Role-Playing on Alpaca-P
86.66LLM-as-a-Judge ScoreLlama-3.1-8B-Instruct
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama-3.1-8B-InstructRatio=0%, Base Model=Llama-3.1-8B-Instruct2026.06 | 86.66 | |
| Llama-3.2-3B-InstructRatio=0%, Base Model=Llama-3.2-3B-Instruct2026.06 | 84.86 | |
| Persona-PrunerRatio=25%, Recovery finetuning=w/, Base Model=Llama-3.1-8B-Instruct2026.06 | 83.9 | |
| Persona-PrunerBackbone=Qwen2.5-32B-Instruct, Sparsity=50%2026.06 | 83.1 | |
| Persona-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=25%, Recovery finetuning=true2026.06 | 82.61 | |
| Persona-PrunerRatio=25%, Recovery finetuning=w/, Base Model=Llama-3.2-3B-Instruct2026.06 | 82.52 | |
| Persona-PrunerRatio=25%, Recovery finetuning=w/o, Base Model=Llama-3.1-8B-Instruct2026.06 | 81.86 | |
| Persona-PrunerRatio=50%, Recovery finetuning=w/, Base Model=Llama-3.1-8B-Instruct2026.06 | 81.73 | |
| Persona-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=25%, Recovery finetuning=true2026.06 | 81.34 | |
| Persona-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=50%, Recovery finetuning=true2026.06 | 81.1 | |
| Persona-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=25%, Recovery finetuning=false2026.06 | 80.45 | |
| Persona-PrunerRatio=25%, Recovery finetuning=w/o, Base Model=Llama-3.2-3B-Instruct2026.06 | 80.19 | |
| Persona-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=50%, Recovery finetuning=true2026.06 | 78.54 | |
| Qwen2.5-7B-InstructBackbone=Qwen2.5-7B-Instruct, Ratio=0%, Recovery finetuning=false2026.06 | 78.47 | |
| Qwen2.5-7B-InstructBackbone=Qwen2.5-7B-Instruct, Ratio=0%, Recovery finetuning=true2026.06 | 78.47 | |
| Persona-PrunerRatio=50%, Recovery finetuning=w/, Base Model=Llama-3.2-3B-Instruct2026.06 | 78.24 | |
| Persona-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=25%, Recovery finetuning=false2026.06 | 75.6 | |
| Persona-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=50%, Recovery finetuning=false2026.06 | 74.91 | |
| Adapt-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=25%, Recovery finetuning=false2026.06 | 73.97 | |
| Qwen2.5-3B-InstructBackbone=Qwen2.5-3B-Instruct, Ratio=0%, Recovery finetuning=false2026.06 | 71.71 | |
| Qwen2.5-3B-InstructBackbone=Qwen2.5-3B-Instruct, Ratio=0%, Recovery finetuning=true2026.06 | 71.71 | |
| LLM-PrunerBackbone=Qwen2.5-32B-Instruct, Sparsity=50%2026.06 | 71.01 | |
| LLM-PrunerRatio=25%, Recovery finetuning=w/o, Base Model=Llama-3.2-3B-Instruct2026.06 | 70.64 | |
| LLM-PrunerRatio=25%, Recovery finetuning=w/o, Base Model=Llama-3.1-8B-Instruct2026.06 | 68.59 | |
| LLM-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=25%, Recovery finetuning=false2026.06 | 67.79 | |
| Persona-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=50%, Recovery finetuning=false2026.06 | 65.8 | |
| Persona-PrunerRatio=50%, Recovery finetuning=w/o, Base Model=Llama-3.1-8B-Instruct2026.06 | 64.23 | |
| LLM-PrunerRatio=25%, Recovery finetuning=w/, Base Model=Llama-3.2-3B-Instruct2026.06 | 62.99 | |
| LLM-PrunerRatio=25%, Recovery finetuning=w/, Base Model=Llama-3.1-8B-Instruct2026.06 | 62.45 | |
