Natural Language Processing on Federated Dataset Personalization 2
90.5Paraphrasing AccuracyFedDPA-T
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| FedDPA-TAdapter=local training adapter2024.03 | 90.5 | 70.54 | 82 | 91.81 | 30.75 | 81 | 75 | 91.07 | 75.33 | |
| FedDPA-FAdapter=local fine-tuning adapter2024.03 | 88 | 64.8 | 84.25 | 89.82 | 29.58 | 78.5 | 72 | 80.89 | 73.48 | |
| Centralized2024.03 | 87 | 64.67 | 77 | 90.65 | 29.12 | 76 | 72.5 | 76.17 | 71.64 | |
| FedLORA2024.03 | 87 | 64.12 | 84.5 | 89.52 | 27.13 | 76.5 | 73.5 | 79.62 | 72.74 | |
| FedIT2024.03 | 86 | 63.13 | 79 | 89.8 | 30.36 | 75.5 | 72 | 81.06 | 72.07 | |
| Local-finetuned2024.03 | 75 | 53.51 | 81 | 91.28 | 27.51 | 69 | 72.5 | 79.31 | 68.64 |