Crisis Classification on CrisisOltea v1 (test)
96.01Macro F1InfoD-R
Evaluation results
| Method | Links | |
|---|---|---|
| InfoD-RBackbone=RoBERTa, Training=Proposed InfoDCL framework2022.03 | 96.01 | |
| Sim-DBackbone=RoBERTa, Training=SimCSE-Distant trained with distantly supervised positive pairs and SSCL loss2022.03 | 95.94 | |
| DCLBackbone=RoBERTa, Training=Trained with LDCL only2022.03 | 95.92 | |
| E-MLMBackbone=RoBERTa, Training=Emoji-based MLM2022.03 | 95.91 | |
| SLPBackbone=RoBERTa, Training=Surrogate label prediction2022.03 | 95.89 | |
| SCLBackbone=RoBERTa, Training=Supervised contrastive loss2022.03 | 95.88 | |
| RBBackbone=RoBERTa, Training=Fine-tuning on original pre-trained2022.03 | 95.87 | |
| LCLBackbone=RoBERTa, Training=label-aware contrastive loss2022.03 | 95.87 | |
| InfoD-BBackbone=BERTweet, Training=Proposed InfoDCL framework2022.03 | 95.84 | |
| WCLBackbone=RoBERTa, Training=Weakly-supervised contrastive learning2022.03 | 95.83 | |
| MLMBackbone=RoBERTa, Training=Further pre-training with MLM objective2022.03 | 95.81 | |
| Mir-BBackbone=RoBERTa, Training=Mirror-BERT2022.03 | 95.79 | |
| BTwBackbone=BERTweet, Training=Original pre-trained2022.03 | 95.76 | |
| Sim-SBackbone=RoBERTa, Training=SimCSE-Self2022.03 | 95.71 |