Question Answering on OBQA, ARC-E, ARC-C, and CQA
53.6OBQA AccuracyCentSFT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CentSFTModel=Mistral-7B, Ratio=-2026.04 | 53.6 | 83.8 | 57.34 | 82.88 | 69.41 | |
| CentSFTModel=LLaMA2-7B, Ratio=-2026.04 | 52.2 | 81.02 | 51.88 | 78.3 | 65.85 | |
| FedProxyModel=Mistral-7B, Ratio=302026.04 | 49.4 | 83.96 | 56.57 | 79.85 | 67.45 | |
| FedProxyModel=Mistral-7B, Ratio=502026.04 | 46.2 | 85.73 | 59.98 | 75.76 | 66.92 | |
| FedProxyModel=LLaMA2-7B, Ratio=302026.04 | 44.6 | 79.63 | 50.68 | 75.51 | 62.61 | |
| OTModel=Mistral-7B, Ratio=302026.04 | 44 | 80.09 | 51.62 | 74.45 | 62.54 | |
| FedOTModel=Mistral-7B, Ratio=302026.04 | 42.8 | 82.66 | 56.66 | 73.63 | 63.94 | |
| OTModel=Mistral-7B, Ratio=502026.04 | 42.2 | 79.55 | 54.52 | 68.8 | 61.27 | |
| FedOTModel=Mistral-7B, Ratio=502026.04 | 40 | 81.82 | 56.83 | 68.88 | 61.88 | |
| FedBiOTModel=Mistral-7B, Ratio=302026.04 | 39.2 | 82.58 | 54.95 | 69.7 | 61.61 | |
| FedProxyModel=LLaMA2-7B, Ratio=502026.04 | 39 | 78.66 | 47.1 | 72.65 | 59.35 | |
| FedBiOTModel=Mistral-7B, Ratio=502026.04 | 37 | 82.28 | 54.69 | 69.45 | 60.86 | |
| ZeroShotModel=Mistral-7B, Ratio=-2026.04 | 36.8 | 81.14 | 54.68 | 68.06 | 60.17 | |
| OTModel=LLaMA2-7B, Ratio=302026.04 | 36.2 | 77.61 | 46.67 | 32.19 | 48.17 | |
| FedOTModel=LLaMA2-7B, Ratio=302026.04 | 36.2 | 77.82 | 46.42 | 32.68 | 48.28 | |
| FedOTModel=LLaMA2-7B, Ratio=502026.04 | 35.4 | 77.15 | 44.62 | 33.5 | 47.67 | |
| OTModel=LLaMA2-7B, Ratio=502026.04 | 35 | 77.53 | 44.37 | 31.12 | 47.01 | |
| FedBiOTModel=LLaMA2-7B, Ratio=302026.04 | 33.6 | 76.26 | 44.54 | 31.29 | 46.42 | |
| FedBiOTModel=LLaMA2-7B, Ratio=502026.04 | 33.4 | 76.52 | 44.88 | 30.88 | 46.42 | |
| ZeroShotModel=LLaMA2-7B, Ratio=-2026.04 | 32.2 | 76.56 | 44.45 | 30.79 | 46 |