Mathematical Reasoning on GSM8K (test) (Accuracy, Runtime)
62.7GSM8K AccuracyParaBlock
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ParaBlockBackbone=Qwen-2.5-1.5B-Instruct2025.11 | 62.7 | 10.56 | |
| FedITBackbone=Qwen-2.5-1.5B-Instruct2025.11 | 62.32 | 19.76 | |
| FedBCDBackbone=Qwen-2.5-1.5B-Instruct2025.11 | 61.79 | 18.48 | |
| FFA-LoRABackbone=Qwen-2.5-1.5B-Instruct2025.11 | 61.56 | 17.12 | |
| Fed full FTBackbone=Qwen-2.5-1.5B-Instruct2025.11 | 60.58 | 9.3 | |
| FLoRABackbone=Qwen-2.5-1.5B-Instruct2025.11 | 57.77 | 52.42 | |
| FedCyBGDBackbone=Qwen-2.5-1.5B-Instruct2025.11 | 54.51 | 42.01 | |
| BaseBackbone=Qwen-2.5-1.5B-Instruct2025.11 | 54.28 | — |