Multi-task NLP on FLAN homogeneous federated setting (test)
84.3Average ROUGE-1Fed-ALoRA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Fed-ALoRABackbone=Qwen2-7B2025.09 | 84.3 | -2.05 | 3.21 | |
| FedSA-LoRABackbone=Qwen2-7B2025.09 | 83.69 | -1.15 | 3.21 | |
| FedDPABackbone=Qwen2-7B2025.09 | 83.17 | -0.6 | 12.85 | |
| FedITBackbone=Qwen2-7B2025.09 | 82.8 | -0.05 | 6.42 | |
| ST BaselineBackbone=Qwen2-7B2025.09 | 82.75 | — | — | |
| Fed-ALoRABackbone=LLaMA2-7B2025.09 | 82.51 | -4.29 | 4.19 | |
| FedITBackbone=LLaMA2-7B2025.09 | 82.47 | -3.92 | 8.39 | |
| FedDPABackbone=LLaMA2-7B2025.09 | 81.96 | -3.3 | 16.78 | |
| FedSA-LoRABackbone=LLaMA2-7B2025.09 | 81.15 | -2.21 | 4.19 | |
| ST BaselineBackbone=LLaMA2-7B2025.09 | 79.67 | — | — |