| Adapt-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=25%, Recovery finetuning=false2026.06 | 57.25 | |
| LLM-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=25%, Recovery finetuning=true2026.06 | 57 | |
| Persona-PrunerRatio=50%, Recovery finetuning=w/o, Base Model=Llama-3.2-3B-Instruct2026.06 | 54.69 | |
| Depth PruningRatio=25%, Recovery finetuning=w/, Base Model=Llama-3.2-3B-Instruct2026.06 | 53.69 | |
| Adapt-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=25%, Recovery finetuning=true2026.06 | 53.05 | |
| Adapt-PrunerRatio=25%, Recovery finetuning=w/, Base Model=Llama-3.2-3B-Instruct2026.06 | 52.86 | |
| Adapt-PrunerRatio=25%, Recovery finetuning=w/, Base Model=Llama-3.1-8B-Instruct2026.06 | 52.63 | |
| Adapt-PrunerRatio=25%, Recovery finetuning=w/o, Base Model=Llama-3.1-8B-Instruct2026.06 | 51.3 | |
| Depth PruningBackbone=Qwen2.5-7B-Instruct, Ratio=25%, Recovery finetuning=false2026.06 | 51.09 | |
| Adapt-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=25%, Recovery finetuning=true2026.06 | 50.9 | |
| Adapt-PrunerRatio=25%, Recovery finetuning=w/o, Base Model=Llama-3.2-3B-Instruct2026.06 | 50.74 | |
| LLM-PrunerRatio=50%, Recovery finetuning=w/, Base Model=Llama-3.1-8B-Instruct2026.06 | 50.37 | |
| Depth PruningRatio=25%, Recovery finetuning=w/, Base Model=Llama-3.1-8B-Instruct2026.06 | 49.07 | |
| Depth PruningBackbone=Qwen2.5-7B-Instruct, Ratio=25%, Recovery finetuning=true2026.06 | 47.99 | |
| Adapt-PrunerBackbone=Qwen2.5-32B-Instruct, Sparsity=50%2026.06 | 47.78 | |
| Depth PruningRatio=25%, Recovery finetuning=w/o, Base Model=Llama-3.1-8B-Instruct2026.06 | 47.3 | |
| LLM-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=25%, Recovery finetuning=true2026.06 | 45.17 | |
| LLM-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=50%, Recovery finetuning=true2026.06 | 43.52 | |
| LLM-PrunerRatio=50%, Recovery finetuning=w/, Base Model=Llama-3.2-3B-Instruct2026.06 | 42.92 | |
| LLM-PrunerRatio=50%, Recovery finetuning=w/o, Base Model=Llama-3.1-8B-Instruct2026.06 | 42.92 | |
| Depth PruningBackbone=Qwen2.5-3B-Instruct, Ratio=25%, Recovery finetuning=true2026.06 | 42.7 | |
| Depth PruningRatio=50%, Recovery finetuning=w/, Base Model=Llama-3.1-8B-Instruct2026.06 | 42.04 | |
| Adapt-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=50%, Recovery finetuning=true2026.06 | 41.74 | |
| Adapt-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=50%, Recovery finetuning=true2026.06 | 39.64 | |
| LLM-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=25%, Recovery finetuning=false2026.06 | 39.07 | |
| Adapt-PrunerRatio=50%, Recovery finetuning=w/, Base Model=Llama-3.2-3B-Instruct2026.06 | 36.26 | |
| Depth PruningBackbone=Qwen2.5-3B-Instruct, Ratio=50%, Recovery finetuning=true2026.06 | 36.11 | |
| Depth PruningRatio=50%, Recovery finetuning=w/, Base Model=Llama-3.2-3B-Instruct2026.06 | 35.6 | |
| LLM-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=50%, Recovery finetuning=true2026.06 | 35.37 | |
| Adapt-PrunerRatio=50%, Recovery finetuning=w/, Base Model=Llama-3.1-8B-Instruct2026.06 | 34.25 | |
