Text Classification on Task represented in Figure 4
95.5AccuracyMultitask Finetuning
Evaluation results
| Method | Links | |
|---|---|---|
| Multitask FinetuningModel=RoBERTa, Warm-start weights=Hessian Weighted2024.12 | 95.5 | |
| Multitask FinetuningModel=RoBERTa, Selection=Global maximum2024.12 | 95.5 | |
| Multitask FinetuningModel=RoBERTa, Warm-start weights=Mixture Weighted2024.12 | 95.4 | |
| Multitask FinetuningModel=RoBERTa, Warm-start weights=Simple Merging2024.12 | 95.2 | |
| Mixture WeightedModel=RoBERTa2024.12 | 94.1 | |
| Hessian WeightedModel=RoBERTa2024.12 | 93.6 | |
| Simple MergingModel=RoBERTa2024.12 | 93.5 |