Hate Speech Detection on HatEval -> GHC (test)
0.642Source F1Source model
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Source modelBackbone=BERT-Base2021.03 | 0.642 | 0.295 | 115.6 | |
| REMOTE (Rs)Backbone=BERT-Base2021.03 | 0.635 | 0.399 | 56.2 | |
| REMOTE (Rs + Ch)Backbone=BERT-Base2021.03 | 0.632 | 0.466 | 49 | |
| REMOTE (Rsoft)Backbone=Bi-LSTM+Attention2021.03 | 0.611 | 0.27 | 69.3 | |
| Distillation (C_sample)Backbone=BERT-Base2021.03 | 0.608 | 0.424 | 315.4 | |
| Fine-tune (C_sample)Backbone=Bi-LSTM+Attention2021.03 | 0.608 | 0.237 | 71.9 | |
| Source modelBackbone=Bi-LSTM+Attention2021.03 | 0.605 | 0.202 | 115.2 | |
| L2-reg (C_sample)Backbone=Bi-LSTM+Attention2021.03 | 0.604 | 0.242 | 21.3 | |
| Distillation (C_sample)Backbone=Bi-LSTM+Attention2021.03 | 0.601 | 0.243 | 16.4 | |
| Fine-tune (C_all)Backbone=BERT-Base2021.03 | 0.6 | 0.515 | 333.3 | |
| L2-reg (C_sample)Backbone=BERT-Base2021.03 | 0.598 | 0.413 | 278.7 | |
| REMOTE (Rsoft + Cstrict)Backbone=Bi-LSTM+Attention2021.03 | 0.587 | 0.313 | 7.4 | |
| Fine-tune (C_all)Backbone=Bi-LSTM+Attention2021.03 | 0.585 | 0.385 | 124.2 | |
| Fine-tune (C_sample)Backbone=BERT-Base2021.03 | 0.58 | 0.41 | 302.6 |