| Adapt-PrunerRatio=50%, Recovery finetuning=w/o, Base Model=Llama-3.1-8B-Instruct2026.06 | 30.25 | |
| SliceGPTBackbone=Qwen2.5-7B-Instruct, Ratio=25%, Recovery finetuning=true2026.06 | 30.09 | |
| Adapt-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=50%, Recovery finetuning=false2026.06 | 27.84 | |
| LLM-PrunerRatio=50%, Recovery finetuning=w/o, Base Model=Llama-3.2-3B-Instruct2026.06 | 26.6 | |
| Depth PruningBackbone=Qwen2.5-7B-Instruct, Ratio=50%, Recovery finetuning=true2026.06 | 26.41 | |
| Depth PruningBackbone=Qwen2.5-3B-Instruct, Ratio=25%, Recovery finetuning=false2026.06 | 22.19 | |
| LLM-PrunerBackbone=Qwen2.5-7B-Instruct, Ratio=50%, Recovery finetuning=false2026.06 | 20.93 | |
| Depth PruningRatio=25%, Recovery finetuning=w/o, Base Model=Llama-3.2-3B-Instruct2026.06 | 20.61 | |
| Depth PruningBackbone=Qwen2.5-32B-Instruct, Sparsity=50%2026.06 | 20.12 | |
| SliceGPTRatio=25%, Recovery finetuning=w/, Base Model=Llama-3.1-8B-Instruct2026.06 | 19.1 | |
| SliceGPTRatio=25%, Recovery finetuning=w/, Base Model=Llama-3.2-3B-Instruct2026.06 | 18.79 | |
| SliceGPTBackbone=Qwen2.5-7B-Instruct, Ratio=50%, Recovery finetuning=true2026.06 | 17.59 | |
| Adapt-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=50%, Recovery finetuning=false2026.06 | 15.27 | |
| SliceGPTRatio=25%, Recovery finetuning=w/o, Base Model=Llama-3.1-8B-Instruct2026.06 | 14.62 | |
| SliceGPTBackbone=Qwen2.5-3B-Instruct, Ratio=25%, Recovery finetuning=true2026.06 | 13.65 | |
| Adapt-PrunerRatio=50%, Recovery finetuning=w/o, Base Model=Llama-3.2-3B-Instruct2026.06 | 10.09 | |
| Depth PruningBackbone=Qwen2.5-3B-Instruct, Ratio=50%, Recovery finetuning=false2026.06 | 7.71 | |
| SliceGPTRatio=50%, Recovery finetuning=w/, Base Model=Llama-3.2-3B-Instruct2026.06 | 7.03 | |
| LLM-PrunerBackbone=Qwen2.5-3B-Instruct, Ratio=50%, Recovery finetuning=false2026.06 | 6.53 | |
| SliceGPTBackbone=Qwen2.5-3B-Instruct, Ratio=50%, Recovery finetuning=true2026.06 | 5.68 | |
| SliceGPTRatio=25%, Recovery finetuning=w/o, Base Model=Llama-3.2-3B-Instruct2026.06 | 5.32 | |
| SliceGPTBackbone=Qwen2.5-7B-Instruct, Ratio=25%, Recovery finetuning=false2026.06 | 4.66 | |
| SliceGPTRatio=50%, Recovery finetuning=w/, Base Model=Llama-3.1-8B-Instruct2026.06 | 3.55 | |
| Depth PruningBackbone=Qwen2.5-7B-Instruct, Ratio=50%, Recovery finetuning=false2026.06 | 3.5 | |
| Depth PruningRatio=50%, Recovery finetuning=w/o, Base Model=Llama-3.2-3B-Instruct2026.06 | 2.39 | |
| Depth PruningRatio=50%, Recovery finetuning=w/o, Base Model=Llama-3.1-8B-Instruct2026.06 | 1.3 | |
| SliceGPTBackbone=Qwen2.5-32B-Instruct, Sparsity=50%2026.06 | 1.04 | |
| SliceGPTRatio=50%, Recovery finetuning=w/o, Base Model=Llama-3.2-3B-Instruct2026.06 | 1.02 | |
| SliceGPTRatio=50%, Recovery finetuning=w/o, Base Model=Llama-3.1-8B-Instruct2026.06 | 0.68 | |
| SliceGPTBackbone=Qwen2.5-3B-Instruct, Ratio=25%, Recovery finetuning=false2026.06 | 0.34 | |
| SliceGPTBackbone=Qwen2.5-7B-Instruct, Ratio=50%, Recovery finetuning=false2026.06 | 0.05 | |
| SliceGPTBackbone=Qwen2.5-3B-Instruct, Ratio=50%, Recovery finetuning=false2026.06 | 0.03 |