Multi-discipline Understanding on MMLU
75.05AccuracyPOP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| POPBackbone=Qwen3-VL-8B-Instruct, Pruning Ratio=33.3%2026.02 | 75.05 | — | |
| Full ModelBackbone=Qwen3-VL-8B-Instruct, Pruning Ratio=0%2026.02 | 74.95 | — | |
| WandaBackbone=Qwen3-VL-8B-Instruct, Pruning Ratio=30%2026.02 | 73.78 | — | |
| Full ModelBackbone=Gemma-3-12B-It, Pruning Ratio=0%2026.02 | 71.46 | — | |
| POPBackbone=Gemma-3-12B-It, Pruning Ratio=33.3%2026.02 | 71.37 | — | |
| WandaBackbone=Gemma-3-12B-It, Pruning Ratio=30%2026.02 | 69.7 | — | |
| Full ModelBackbone=Llama-3.1-8B-Instruct, Pruning Ratio=0%2026.02 | 68.33 | — | |
| POPBackbone=Llama-3.1-8B-Instruct, Pruning Ratio=31.25%2026.02 | 67.43 | — | |
| WandaBackbone=Llama-3.1-8B-Instruct, Pruning Ratio=30%2026.02 | 65.87 | — | |
| ShortGPTBackbone=Llama-3.1-8B-Instruct, Pruning Ratio=25%2026.02 | 65.8 | — | |
| JETBackbone=DS-Qwen-7B2025.09 | 63.9 | 860 | |
| Laser-DEBackbone=DS-Qwen-7B2025.09 | 63.4 | 1,027 | |
| Laser-DBackbone=DS-Qwen-7B2025.09 | 62.3 | 1,059 | |
| SFTBackbone=DS-Qwen-7B2025.09 | 60.2 | 1,044 | |
| BaseBackbone=DS-Qwen-7B2025.09 | 60.1 | 1,022 | |
| DPOBackbone=DS-Qwen-7B2025.09 | 60.1 | 947 | |
| DAPOBackbone=DS-Qwen-7B2025.09 | 58.9 | 907 | |
| AdaptThinkBackbone=DS-Qwen-7B2025.09 | 57.3 | 840 | |
| LCR1Backbone=DS-Qwen-7B2025.09 | 56.5 | 611 | |
| Laser-DBackbone=DS-Qwen-1.5B2025.09 | 47.2 | 1,143 | |
| DAPOBackbone=DS-Qwen-1.5B2025.09 | 46.5 | 856 | |
| Laser-DEBackbone=DS-Qwen-1.5B2025.09 | 46.2 | 1,217 | |
| SFTBackbone=DS-Qwen-1.5B2025.09 | 44.7 | 1,122 | |
| JETBackbone=DS-Qwen-1.5B2025.09 | 44.6 | 715 | |
| DPOBackbone=DS-Qwen-1.5B2025.09 | 44.5 | 1,143 | |
| AdaptThinkBackbone=DS-Qwen-1.5B2025.09 | 44.5 | 1,224 | |
| BaseBackbone=DS-Qwen-1.5B2025.09 | 43.8 | 1,306 | |
| LCR1Backbone=DS-Qwen-1.5B2025.09 | 42.7 | 744 | |
| SliceGPTBackbone=Qwen3-VL-8B-Instruct, Pruning Ratio=25%2026.02 | 39.16 | — | |
| SliceGPTBackbone=Llama-3.1-8B-Instruct, Pruning Ratio=25%2026.02 | 34.97 | — | |
| ShortGPTBackbone=Qwen3-VL-8B-Instruct, Pruning Ratio=25%2026.02 | 33.85 | — | |
| ShortGPTBackbone=Gemma-3-12B-It, Pruning Ratio=25%2026.02 | 23.81 | — | |
| SliceGPTBackbone=Gemma-3-12B-It, Pruning Ratio=25%2026.02 | 22.95 | — |