Zero-shot Commonsense Reasoning on Commonsense Reasoning Benchmarks (LLaMA-2-13B)
80.92BoolQ Accuracy (Zero-shot)Dense
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DensePruning Ratio=0%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 80.92 | 80.52 | 79.36 | 71.98 | 79.63 | 49.15 | 45 | 69.51 | |
| SlimGPT w/oPruning Ratio=20%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 80.37 | 78.45 | 77.07 | 71.51 | 75.84 | 45.14 | 43.6 | 67.43 | |
| Token FilteringPruning Ratio=20%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 80.18 | 78.62 | 77.78 | 72.22 | 78.96 | 48.04 | 45 | 68.69 | |
| Token FilteringPruning Ratio=25%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 80.09 | 78.56 | 77.82 | 71.03 | 77.57 | 47.95 | 44.2 | 68.17 | |
| Token FilteringPruning Ratio=33%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 79.79 | 77.15 | 76.07 | 71.35 | 77.44 | 47.35 | 44 | 67.65 | |
| Token FilteringPruning Ratio=50%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 79.76 | 77.04 | 74.56 | 71.03 | 71.72 | 45.22 | 42 | 65.9 | |
| SlimGPT w/oPruning Ratio=25%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 78.78 | 77.91 | 75.65 | 70.64 | 73.06 | 44.11 | 43 | 66.16 | |
| FLAPPruning Ratio=20%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 77.21 | 77.95 | 78.06 | 71.32 | 76.44 | 45.27 | 42 | 66.89 | |
| PPPruning Ratio=20%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 76.29 | 79.55 | 76.53 | 68.57 | 76.85 | 44.57 | 42 | 66.33 | |
| SlimGPT w/oPruning Ratio=33%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 76.27 | 76.44 | 72.76 | 70.8 | 70.83 | 41.21 | 41 | 64.19 | |
| FLAPPruning Ratio=25%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 75.81 | 75.61 | 74.74 | 69.59 | 73.83 | 43.57 | 41 | 64.88 | |
| FLAPPruning Ratio=33%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 74.37 | 74.42 | 70.45 | 68.17 | 70.22 | 42.29 | 39.2 | 62.73 | |
| FLAPPruning Ratio=50%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 74.31 | 70.49 | 58.39 | 62.96 | 61.67 | 36.83 | 37.2 | 57.41 | |
| PPPruning Ratio=25%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 73.82 | 78.47 | 75.82 | 67.78 | 75.42 | 42.95 | 42.4 | 65.24 | |
| PPPruning Ratio=33%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 70.3 | 76.74 | 71.67 | 63.65 | 71.08 | 39.96 | 42.6 | 62.29 | |
| SlimGPT w/oPruning Ratio=50%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 66.06 | 73.61 | 61.76 | 65.82 | 60.44 | 34.39 | 38 | 57.15 | |
| PPPruning Ratio=50%, Base Model=LLaMA-2-13B, Zero-shot=true2025.12 | 62.17 | 69.22 | 49.88 | 55.07 | 59.26 | 29.63 | 36.2 | 51.93 |