Stance Detection on RumorEval S (val)
59.2Macro F1JSDRV
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| JSDRVBackbone=Llama 2, Split=Val, Seeding claims percentage=50%2024.06 | 59.2 | 76.3 | — | — | — | — | 76.5 | 48.6 | 51 | 61.2 | |
| JSDRV-BertBackbone=Bert, Split=Val, Seeding claims percentage=50%2024.06 | 58 | 75.2 | — | — | — | — | 75.8 | 48.8 | 50 | 57.4 | |
| JSDRVData Split=T162024.06 | 57.4 | 69.7 | 53.6 | 38 | 58 | 79.8 | — | — | — | — | |
| JSDRV-BertBackbone=BERT, Data Split=T162024.06 | 56 | 68.1 | 52.7 | 38.1 | 57.3 | 75.9 | — | — | — | — | |
| JSDRVData Split=Val2024.06 | 55 | 67.2 | 53.6 | 31 | 57.6 | 77.9 | — | — | — | — | |
| JSDRV-BertBackbone=BERT, Data Split=Val2024.06 | 54.1 | 66.8 | 53.1 | 31.6 | 56.2 | 75.5 | — | — | — | — | |
| JointCLData Split=Val2024.06 | 50.5 | 63.9 | 53.2 | 21 | 51.6 | 76 | — | — | — | — | |
| Llama 2-STTraining Mode=Zero-shot2024.06 | 50 | 63 | 50.1 | 20.3 | 53.2 | 76.3 | — | — | — | — | |
| Llama 2-MTTraining Mode=Zero-shot2024.06 | 50 | 63.2 | 50.2 | 19.9 | 53.3 | 76.6 | — | — | — | — | |
| BerTweetTraining Mode=Fine-tuned2024.06 | 49.2 | 61.9 | 49.7 | 20.3 | 51.3 | 75.3 | — | — | — | — | |
| MT-GRUSplit=Val2024.06 | 45.2 | 76.8 | — | — | — | — | 66.2 | 29.8 | 37.3 | 45.7 | |
| MT-GRUData Split=Val2024.06 | 43.2 | 63.6 | 31.3 | 15.6 | 50.6 | 74.8 | — | — | — | — | |
| MTL2Split=Val2024.06 | 43 | 65.3 | — | — | — | — | 62.2 | 27.9 | 35.2 | 45.7 | |
| BrLSTMData Split=Val2024.06 | 42 | 66 | 46 | 0 | 39.1 | 75.8 | — | — | — | — | |
| BIGRUData Split=Val2024.06 | 41.7 | 63 | 39.2 | 16.2 | 36 | 75.4 | — | — | — | — | |
| TGATraining Mode=Zero-shot2024.06 | 32.4 | — | 30.1 | 16.8 | 34.2 | 48.6 | — | — | — | — |