Second-level Implicit Discourse Relation Recognition on PDTB 2.0 (Lin split)
67.62AccuracyDiscoPrompt (T5-11b)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DiscoPrompt (T5-11b)Backbone=T5-11b2023.05 | 67.62 | 47.18 | |
| DiscoPromptBackbone=T5, Model Scale=11b2023.05 | 67.62 | 47.18 | |
| DiscoPrompt (T5-large)Backbone=T5-large2023.05 | 63.05 | 45.25 | |
| DiscoPromptBackbone=T5, Model Scale=large2023.05 | 63.05 | 45.25 | |
| OTMT (XLNet-large)Backbone=XLNet-large2023.05 | 61.62 | — | |
| OTMTBackbone=XLNet-large2023.05 | 61.62 | — | |
| Prompt-Tuning (T5-large)Backbone=T5-large, Training Strategy=Prompt-Tuning2023.05 | 60.44 | 40.75 | |
| PCP w/ ROBERTa-largeBackbone=RoBERTa-large2023.05 | 60.44 | 43.04 | |
| Prompt-TuningBackbone=T5, Model Scale=large2023.05 | 60.44 | 40.75 | |
| PCPBackbone=RoBERTa-large2023.05 | 60.44 | 43.04 | |
| Fine-Tuning (T5-large)Backbone=T5-large, Training Strategy=Fine-Tuning2023.05 | 59.4 | 42.04 | |
| Fine-TuningBackbone=T5, Model Scale=large2023.05 | 59.4 | 42.04 | |
| DiscoPrompt (T5-base)Backbone=T5-base2023.05 | 59.27 | 41.82 | |
| DiscoPromptBackbone=T5, Model Scale=base2023.05 | 59.27 | 41.82 | |
| XLNet (large, cased)Backbone=XLNet-large2023.05 | 58.77 | 40.71 | |
| XLNetBackbone=XLNet, Model Scale=large, Casing=cased2023.05 | 58.77 | 40.71 | |
| BERT-largeBackbone=BERT-large2023.05 | 56.53 | 38.68 | |
| Prefix-Tuning (T5-large)Backbone=T5-large, Training Strategy=Prefix-Tuning2023.05 | 56.53 | 36.86 | |
| Prefix-TuningBackbone=T5, Model Scale=large2023.05 | 56.53 | 36.86 | |
| OTMT (XLNet-base)Backbone=XLNet-base2023.05 | 56.37 | — | |
| OTMTBackbone=XLNet-base2023.05 | 56.37 | — | |
| PCP w/ ROBERTa-baseBackbone=RoBERTa-base2023.05 | 56.14 | 41.19 | |
| PCPBackbone=RoBERTa-base2023.05 | 56.14 | 41.19 | |
| XLNet(base, cased)Backbone=XLNet-base2023.05 | 55.82 | 36.23 | |
| XLNetBackbone=XLNet, Model Scale=base, Casing=cased2023.05 | 55.82 | 36.23 | |
| Liu et al.2023.05 | 53.96 | 39.15 | |
| BMGF-ROBERTaBackbone=RoBERTa2023.05 | 53.96 | 39.15 | |
| Fine-Tuning (T5-base)Backbone=T5-base, Training Strategy=Fine-Tuning2023.05 | 51.96 | 36.49 | |
| Fine-TuningBackbone=T5, Model Scale=base2023.05 | 51.96 | 36.49 | |
| BERT-baseBackbone=BERT-base2023.05 | 49.87 | 26.7 | |
| Nguyen et al.2023.05 | 46.48 | — | |
| MTL-MLoss2023.05 | 46.48 | — | |
| Shi and Demberg2023.05 | 45.82 | — | |
| Bai and Zhao2023.05 | 45.73 | — | |
| Qin et al.2023.05 | 44.65 | — | |
| Qin et al.2023.05 | 43.81 | — | |
| Lin et al.2023.05 | 40.2 | — | |
| Prompt-Tuning (T5-base)Backbone=T5-base, Training Strategy=Prompt-Tuning2023.05 | 27.68 | 8.97 | |
| Prompt-TuningBackbone=T5, Model Scale=base2023.05 | 27.68 | 8.97 | |
| Prefix-Tuning (T5-base)Backbone=T5-base, Training Strategy=Prefix-Tuning2023.05 | 26.37 | 8.45 | |
| Prefix-TuningBackbone=T5, Model Scale=base2023.05 | 26.37 | 8.45 |