Common Sense Reasoning on ARC Challenge
89.6AccuracyDASD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DASDBase Model=Qwen3-4B2026.05 | 89.6 | 74.3 | |
| Hint-decodingBase Model=Qwen3-4B2026.05 | 88.9 | 71.6 | |
| Self-DistillationBase Model=Qwen3-4B2026.05 | 88.8 | 69.7 | |
| Mask-PPLBase Model=Qwen3-4B2026.05 | 88.4 | 71.1 | |
| OriginBase Model=Qwen3-4B2026.05 | 88 | 70.8 | |
| DASDBase Model=Llama3.2-3B2026.05 | 78 | 56.7 | |
| Hint-decodingBase Model=Llama3.2-3B2026.05 | 77 | 56.3 | |
| Mask-PPLBase Model=Llama3.2-3B2026.05 | 76.2 | 54.7 | |
| DASDBase Model=Gemma2-2B2026.05 | 75.5 | 47 | |
| Mask-PPLBase Model=Gemma2-2B2026.05 | 74.9 | 45.4 | |
| Hint-decodingBase Model=Gemma2-2B2026.05 | 74.5 | 45.9 | |
| Self-DistillationBase Model=Llama3.2-3B2026.05 | 74.4 | 53.1 | |
| OriginBase Model=Llama3.2-3B2026.05 | 74 | 52.3 | |
| Self-DistillationBase Model=Gemma2-2B2026.05 | 73.3 | 44.8 | |
| OriginBase Model=Gemma2-2B2026.05 | 70 | 43 | |
| VanillaBackbone=Mixtral-8x7B-v0.1, Sparsity=0%2025.11 | 56.7 | — | |
| PuzzleMoEBackbone=Mixtral-8x7B-v0.1, Sparsity=25%2025.11 | 55.3 | — | |
| PuzzleMoEBackbone=Mixtral-8x7B-v0.1, Sparsity=50%2025.11 | 53.8 | — | |
| STUNBackbone=Mixtral-8x7B-v0.1, Sparsity=25%2025.11 | 52.7 | — | |
| VanillaBackbone=Qwen3-MoE-30B-A3B, Sparsity=0%2025.11 | 52.7 | — | |
| NAEEBackbone=Mixtral-8x7B-v0.1, Sparsity=25%2025.11 | 52 | — | |
| PuzzleMoEBackbone=Qwen3-MoE-30B-A3B, Sparsity=25%2025.11 | 51.6 | — | |
| D2Backbone=Mixtral-8x7B-v0.1, Sparsity=20%2025.11 | 51 | — | |
| PuzzleMoEBackbone=Qwen3-MoE-30B-A3B, Sparsity=50%2025.11 | 51 | — | |
| HC-SMoEBackbone=Mixtral-8x7B-v0.1, Sparsity=25%2025.11 | 50.3 | — | |
| Sub-MoEBackbone=Mixtral-8x7B-v0.1, Sparsity=25%2025.11 | 49 | — | |
| WandaBackbone=Mixtral-8x7B-v0.1, Sparsity=2:42025.11 | 48.3 | — | |
| WandaBackbone=Qwen3-MoE-30B-A3B, Sparsity=2:42025.11 | 48.2 | — | |
| NAEEBackbone=Mixtral-8x7B-v0.1, Sparsity=50%2025.11 | 48.1 | — | |
| D2Backbone=Mixtral-8x7B-v0.1, Sparsity=40%2025.11 | 47 | — | |
| Sub-MoEBackbone=Mixtral-8x7B-v0.1, Sparsity=50%2025.11 | 45 | — | |
| VanillaBackbone=Deepseek-MoE-16b, Sparsity=0%2025.11 | 44.6 | — | |
| PuzzleMoEBackbone=Deepseek-MoE-16b, Sparsity=25%2025.11 | 44 | — | |
| Sub-MoEBackbone=Qwen3-MoE-30B-A3B, Sparsity=25%2025.11 | 44 | — | |
| PuzzleMoEBackbone=Deepseek-MoE-16b, Sparsity=50%2025.11 | 43 | — | |
| D2Backbone=Qwen3-MoE-30B-A3B, Sparsity=20%2025.11 | 42.1 | — | |
| HC-SMoEBackbone=Mixtral-8x7B-v0.1, Sparsity=50%2025.11 | 41.1 | — | |
| D2Backbone=Deepseek-MoE-16b, Sparsity=20%2025.11 | 41 | — | |
| VanillaBackbone=Qwen1.5-MoE-A2.7B, Sparsity=0%2025.11 | 41 | — | |
| PuzzleMoEBackbone=Qwen1.5-MoE-A2.7B, Sparsity=25%2025.11 | 40.9 | — | |
| PuzzleMoEBackbone=Qwen1.5-MoE-A2.7B, Sparsity=50%2025.11 | 40.7 | — | |
| HC-SMoEBackbone=Qwen3-MoE-30B-A3B, Sparsity=25%2025.11 | 40.1 | — | |
| Sub-MoEBackbone=Deepseek-MoE-16b, Sparsity=25%2025.11 | 40 | — | |
| Sub-MoEBackbone=Qwen3-MoE-30B-A3B, Sparsity=50%2025.11 | 40 | — | |
| WandaBackbone=Qwen1.5-MoE-A2.7B, Sparsity=2:42025.11 | 38.7 | — | |
| WandaBackbone=Deepseek-MoE-16b, Sparsity=2:42025.11 | 37.4 | — | |
| HC-SMoEBackbone=Deepseek-MoE-16b, Sparsity=25%2025.11 | 36.7 | — | |
| D2Backbone=Deepseek-MoE-16b, Sparsity=40%2025.11 | 36 | — | |
| D2Backbone=Qwen3-MoE-30B-A3B, Sparsity=40%2025.11 | 35.7 | — | |
| HC-SMoEBackbone=Qwen1.5-MoE-A2.7B, Sparsity=25%2025.11 | 34.8 | — | |
| HC-SMoEBackbone=Qwen3-MoE-30B-A3B, Sparsity=50%2025.11 | 34.3 | — | |
| D2Backbone=Qwen1.5-MoE-A2.7B, Sparsity=20%2025.11 | 33.6 | — | |
| Sub-MoEBackbone=Deepseek-MoE-16b, Sparsity=50%2025.11 | 32 | — | |
| D2Backbone=Qwen1.5-MoE-A2.7B, Sparsity=40%2025.11 | 27.8 | — | |
| RegMix-DProxy models=512, Mode=Online2026.06 | 27.57 | — | |
| RegMix-DProxy models=128, Mode=Offline2026.06 | 27.23 | — | |
| RegMix-DProxy models=128, Mode=Online2026.06 | 26.55 | — | |
| RegMix2026.06 | 26.39 | — | |
| RegMix-DProxy models=512, Mode=Offline2026.06 | 26.32 | — | |
| DoReMi2026.06 | 25.78 | — | |
| Human2026.06 | 25.61 | — | |
| HC-SMoEBackbone=Qwen1.5-MoE-A2.7B, Sparsity=50%2025.11 | 23.9 | — | |
| HC-SMoEBackbone=Deepseek-MoE-16b, Sparsity=50%2025.11 | 22.3 | — |