LLM Utility Evaluation on LIMA original and rewritten variants
77.6ARC-E AccuracyCBBR
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| CBBRModel=Llama3.1-8B, Rewriter=CBBR2026.05 | 77.6 | 47.9 | 80.8 | 78.7 | 73.6 | 63.2 | 71.9 | -5.9 | |
| CBBRModel=Qwen2.5-7B, Rewriter=CBBR2026.05 | 76.5 | 48.1 | 79.8 | 79.1 | 70.3 | 71.1 | 67.5 | -8.1 | |
| Original (No Rewriter)Model=Qwen2.5-1.5B, Rewriter=None2026.05 | 75.3 | 44.3 | 67.3 | 75.1 | 63.5 | 58.5 | 38.1 | — | |
| OBBRModel=Qwen2.5-1.5B, Rewriter=OBBR2026.05 | 75 | 44 | 67.2 | 75 | 63.5 | 58.5 | 38.6 | 0 | |
| CBBRModel=Qwen2.5-1.5B, Rewriter=CBBR2026.05 | 74.7 | 44.2 | 67.8 | 75 | 64.1 | 57.9 | 36.1 | 0.8 | |
| DPRModel=Qwen2.5-1.5B, Rewriter=DPR2026.05 | 74.3 | 43.6 | 67.2 | 75 | 64.2 | 58.1 | 40.3 | -0.5 | |
| ParaphraseModel=Qwen2.5-7B, Rewriter=Para.2026.05 | 74.3 | 47 | 79.8 | 78 | 66.1 | 67.2 | 60.5 | -3.6 | |
| ParaphraseModel=Qwen2.5-1.5B, Rewriter=Para.2026.05 | 73.8 | 43.8 | 67 | 74.7 | 63.4 | 58.3 | 39.3 | 0.2 | |
| DPRModel=Llama3.1-8B, Rewriter=DPR2026.05 | 73.8 | 47.2 | 81.2 | 78.6 | 72 | 61.2 | 68.9 | -3.5 | |
| ParaphraseModel=Llama3.1-8B, Rewriter=Para.2026.05 | 72.4 | 45.3 | 80.9 | 77.6 | 69.3 | 60.6 | 69.3 | -1.8 | |
| OBBRModel=Qwen2.5-7B, Rewriter=OBBR2026.05 | 70.4 | 44.1 | 79.8 | 76.3 | 67.2 | 68 | 62.4 | -2.4 | |
| Original (No Rewriter)Model=Llama3.1-8B, Rewriter=None2026.05 | 69.9 | 44.5 | 80.4 | 76.3 | 71 | 61.2 | 64 | — | |
| DPRModel=Qwen2.5-7B, Rewriter=DPR2026.05 | 69.6 | 42.8 | 80.2 | 77.4 | 65.3 | 68 | 60.1 | -1.1 | |
| OBBRModel=Llama3.1-8B, Rewriter=OBBR2026.05 | 68.9 | 43.6 | 80.6 | 77.5 | 71.4 | 61 | 66.6 | -0.4 | |
| OBBRModel=Llama3.2-1B, Rewriter=OBBR2026.05 | 67.9 | 34.9 | 60.8 | 74.5 | 62.9 | 45.8 | 51.3 | -1.5 | |
| DPRModel=Llama3.2-1B, Rewriter=DPR2026.05 | 67.9 | 34.8 | 61 | 74 | 62.4 | 45.3 | 49.6 | -0.7 | |
| CBBRModel=Llama3.2-1B, Rewriter=CBBR2026.05 | 67.5 | 35.1 | 61.4 | 74.6 | 62.2 | 44.9 | 50.5 | -1 | |
| Original (No Rewriter)Model=Llama3.2-1B, Rewriter=None2026.05 | 67.3 | 34.7 | 62 | 73.6 | 62.4 | 44 | 48.9 | — | |
| Original (No Rewriter)Model=Qwen2.5-7B, Rewriter=None2026.05 | 67.2 | 43.8 | 79.8 | 76.6 | 63.3 | 67.7 | 59.4 | — | |
| ParaphraseModel=Llama3.2-1B, Rewriter=Para.2026.05 | 67.1 | 35.8 | 62.1 | 73.9 | 62.4 | 43.3 | 51.4 | -1 |