Zero-shot Evaluation on ARC (c/e), WinoGrande, BoolQ, HellaSwag, OBQA, PIQA, and MMLU (test val)
0.7361Average AccuracyDense
Evaluation Results
| Method | Links | |
|---|---|---|
| DenseModel=LLaMA2-70B, Sparsity=0%2025.12 | 0.7361 | |
| O-FANGModel=LLaMA2-70B, Sparsity=20%2025.12 | 0.7224 | |
| F-FANGModel=LLaMA2-70B, Sparsity=20%2025.12 | 0.7211 | |
| FLAPModel=LLaMA2-70B, Sparsity=20%2025.12 | 0.716 | |
| O-FANGModel=LLaMA2-70B, Sparsity=30%2025.12 | 0.7134 | |
| SoBPModel=LLaMA2-70B, Sparsity=20%2025.12 | 0.7124 | |
| F-FANGModel=LLaMA2-70B, Sparsity=30%2025.12 | 0.7113 | |
| OBCModel=LLaMA2-70B, Sparsity=20%2025.12 | 0.7054 | |
| SoBPModel=LLaMA2-70B, Sparsity=30%2025.12 | 0.703 | |
| FLAPModel=LLaMA2-70B, Sparsity=30%2025.12 | 0.6968 | |
| SliceGPTModel=LLaMA2-70B, Sparsity=20%2025.12 | 0.696 | |
| F-FANGModel=LLaMA2-70B, Sparsity=40%2025.12 | 0.693 | |
| DenseModel=LLaMA2-13B, Sparsity=0%2025.12 | 0.6924 | |
| O-FANGModel=LLaMA2-70B, Sparsity=40%2025.12 | 0.6877 | |
| SoBPModel=LLaMA2-70B, Sparsity=40%2025.12 | 0.6858 | |
| SVD-LLMModel=LLaMA2-70B, Sparsity=20%2025.12 | 0.6814 | |
| FLAPModel=LLaMA2-70B, Sparsity=40%2025.12 | 0.6796 | |
| OBCModel=LLaMA2-70B, Sparsity=30%2025.12 | 0.6794 | |
| F-FANGModel=LLaMA2-13B, Sparsity=20%2025.12 | 0.6791 | |
| SoBPModel=LLaMA2-13B, Sparsity=20%2025.12 | 0.6773 | |
| OBCModel=LLaMA2-70B, Sparsity=40%2025.12 | 0.6723 | |
| O-FANGModel=LLaMA2-13B, Sparsity=20%2025.12 | 0.6706 | |
| SoBPModel=LLaMA2-13B, Sparsity=30%2025.12 | 0.6682 | |
| DenseModel=LLaMA2-7B, Sparsity=0%2025.12 | 0.6669 | |
| OBCModel=LLaMA2-13B, Sparsity=20%2025.12 | 0.6659 | |
| DenseModel=LLaMA1-7B, Sparsity=0%2025.12 | 0.6605 | |
| F-FANGModel=LLaMA2-13B, Sparsity=30%2025.12 | 0.655 | |
| SVD-LLMModel=LLaMA2-70B, Sparsity=30%2025.12 | 0.6465 | |
| O-FANGModel=LLaMA2-13B, Sparsity=30%2025.12 | 0.6421 | |
| F-FANGModel=LLaMA2-7B, Sparsity=20%2025.12 | 0.6329 | |
| SoBPModel=LLaMA2-7B, Sparsity=20%2025.12 | 0.6327 | |
| F-FANGModel=LLaMA1-7B, Sparsity=20%2025.12 | 0.6321 | |
| O-FANGModel=LLaMA2-7B, Sparsity=20%2025.12 | 0.6288 | |
| O-FANGModel=LLaMA1-7B, Sparsity=20%2025.12 | 0.6265 | |
| OBCModel=LLaMA2-7B, Sparsity=20%2025.12 | 0.6227 | |
| SoBPModel=LLaMA1-7B, Sparsity=20%2025.12 | 0.6219 | |
| SliceGPTModel=LLaMA2-70B, Sparsity=30%2025.12 | 0.6161 | |
| FLAPModel=LLaMA2-13B, Sparsity=20%2025.12 | 0.6155 | |
| OBCModel=LLaMA1-7B, Sparsity=20%2025.12 | 0.6153 | |
| FLAPModel=LLaMA1-7B, Sparsity=20%2025.12 | 0.6119 | |
| O-FANGModel=LLaMA2-13B, Sparsity=40%2025.12 | 0.6116 | |
| SoBPModel=LLaMA2-13B, Sparsity=40%2025.12 | 0.6086 | |
| F-FANGModel=LLaMA1-7B, Sparsity=30%2025.12 | 0.6034 | |
