Science Question Answering on GPQA (accuracy, Avg, Delta)
79.29AccuracyAct-Logits
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Act-LogitsBase Model=openPangu-Ultra, Sparsity=31.25%2025.05 | 79.29 | 76.91 | -2.34 | |
| FullBase Model=openPangu-Ultra, Sparsity=31.25%2025.05 | 76.77 | 79.25 | — | |
| EASY-EPBase Model=openPangu-Ultra, Sparsity=31.25%2025.05 | 76.74 | 76.45 | -2.8 | |
| PreMoEBase Model=openPangu-Ultra, Sparsity=31.25%2025.05 | 75.76 | 78.45 | -0.8 | |
| FullBase Model=DeepSeek-R1, Sparsity=50%2025.05 | 73.23 | 75.51 | — | |
| PreMoEBase Model=DeepSeek-R1, Sparsity=50%2025.05 | 72.22 | 76.52 | 1.01 | |
| EASY-EPBase Model=DeepSeek-R1, Sparsity=50%2025.05 | 70.12 | 74.69 | -0.82 | |
| FullBase Model=Qwen3-30B-A3B, Sparsity=50%2025.05 | 68.69 | 80.69 | — | |
| PreMoEBase Model=Qwen3-30B-A3B, Sparsity=50%2025.05 | 68.18 | 79.92 | -0.77 | |
| SEER (L)Base Model=openPangu-Ultra, Sparsity=31.25%2025.05 | 62.63 | 71.29 | -7.96 | |
| RandomBase Model=openPangu-Ultra, Sparsity=31.25%2025.05 | 60.1 | 50.13 | -29.12 | |
| SEER (G)Base Model=openPangu-Ultra, Sparsity=31.25%2025.05 | 57.58 | 70.93 | -8.32 | |
| FrequencyBase Model=openPangu-Ultra, Sparsity=31.25%2025.05 | 52.02 | 69.67 | -9.58 | |
| EASY-EPBase Model=Qwen3-30B-A3B, Sparsity=50%2025.05 | 50.48 | 54.26 | -26.43 | |
| Act-LogitsBase Model=DeepSeek-R1, Sparsity=50%2025.05 | 48.48 | 62.87 | -12.64 | |
| FrequencyBase Model=Qwen3-30B-A3B, Sparsity=50%2025.05 | 47.98 | 54.26 | -32.76 | |
| RandomBase Model=DeepSeek-R1, Sparsity=50%2025.05 | 39.9 | 26.69 | -48.82 | |
| SEER (L)Base Model=DeepSeek-R1, Sparsity=50%2025.05 | 36.87 | 23.56 | -51.95 | |
| SEER (G)Base Model=DeepSeek-R1, Sparsity=50%2025.05 | 36.87 | 24.31 | -51.2 | |
| FrequencyBase Model=DeepSeek-R1, Sparsity=50%2025.05 | 33.33 | 49.93 | -25.58 | |
| All-LogitsBase Model=DeepSeek-R1, Sparsity=50%2025.05 | 28.79 | 5.46 | -70.05 | |
| RandomBase Model=Qwen3-30B-A3B, Sparsity=50%2025.05 | 17.68 | 13.07 | -67.62 | |
| All-LogitsBase Model=openPangu-Ultra, Sparsity=31.25%2025.05 | 16.67 | 5.39 | -73.86 | |
| Act-LogitsBase Model=Qwen3-30B-A3B, Sparsity=50%2025.05 | 3.54 | 0.82 | -79.87 | |
| All-LogitsBase Model=Qwen3-30B-A3B, Sparsity=50%2025.05 | 0.51 | 0.35 | -80.34 | |
| SEER (L)Base Model=Qwen3-30B-A3B, Sparsity=50%2025.05 | 0 | 0.64 | -80.05 | |
| SEER (G)Base Model=Qwen3-30B-A3B, Sparsity=50%2025.05 | 0 | 0.34 | -80.35 |