Multi-task NLP on FLAN Heterogeneous Federated Setting
84.13Avg. ROUGE-1Fed-ALoRA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Fed-ALoRABackbone=Qwen2-7B2025.09 | 84.13 | -1.02 | 9.23 | |
| ST BaselineBackbone=Qwen2-7B2025.09 | 83.43 | — | — | |
| ZeroPaddingBackbone=Qwen2-7B2025.09 | 83.32 | 0.06 | 37.73 | |
| Fed-ALoRABackbone=LLaMA2-7B2025.09 | 82.5 | -1.07 | 12.12 | |
| ZeroPaddingBackbone=LLaMA2-7B2025.09 | 82.29 | -0.91 | 49.28 | |
| ST BaselineBackbone=LLaMA2-7B2025.09 | 81.73 | — | — | |
| FLoRABackbone=Qwen2-7B2025.09 | 81.65 | 2.13 | 108.38 | |
| FLoRABackbone=LLaMA2-7B2025.09 | 80.45 | 1.54 | 141.56 |