2017.11 | 26.4 | 26.4 |