Average Performance on AIME 2024, GPQA Diamond, IFEval, and LiveCodeBench v6
73.86Average ScoreOriginal
Evaluation Results
| Method | Links | |
|---|---|---|
| OriginalModel=Seed-OSS-36B-Instruct, MLP Sparsity=0%2026.06 | 73.86 | |
| SpenseGPT+Model=Seed-OSS-36B-Instruct, MLP Sparsity=25%2026.06 | 73.12 | |
| SpenseGPTModel=Seed-OSS-36B-Instruct, MLP Sparsity=25%2026.06 | 71.69 | |
| OriginalModel=Qwen3-32B, MLP Sparsity=0%2026.06 | 71.68 | |
| SpenseGPT+Model=Qwen3-32B, MLP Sparsity=25%2026.06 | 70.76 | |
| SpenseGPT+Model=Seed-OSS-36B-Instruct, MLP Sparsity=37.5%2026.06 | 70.35 | |
| SpenseGPTModel=Seed-OSS-36B-Instruct, MLP Sparsity=37.5%2026.06 | 70.25 | |
| SpenseGPT+Model=Qwen3-32B, MLP Sparsity=37.5%2026.06 | 67.79 | |
| SpenseGPTModel=Qwen3-32B, MLP Sparsity=25%2026.06 | 67.21 | |
| SpenseGPTModel=Qwen3-32B, MLP Sparsity=37.5%2026.06 | 65.38 | |
| SparseGPTModel=Seed-OSS-36B-Instruct, MLP Sparsity=50%2026.06 | 65.18 | |
| SparseGPTModel=Qwen3-32B, MLP Sparsity=50%2026.06 | 60.27 | |
| HyperPruneModel=Qwen3-32B, MLP Sparsity=50%2026.06 | 24.22 | |
| WandaModel=Qwen3-32B, MLP Sparsity=50%2026.06 | 18.96 | |
| HyperPruneModel=Seed-OSS-36B-Instruct, MLP Sparsity=50%2026.06 | 10.71 | |
| WandaModel=Seed-OSS-36B-Instruct, MLP Sparsity=50%2026.06 | 9.88 |