Pruning Performance Evaluation on Aggregate Benchmarks Table 6
100Relative Performance (%)Full Model
Evaluation results
| Method | Links | |
|---|---|---|
| Full ModelBackbone=LLaVA-OneVision, Sparsity Ratio=0%2026.04 | 100 | |
| TopoVLMBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.04 | 94.8 | |
| TAMPBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.04 | 92.3 | |
| ECoFLaPBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.04 | 92 | |
| SparseGPTBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.04 | 90.9 | |
| LLM-StreamlineBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.04 | 90.3 | |
| WandaBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.04 | 87 | |
| LLM-PrunerBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.04 | 86.7 | |
| OWLBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.04 | 82.8 | |
| MagnitudeBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.04 | 7.4 |