Language Understanding on MMLU (Accuracy and Drop)
59.01AccuracyPersona-Pruner
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Persona-PrunerBackbone=Qwen2.5-32B-Instruct, Sparsity=50%2026.06 | 59.01 | — | |
| LLM-PrunerBackbone=Qwen2.5-32B-Instruct, Sparsity=50%2026.06 | 53.45 | — | |
| BF16shot-count=5-shot, Learning Rate=10^-52026.05 | 43.36 | — | |
| HiF8 QATshot-count=5-shot, Learning Rate=10^-5, amax_algo=max, history=64, BF16 warmup=500 steps2026.05 | 43.17 | 0.43 | |
| Adapt-PrunerBackbone=Qwen2.5-32B-Instruct, Sparsity=50%2026.06 | 37.86 | — | |
| Depth PruningBackbone=Qwen2.5-32B-Instruct, Sparsity=50%2026.06 | 24.01 | — | |
| SliceGPTBackbone=Qwen2.5-32B-Instruct, Sparsity=50%2026.06 | 23.02 | — |