| O-FANGModel=LLaMA2-7B, Sparsity=30%2025.12 | 0.5983 | |
| SoBPModel=LLaMA1-7B, Sparsity=30%2025.12 | 0.5961 | |
| O-FANGModel=LLaMA1-7B, Sparsity=30%2025.12 | 0.595 | |
| F-FANGModel=LLaMA2-7B, Sparsity=30%2025.12 | 0.5944 | |
| SoBPModel=LLaMA2-7B, Sparsity=30%2025.12 | 0.5915 | |
| SVD-LLMModel=LLaMA2-13B, Sparsity=20%2025.12 | 0.5898 | |
| F-FANGModel=LLaMA2-13B, Sparsity=40%2025.12 | 0.5862 | |
| SVD-LLMModel=LLaMA2-70B, Sparsity=40%2025.12 | 0.5805 | |
| OBCModel=LLaMA2-7B, Sparsity=30%2025.12 | 0.5788 | |
| OBCModel=LLaMA1-7B, Sparsity=30%2025.12 | 0.5756 | |
| FLAPModel=LLaMA1-7B, Sparsity=30%2025.12 | 0.573 | |
| FLAPModel=LLaMA2-13B, Sparsity=30%2025.12 | 0.5727 | |
| SliceGPTModel=LLaMA2-13B, Sparsity=20%2025.12 | 0.5678 | |
| FLAPModel=LLaMA2-7B, Sparsity=20%2025.12 | 0.5662 | |
| SVD-LLMModel=LLaMA1-7B, Sparsity=20%2025.12 | 0.5625 | |
| SliceGPTModel=LLaMA1-7B, Sparsity=20%2025.12 | 0.5616 | |
| SoBPModel=LLaMA1-7B, Sparsity=40%2025.12 | 0.561 | |
| SoBPModel=LLaMA2-7B, Sparsity=40%2025.12 | 0.5606 | |
| O-FANGModel=LLaMA1-7B, Sparsity=40%2025.12 | 0.5516 | |
| O-FANGModel=LLaMA2-7B, Sparsity=40%2025.12 | 0.5496 | |
| SliceGPTModel=LLaMA2-7B, Sparsity=20%2025.12 | 0.5425 | |
| F-FANGModel=LLaMA1-7B, Sparsity=40%2025.12 | 0.5382 | |
| FLAPModel=LLaMA2-13B, Sparsity=40%2025.12 | 0.5318 | |
| SVD-LLMModel=LLaMA2-13B, Sparsity=30%2025.12 | 0.5316 | |
| OBCModel=LLaMA1-7B, Sparsity=40%2025.12 | 0.5249 | |
| OBCModel=LLaMA2-7B, Sparsity=40%2025.12 | 0.5247 | |
| FLAPModel=LLaMA1-7B, Sparsity=40%2025.12 | 0.5245 | |
| SliceGPTModel=LLaMA2-70B, Sparsity=40%2025.12 | 0.52 | |
| OBCModel=LLaMA2-13B, Sparsity=30%2025.12 | 0.5156 | |
| SVD-LLMModel=LLaMA1-7B, Sparsity=30%2025.12 | 0.5139 | |
| F-FANGModel=LLaMA2-7B, Sparsity=40%2025.12 | 0.5094 | |
| FLAPModel=LLaMA2-7B, Sparsity=30%2025.12 | 0.5091 | |
| SliceGPTModel=LLaMA2-13B, Sparsity=30%2025.12 | 0.501 | |
| OBCModel=LLaMA2-13B, Sparsity=40%2025.12 | 0.488 | |
| SVD-LLMModel=LLaMA2-7B, Sparsity=20%2025.12 | 0.487 | |
| FLAPModel=LLaMA2-7B, Sparsity=40%2025.12 | 0.487 | |
| SliceGPTModel=LLaMA1-7B, Sparsity=30%2025.12 | 0.469 | |
| SliceGPTModel=LLaMA2-7B, Sparsity=30%2025.12 | 0.467 | |
| SVD-LLMModel=LLaMA2-13B, Sparsity=40%2025.12 | 0.454 | |
| SVD-LLMModel=LLaMA2-7B, Sparsity=30%2025.12 | 0.4477 | |
| SliceGPTModel=LLaMA2-13B, Sparsity=40%2025.12 | 0.4431 | |
| SVD-LLMModel=LLaMA1-7B, Sparsity=40%2025.12 | 0.4295 | |
| SliceGPTModel=LLaMA2-7B, Sparsity=40%2025.12 | 0.4147 | |
| SVD-LLMModel=LLaMA2-7B, Sparsity=40%2025.12 | 0.4047 | |
| SliceGPTModel=LLaMA1-7B, Sparsity=40%2025.12 | 0.3964 |