Answer Selection 벤치마크
Answer Selection on ASNQ
MAP
- 2022-05-02 — RoBERTa-Base Joint MSPP: MAP 0.673
- 2022-05-20 — DeBERTa-V3-Large + SSP: MAP 0.743
| Rank | Model | MAP | MRR | Paper | Code | Year |
|---|---|---|---|---|---|---|
| 1 | DeBERTa-V3-Large + SSP | 0.743 | 0.800 | Pre-training Transformer Models with Sentence-Level Objectives for Answer Sentence Selection | 2022 | |
| 2 | ELECTRA-Base + SSP | 0.697 | 0.757 | Pre-training Transformer Models with Sentence-Level Objectives for Answer Sentence Selection | 2022 | |
| 3 | RoBERTa-Base Joint MSPP | 0.673 | 0.737 | Paragraph-based Transformer Pre-training for Multi-Sentence Inference | amazon-research/wqa-multi-sentence-inference | 2022 |