Answer Sentence Selection on WikiQA clean (test)
92MAPTANDA (RoBERTa-Large)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TANDA (RoBERTa-Large)Backbone=RoBERTa-Large, Transfer Source=ASNQ, Adaptation Target=WikiQA2019.11 | 92 | 93.3 | |
| RoBERTa-Large (Fine-Tuning on ASNQ)Backbone=RoBERTa-Large, Training Protocol=Fine-Tuning, Training Dataset=ASNQ2019.11 | 91 | 91.9 | |
| TANDA (BERT-Large)Backbone=BERT-Large, Transfer Source=ASNQ, Adaptation Target=WikiQA2019.11 | 90.4 | 91.2 | |
| TANDA (BERT-Base)Backbone=BERT-Base, Transfer Source=ASNQ, Adaptation Target=WikiQA2019.11 | 89.3 | 90.3 | |
| BERT-Large (Fine-Tuning on ASNQ)Backbone=BERT-Large, Training Protocol=Fine-Tuning, Training Dataset=ASNQ2019.11 | 89.2 | 90.4 | |
| TANDA (RoBERTa-Base)Backbone=RoBERTa-Base, Transfer Source=ASNQ, Adaptation Target=WikiQA2019.11 | 88.9 | 90.1 | |
| BERT-Base (Fine-Tuning on ASNQ)Backbone=BERT-Base, Training Protocol=Fine-Tuning, Training Dataset=ASNQ2019.11 | 88.4 | 89.8 | |
| RoBERTa-Base (Fine-Tuning on ASNQ)Backbone=RoBERTa-Base, Training Protocol=Fine-Tuning, Training Dataset=ASNQ2019.11 | 88.2 | 89.4 | |
| BERT-Large (Fine-Tuning on WikiQA)Backbone=BERT-Large, Training Protocol=Fine-Tuning, Training Dataset=WikiQA2019.11 | 83.6 | 85.3 | |
| Comp-Agg + LM + LC + TL(QNLI)Backbone=Compare-Aggregate, Transfer Source=QNLI2019.11 | 83.1 | 81.8 | |
| BERT-Base (Fine-Tuning on WikiQA)Backbone=BERT-Base, Training Protocol=Fine-Tuning, Training Dataset=WikiQA2019.11 | 81.3 | 82.8 | |
| Comp-Agg + LM + LCBackbone=Compare-Aggregate, Components=Language Modeling, Latent Clustering2019.11 | 76.4 | 78.4 |