Instruction Tuning on WizardLM
75.07Reasoning ScoreADG
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ADGBackbone=Qwen2.5-7B, Selection Budget=10K2026.04 | 75.07 | 66.67 | 61.19 | 67.64 | |
| MIGBackbone=Qwen2.5-7B, Selection Budget=10K2026.04 | 74.1 | 65.68 | 59.28 | 66.35 | |
| SuperFilteringBackbone=Qwen2.5-7B, Selection Budget=10K2026.04 | 73.94 | 65.23 | 57.55 | 65.57 | |
| SCARBackbone=Qwen2.5-7B, Selection Budget=10K2026.04 | 73.65 | 64.44 | 57.36 | 65.15 | |
| IFDBackbone=Qwen2.5-7B, Selection Budget=10K2026.04 | 73.22 | 65.99 | 53.19 | 64.13 | |
| RethinkingBackbone=Qwen2.5-7B, Selection Budget=10K2026.04 | 72.49 | 64.59 | 55.85 | 64.31 | |
| RandomBackbone=Qwen2.5-7B, Selection Budget=10K2026.04 | 72.26 | 64 | 57.97 | 64.74 | |
| SelectITBackbone=Qwen2.5-7B, Selection Budget=10K2026.04 | 71.79 | 64.16 | 55.17 | 63.71 | |
| ZIPBackbone=Qwen2.5-7B, Selection Budget=10K2026.04 | 71.12 | 65.22 | 57.26 | 64.53 | |
| All DataBackbone=Qwen2.5-7B, Selection Budget=Full2026.04 | 70.79 | 64 | 55.11 | 63.3 | |
| ADGBackbone=LLaMA3-8B, Selection Budget=10K2026.04 | 52.53 | 57.61 | 45.06 | 51.73 | |
| All DataBackbone=LLaMA3-8B, Selection Budget=Full2026.04 | 51.6 | 54.2 | 42.85 | 49.55 | |
| SuperFilteringBackbone=LLaMA3-8B, Selection Budget=10K2026.04 | 51.09 | 57.87 | 42.11 | 50.35 | |
| IFDBackbone=LLaMA3-8B, Selection Budget=10K2026.04 | 50.34 | 57.73 | 42.52 | 50.19 | |
| RandomBackbone=LLaMA3-8B, Selection Budget=10K2026.04 | 48.38 | 56.89 | 38.81 | 48.03 | |
| MIGBackbone=LLaMA3-8B, Selection Budget=10K2026.04 | 46.95 | 55.74 | 41.32 | 48 | |
| RethinkingBackbone=LLaMA3-8B, Selection Budget=10K2026.04 | 45.05 | 54.03 | 36.68 | 45.25 | |
| SCARBackbone=LLaMA3-8B, Selection Budget=10K2026.04 | 45.04 | 54.21 | 33.64 | 44.29 | |
| SelectITBackbone=LLaMA3-8B, Selection Budget=10K2026.04 | 45.02 | 53.42 | 37.67 | 45.37 | |
| ZIPBackbone=LLaMA3-8B, Selection Budget=10K2026.04 | 44.66 | 56.7 | 36.47 | 45.94 |