Instruction Tuning on Instruction Tuning Datasets 1.0 (train test)
1.45Model PerformanceK-Center-Greedy
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| K-Center-GreedyBackbone Model=Qwen-2.5-7B, Data Selection Strategy=Core-set based2025.02 | 1.45 | 0.505 | 3.09 | 0.44 | 3.42 | |
| Repr FilterBackbone Model=Qwen-2.5-7B2025.02 | 1.35 | 0.495 | 7.15 | 0.421 | 3.46 | |
| K-meansBackbone Model=LLaMA-3-8B, Data Selection Strategy=Clustering-based2025.02 | 1.32 | 0.693 | 1.7 | 0.648 | 2.99 | |
| K-Center-GreedyBackbone Model=LLaMA-3-8B, Data Selection Strategy=Core-set based2025.02 | 1.31 | 0.687 | 2.53 | 0.746 | 2.73 | |
| QDITBackbone Model=LLaMA-3-8B2025.02 | 1.25 | 0.673 | 1.59 | 0.629 | 2.99 | |
| QDITBackbone Model=Qwen-2.5-7B2025.02 | 1.23 | 0.403 | 2.6 | 0.223 | 3.54 | |
| X_allBackbone Model=LLaMA-3-8B, Selection Strategy=Random2025.02 | 1.2 | 0.675 | 1.61 | 0.634 | 2.99 | |
| K-meansBackbone Model=Qwen-2.5-7B, Data Selection Strategy=Clustering-based2025.02 | 1.06 | 0.44 | 1.6 | 0.26 | 3.54 | |
| Repr FilterBackbone Model=LLaMA-3-8B2025.02 | 1.05 | 0.671 | 2.23 | 0.703 | 2.86 | |
| X_allBackbone Model=Qwen-2.5-7B, Selection Strategy=Random2025.02 | 0.87 | 0.408 | 1.41 | 0.23 | 3.54 | |
| ShareGPTBackbone Model=LLaMA-3-8B, Selection Strategy=Random2025.02 | 0.83 | 0.628 | 1.7 | 0.656 | 2.83 | |
| WizardLMBackbone Model=LLaMA-3-8B, Selection Strategy=Random2025.02 | 0.72 | 0.591 | 1.44 | 0.578 | 2.88 | |
| AlpacaBackbone Model=LLaMA-3-8B, Selection Strategy=Random2025.02 | 0.07 | 0.572 | 1.32 | 0.605 | 2.83 | |
| ShareGPTBackbone Model=Qwen-2.5-7B, Selection Strategy=Random2025.02 | 0.07 | 0.392 | 3.36 | 0.285 | 3.51 | |
| WizardLMBackbone Model=Qwen-2.5-7B, Selection Strategy=Random2025.02 | -0.08 | 0.349 | 2.65 | 0.211 | 3.5 | |
| DollyBackbone Model=LLaMA-3-8B, Selection Strategy=Random2025.02 | -0.14 | 0.5 | 1.44 | 0.603 | 2.59 | |
| AlpacaBackbone Model=Qwen-2.5-7B, Selection Strategy=Random2025.02 | -0.38 | 0.336 | 1.89 | 0.189 | 3.5 | |
| DuplicateBackbone Model=Qwen-2.5-7B, Selection Strategy=Random2025.02 | -0.43 | 0.309 | 0.2 | 0.243 | 3.48 | |
| DollyBackbone Model=Qwen-2.5-7B, Selection Strategy=Random2025.02 | -0.49 | 0.32 | 3.04 | 0.221 | 3.46 | |
| DuplicateBackbone Model=LLaMA-3-8B, Selection Strategy=Random2025.02 | -1.35 | 0.461 | 0.05 | 0.634 | 2.52 |