Multi-task Language Understanding on MMLU (Accuracy and Supernodes Pruned)
69AccuracyUnpruned
Evaluation results
| Method | Links | ||
|---|---|---|---|
| UnprunedBackbone=Llama-3.1-8B2026.04 | 69 | 0 | |
| Magnitude (channel)Backbone=Llama-3.1-8B, Sparsity=50%, Sparsity Target=FFN channel2026.04 | 30 | 41.7 | |
| SCAR-ProtBackbone=Llama-3.1-8B, Sparsity=50%, Sparsity Target=FFN channel2026.04 | 30 | 0 | |
| SCAR-LPBackbone=Llama-3.1-8B, Sparsity=50%, Sparsity Target=FFN channel2026.04 | 29 | 0 | |
| SCAR-ConnBackbone=Llama-3.1-8B, Sparsity=50%, Sparsity Target=FFN channel2026.04 | 29 | 0 | |
| Act. L2Backbone=Llama-3.1-8B, Sparsity=50%, Sparsity Target=FFN channel2026.04 | 28 | 19.6 | |
| SparseGPT (channel)Backbone=Llama-3.1-8B, Sparsity=50%, Sparsity Target=FFN channel2026.04 | 22 | 34.6 | |
| Wanda (channel)Backbone=Llama-3.1-8B, Sparsity=50%, Sparsity Target=FFN channel2026.04 | 21 | 24.3 |