General LLM Benchmarking on ProSparse-LLaMA2-7B Evaluation Suite (test)
62.94TriviaQA AccuracyDense baseline
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Dense baselinePredicted sparsity (%)=−2026.03 | 62.94 | 75.42 | 53.58 | 17.68 | 22.59 | 11.45 | 35.85 | 31.84 | 38.92 | 1 | |
| SVDP (seq.)Predictor category=Ours, Predicted sparsity (%)=50, r=256 sequential, Execution pipeline=sequential2026.03 | 62.74 | 75.34 | 53.67 | 17.68 | 22.07 | 11.22 | 35.58 | 31.94 | 38.78 | 1.64 | |
| SVDPPredictor category=Ours, Predicted sparsity (%)=75, r=1024 parallel2026.03 | 62.66 | 75.38 | 53.58 | 17.68 | 22.18 | 10.54 | 35.86 | 31.85 | 38.72 | 1.61 | |
| PowerInferPredictor category=Trained predictors, Predicted sparsity (%)=80, r=1024 parallel2026.03 | 61.1 | 74.54 | 52.99 | 15.85 | 21.15 | 10.69 | 35.29 | 31.8 | 37.93 | 1.63 | |
| GRIFFINPredictor category=Training-free methods, Predicted sparsity (%)=20, r=1024 parallel2026.03 | 59.77 | 74.75 | 53.33 | 15.85 | 22.18 | 8.95 | 35.49 | 32.01 | 37.79 | 1.14 | |
| Deja VuPredictor category=Trained predictors, Predicted sparsity (%)=20, r=1024 parallel2026.03 | 59.47 | 74.87 | 53.41 | 3.66 | 2.98 | 3.03 | 34.49 | 30.97 | 32.86 | 1.38 | |
| GRIFFINPredictor category=Training-free methods, Predicted sparsity (%)=50, r=1024 parallel2026.03 | 35.74 | 72.77 | 51.96 | 12.2 | 18.17 | 4.47 | 35.32 | 31.88 | 32.81 | 1.46 |