Question Answering on Commonsense Reasoning Suite (test)
48.27ARC-cEVA
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| EVABackbone=LLaMA-3-8B, Rank (r)=16, Seeds=3, Evaluation Protocol=Zero-shot2026.05 | 48.27 | 71.34 | 84.15 | 80.81 | 45.53 | 79.04 | 71.85 | 68.71 | |
| FIM-LoRABackbone=LLaMA-3-8B, Rank (r)=16, Seeds=3, Evaluation Protocol=Zero-shot, rmin=82026.05 | 47.92 | 72.59 | 83.17 | 79.62 | 45.47 | 79.47 | 71.06 | 68.47 | |
| LoRABackbone=LLaMA-3-8B, Rank (r)=16, Seeds=3, Evaluation Protocol=Zero-shot2026.05 | 47.67 | 72.78 | 83.52 | 80.8 | 45.47 | 79.13 | 71.85 | 68.74 | |
| FIM-LoRABackbone=LLaMA-3-8B, Rank (r)=16, Seeds=3, Evaluation Protocol=Zero-shot, rmin=42026.05 | 46.73 | 70.93 | 83.02 | 79.27 | 45.47 | 78.8 | 69.98 | 67.74 | |
| FIM-LoRABackbone=LLaMA-3-8B, Rank (r)=16, Seeds=3, Evaluation Protocol=Zero-shot, rmin=12026.05 | 45.28 | 67.13 | 83.04 | 79.27 | 45.13 | 78.94 | 70.24 | 67.01 |