Extractive Question Answering on TriviaQA MRQA
78.2F1 ScoreLinkBERT
Evaluation results
| Method | Links | |
|---|---|---|
| LinkBERTModel Scale=large2022.03 | 78.2 | |
| SpanBERTNumber of training examples=Full Dataset2021.01 | 77.2 | |
| SplinterNumber of training examples=Full Dataset2021.01 | 76.5 | |
| SpanBERT (Reimpl)Number of training examples=Full Dataset2021.01 | 75.8 | |
| RoBERTaNumber of training examples=Full Dataset2021.01 | 74 | |
| LinkBERTModel Scale=base2022.03 | 73.9 | |
| BERTModel Scale=large2022.03 | 73.7 | |
| BERTModel Scale=base2022.03 | 70.3 | |
| SplinterNumber of training examples=10242021.01 | 64.8 | |
| SpanBERT (Reimpl)Number of training examples=10242021.01 | 55.5 | |
| SpanBERTNumber of training examples=10242021.01 | 50.3 | |
| LinkBERTModel Scale=tiny2022.03 | 50 | |
| RoBERTaNumber of training examples=10242021.01 | 46.8 | |
| SplinterNumber of training examples=1282021.01 | 44.7 | |
| BERTModel Scale=tiny2022.03 | 43.4 | |
| SpanBERT (Reimpl)Number of training examples=1282021.01 | 26.3 | |
| SpanBERTNumber of training examples=1282021.01 | 24.2 | |
| RoBERTaNumber of training examples=1282021.01 | 19.1 | |
| SplinterNumber of training examples=162021.01 | 18.9 | |
| SpanBERTNumber of training examples=162021.01 | 12.8 | |
| SpanBERT (Reimpl)Number of training examples=162021.01 | 11.6 | |
| RoBERTaNumber of training examples=162021.01 | 7.5 |