Instruction Following on IFEval (Precision, Intent, and Utility Scores)
84.1Precision (Short)SPLoRA
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SPLoRAModel=Gemma-3-4B-it2026.05 | 84.1 | 88.49 | 85.03 | 89.45 | 86.77 | 2.35 | |
| CSULoRAModel=Gemma-3-4B-it2026.05 | 81.89 | 87.41 | 84.47 | 89.33 | 85.78 | 1.36 | |
| SaLoRAModel=Gemma-3-4B-it2026.05 | 81.52 | 86.93 | 83.18 | 88.37 | 85 | 0.58 | |
| SafeLoRAModel=Gemma-3-4B-it2026.05 | 80.96 | 86.69 | 83.73 | 88.85 | 85.06 | 0.64 | |
| AlignGuardModel=Gemma-3-4B-it2026.05 | 80.96 | 86.57 | 82.26 | 87.77 | 84.39 | -0.03 | |
| LoRAModel=Gemma-3-4B-it2026.05 | 80.41 | 86.57 | 82.44 | 88.25 | 84.42 | 0 | |
| LoRAModel=Llama-3.2-3B-Instruct2026.05 | 79.3 | 85.13 | 80.96 | 86.45 | 82.96 | 0 | |
| SaLoRAModel=Llama-3.2-3B-Instruct2026.05 | 77.63 | 83.81 | 79.3 | 85.61 | 81.59 | -1.37 | |
| AlignGuardModel=Llama-3.2-3B-Instruct2026.05 | 77.26 | 84.05 | 79.67 | 86.09 | 81.77 | -1.19 | |
| SPLoRAModel=Llama-3.2-3B-Instruct2026.05 | 76.71 | 83.09 | 79.67 | 85.37 | 81.21 | -1.75 | |
| SafeLoRAModel=Llama-3.2-3B-Instruct2026.05 | 76.52 | 83.57 | 78.56 | 84.89 | 80.89 | -2.07 | |
| CSULoRAModel=Llama-3.2-3B-Instruct2026.05 | 74.12 | 81.53 | 77.08 | 84.05 | 79.2 | -3.76 | |
| Base modelModel=Gemma-3-4B-it2026.05 | 69.69 | 78.9 | 74.68 | 82.73 | 76.5 | -7.92 | |
| Base modelModel=Llama-3.2-3B-Instruct2026.05 | 66.54 | 75.54 | 72.09 | 80.1 | 73.57 | -9.39 | |
| RESTAModel=Llama-3.2-3B-Instruct2026.05 | 60.26 | 71.22 | 62.11 | 72.3 | 66.47 | -16.49 | |
| RESTAModel=Gemma-3-4B-it2026.05 | 57.3 | 67.99 | 59.33 | 69.66 | 63.57 | -20.85 |