Natural Language Understanding on GLUE (val) - Subset (SST-2, QNLI, QQP, MNLI)
94.61SST-2 AccuracyRoLoRA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RoLoRABackbone=RoBERTa-Large, Rank=2, Parameter Count=34K, Data Distribution=severe heterogeneity2024.10 | 94.61 | 89.83 | 85.15 | 85.55 | 88.78 | |
| FedTT+Backbone=RoBERTa-Large, Parameter Count=28K, Data Distribution=severe heterogeneity2024.10 | 94.5 | 90.17 | 86.65 | 86.28 | 89.4 | |
| FedTTBackbone=RoBERTa-Large, Parameter Count=51K, Data Distribution=severe heterogeneity2024.10 | 94.38 | 90.55 | 85.47 | 85.27 | 88.92 | |
| LoRABackbone=RoBERTa-Large, Rank=2, Parameter Count=68K, Data Distribution=severe heterogeneity2024.10 | 93.23 | 82.57 | 58.96 | 76.96 | 77.93 | |
| FFA-LoRABackbone=RoBERTa-Large, Rank=2, Parameter Count=34K, Data Distribution=severe heterogeneity2024.10 | 92.32 | 85.15 | 62.79 | 77.78 | 79.51 |