Top-level Implicit Discourse Relation Recognition on CoNLL16 (blind)
72.33F1 ScoreDiscoPrompt
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DiscoPromptBackbone=T5, Model Size=11b2023.05 | 72.33 | 84.94 | |
| DiscoPromptBackbone=T5, Model Size=large2023.05 | 67.89 | 80.47 | |
| PCPBackbone=RoBERTa, Model Size=large2023.05 | 64.74 | 76.47 | |
| Prompt-TuningBackbone=T5, Model Size=large2023.05 | 63.35 | 77.41 | |
| DiscoPromptBackbone=T5, Model Size=base2023.05 | 62.98 | 76.94 | |
| BMGF-ROBERTaBackbone=RoBERTa2023.05 | 58.3 | 74.43 | |
| Fine-TuningBackbone=T5, Model Size=large2023.05 | 56.28 | 73.07 | |
| Prefix-TuningBackbone=T5, Model Size=large2023.05 | 55.61 | 74.12 | |
| PCPBackbone=RoBERTa, Model Size=base2023.05 | 55.3 | 72 | |
| Fine-TuningBackbone=T5, Model Size=base2023.05 | 50.94 | 71.3 | |
| Prompt-TuningBackbone=T5, Model Size=base2023.05 | 27.21 | 64.71 | |
| Prefix-TuningBackbone=T5, Model Size=base2023.05 | 27.17 | 65.88 | |
| XLNetBackbone=XLNet, Model Size=large, Case=cased2023.05 | 22.37 | 66.59 | |
| XLNetBackbone=XLNet, Model Size=base, Case=cased2023.05 | 19.9 | 66.12 |