Science Question Answering on DeepSeek-distilled ARC-C (test)
77.22AccuracyFull Synthetic
Evaluation Results
| Method | Links | |
|---|---|---|
| Full SyntheticBackbone Model=Mistral-v0.2-7B, Selection Ratio=100%2026.05 | 77.22 | |
| TMAPBackbone Model=Mistral-v0.2-7B, Selection Ratio=30%2026.05 | 77.05 | |
| AlpagasusBackbone Model=Mistral-v0.2-7B, Selection Ratio=30%2026.05 | 76.45 | |
| DEITABackbone Model=Mistral-v0.2-7B, Selection Ratio=30%2026.05 | 75.1 | |
| LongestBackbone Model=Mistral-v0.2-7B, Selection Ratio=30%2026.05 | 74.4 | |
| IFDBackbone Model=Mistral-v0.2-7B, Selection Ratio=30%2026.05 | 74.15 | |
| RandomBackbone Model=Mistral-v0.2-7B, Selection Ratio=30%2026.05 | 69.37 | |
| Full SyntheticBackbone Model=Llama2-13B, Selection Ratio=100%2026.05 | 67.23 | |
| TMAPBackbone Model=Llama2-13B, Selection Ratio=30%2026.05 | 66.13 | |
| DEITABackbone Model=Llama2-13B, Selection Ratio=30%2026.05 | 65.52 | |
| IFDBackbone Model=Llama2-13B, Selection Ratio=30%2026.05 | 64.76 | |
| AlpagasusBackbone Model=Llama2-13B, Selection Ratio=30%2026.05 | 64.51 | |
| RandomBackbone Model=Llama2-13B, Selection Ratio=30%2026.05 | 64.33 | |
| LongestBackbone Model=Llama2-13B, Selection Ratio=30%2026.05 | 64.25 | |
| BaseBackbone Model=Llama2-13B2026.05 | 60.58 | |
| BaseBackbone Model=Mistral-v0.2-7B2026.05 | 56.65 |