Question Answering on 7 QA tasks
69.44AccuracyDense Model
Evaluation Results
| Method | Links | |
|---|---|---|
| Dense ModelPR=0%, Backbone=Llama3.1, Model Size=70B2026.02 | 69.44 | |
| Dense ModelPR=0%, Backbone=Qwen3, Model Size=32B2026.02 | 68.28 | |
| Dense ModelPR=0%, Backbone=Llama2, Model Size=70B2026.02 | 66.96 | |
| PPPR=20%, Online=true, Backbone=Llama2, Model Size=70B2026.02 | 66.89 | |
| DensePruning Ratio=0%, Model=Qwen3-MoE-A3B2026.02 | 66.76 | |
| Tyr´PR=40%, Online=false, Backbone=Llama3.1, Model Size=70B2026.02 | 65.9 | |
| Dense ModelPR=0%, Backbone=Qwen3, Model Size=8B2026.02 | 65.74 | |
| FLAPPruning Ratio=20%, Model=Qwen2-MoE-A14B2026.02 | 65.04 | |
| DensePruning Ratio=0%, Model=Qwen2-MoE-A14B2026.02 | 64.71 | |
| Dense ModelPR=0%, Backbone=Llama3.1, Model Size=8B2026.02 | 64.68 | |
| POPPruning Ratio=20%, Model=Qwen2-MoE-A14B2026.02 | 64.64 | |
| POPPruning Ratio=20%, Model=Qwen3-MoE-A3B2026.02 | 64.28 | |
| POPPR=20%, Online=true, Backbone=Qwen3, Model Size=32B2026.02 | 64.24 | |
| POPPruning Ratio=40%, Model=Qwen2-MoE-A14B2026.02 | 64.24 | |
| Tyr´Pruning Ratio=20%, Model=Qwen3-MoE-A3B2026.02 | 63.8 | |
| PPPR=40%, Online=true, Backbone=Llama2, Model Size=70B2026.02 | 63.23 | |
| Dense ModelPR=0%, Backbone=Llama2, Model Size=13B2026.02 | 63 | |
| DensePruning Ratio=0%, Model=Qwen1.5-MoE-A2.7B2026.02 | 60.21 | |
| Tyr´PR=20%, Online=false, Backbone=Llama3.1, Model Size=8B2026.02 | 59.85 | |
| Dense ModelPR=0%, Backbone=Llama2, Model Size=7B2026.02 | 59.7 | |
| Tyr´Pruning Ratio=20%, Model=Qwen1.5-MoE-A2.7B2026.02 | 59.61 | |
| POPPruning Ratio=20%, Model=Qwen1.5-MoE-A2.7B2026.02 | 58.4 | |
| FLAPPruning Ratio=40%, Model=Qwen2-MoE-A14B2026.02 | 58.12 | |
| FLAPPR=20%, Online=false, Backbone=Llama2, Model Size=13B2026.02 | 57.87 | |
| Wanda-spPR=20%, Online=false, Backbone=Llama2, Model Size=7B2026.02 | 55.02 | |
| POPPruning Ratio=40%, Model=Qwen3-MoE-A3B2026.02 | 54.94 | |
| Wanda-spPruning Ratio=20%, Model=Qwen1.5-MoE-A2.7B2026.02 | 53.97 | |
| POPPruning Ratio=40%, Model=Qwen1.5-MoE-A2.7B2026.02 | 53.95 | |
| Tyr´Pruning Ratio=40%, Model=Qwen3-MoE-A3B2026.02 | 53.74 | |
| FLAPPruning Ratio=20%, Model=Qwen1.5-MoE-A2.7B2026.02 | 53.58 | |
| Wanda-spPruning Ratio=20%, Model=Qwen3-MoE-A3B2026.02 | 52.37 | |
| POPPR=40%, Online=true, Backbone=Qwen3, Model Size=8B2026.02 | 50.98 | |
| FLAPPruning Ratio=40%, Model=Qwen1.5-MoE-A2.7B2026.02 | 49.81 | |
| Tyr´Pruning Ratio=40%, Model=Qwen1.5-MoE-A2.7B2026.02 | 48.74 | |
| FLAPPR=40%, Online=false, Backbone=Llama2, Model Size=13B2026.02 | 47.59 | |
| Wanda-spPR=40%, Online=false, Backbone=Llama2, Model Size=7B2026.02 | 41.39 | |
| Wanda-spPruning Ratio=20%, Model=Qwen2-MoE-A14B2026.02 | 40.61 | |
| Wanda-spPruning Ratio=40%, Model=Qwen3-MoE-A3B2026.02 | 39.01 | |
| Wanda-spPruning Ratio=40%, Model=Qwen1.5-MoE-A2.7B2026.02 | 38.04 | |
| Wanda-spPruning Ratio=40%, Model=Qwen2-MoE-A14B2026.02 | 34.43 | |
| FLAPPruning Ratio=20%, Model=Qwen3-MoE-A3B2026.02 | 32.34 | |
| FLAPPruning Ratio=40%, Model=Qwen3-MoE-A3B2026.02 | 32.34 |