Preference Alignment Data Selection on Aggregate Utility-Robustness Preference Pool (test)
97.68Robustness Score (R)Whole
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| WholeBase Model=LLaMA3.2-3B, Selection Budget=10k, Tier=T32026.06 | 97.68 | 60.25 | |
| WholeBase Model=LLaMA3.2-3B, Selection Budget=30k, Tier=T32026.06 | 97.68 | 60.25 | |
| DOG-DBase Model=LLaMA3.2-3B, Selection Budget=30k, Tier=T22026.06 | 96.57 | 57.6 | |
| DOGBase Model=LLaMA3.2-3B, Selection Budget=30k, Tier=T22026.06 | 96.08 | 57.2 | |
| RM-RSBase Model=LLaMA3.2-3B, Selection Budget=30k, Tier=T22026.06 | 96.03 | 55.9 | |
| DOG-DBase Model=LLaMA3.2-3B, Selection Budget=10k, Tier=T22026.06 | 95.63 | 55.8 | |
| SDPOBase Model=LLaMA3.2-3B, Selection Budget=10k, Tier=T32026.06 | 93.58 | 55.65 | |
| SDPOBase Model=LLaMA3.2-3B, Selection Budget=30k, Tier=T32026.06 | 91.63 | 56.25 | |
| DOGBase Model=LLaMA3.2-3B, Selection Budget=10k, Tier=T22026.06 | 90.83 | 55.7 | |
| KMeansBase Model=LLaMA3.2-3B, Selection Budget=10k, Tier=T22026.06 | 90.71 | 55.15 | |
| KMeansBase Model=LLaMA3.2-3B, Selection Budget=30k, Tier=T22026.06 | 89.83 | 56.95 | |
| LaserBase Model=LLaMA3.2-3B, Selection Budget=10k, Tier=T22026.06 | 89.65 | 56.3 | |
| LaserBase Model=LLaMA3.2-3B, Selection Budget=30k, Tier=T22026.06 | 86 | 56.3 | |
| RM-RSBase Model=LLaMA3.2-3B, Selection Budget=10k, Tier=T22026.06 | 85.6 | 54.85 | |
| DEITABase Model=LLaMA3.2-3B, Selection Budget=30k, Tier=T22026.06 | 82.86 | 54.64 | |
| RandomBase Model=LLaMA3.2-3B, Selection Budget=30k, Tier=T12026.06 | 81.13 | 54.05 | |
| RandomBase Model=LLaMA3.2-3B, Selection Budget=10k, Tier=T12026.06 | 80.95 | 53.8 | |
| DEITABase Model=LLaMA3.2-3B, Selection Budget=10k, Tier=T22026.06 | 80.26 | 54.38 | |
| SDPOBase Model=Gemma3-1B, Selection Budget=30k, Tier=T32026.06 | 78.88 | 56.25 | |
| DOG-DBase Model=Gemma3-1B, Selection Budget=30k, Tier=T22026.06 | 77.25 | 47.1 | |
| DOGBase Model=Gemma3-1B, Selection Budget=30k, Tier=T22026.06 | 76.9 | 46.2 | |
| DOG-DBase Model=Gemma3-1B, Selection Budget=10k, Tier=T22026.06 | 76.5 | 45.8 | |
| LaserBase Model=Gemma3-1B, Selection Budget=30k, Tier=T22026.06 | 76.5 | 45.55 | |
| KMeansBase Model=Gemma3-1B, Selection Budget=30k, Tier=T22026.06 | 74.85 | 48.65 | |
| BaseBase Model=LLaMA3.2-3B, Selection Budget=10k, Tier=T12026.06 | 74.6 | 52.45 | |
| BaseBase Model=LLaMA3.2-3B, Selection Budget=30k, Tier=T12026.06 | 74.6 | 52.45 | |
| SDPOBase Model=Gemma3-1B, Selection Budget=10k, Tier=T32026.06 | 74.38 | 55.65 | |
| WholeBase Model=Gemma3-1B, Selection Budget=10k, Tier=T32026.06 | 74.38 | 45.89 | |
| WholeBase Model=Gemma3-1B, Selection Budget=30k, Tier=T32026.06 | 74.38 | 45.89 | |
| LaserBase Model=Gemma3-1B, Selection Budget=10k, Tier=T22026.06 | 73.2 | 46.05 | |
| DOGBase Model=Gemma3-1B, Selection Budget=10k, Tier=T22026.06 | 72.7 | 43.8 | |
| RM-RSBase Model=Gemma3-1B, Selection Budget=30k, Tier=T22026.06 | 71.65 | 45.95 | |
| KMeansBase Model=Gemma3-1B, Selection Budget=10k, Tier=T22026.06 | 71.3 | 44.85 | |
| RandomBase Model=Gemma3-1B, Selection Budget=30k, Tier=T12026.06 | 69.2 | 43.5 | |
| RandomBase Model=Gemma3-1B, Selection Budget=10k, Tier=T12026.06 | 67.88 | 42.75 | |
| DEITABase Model=Gemma3-1B, Selection Budget=10k, Tier=T22026.06 | 67.65 | 41.44 | |
| DEITABase Model=Gemma3-1B, Selection Budget=30k, Tier=T22026.06 | 67.65 | 43.47 | |
| RM-RSBase Model=Gemma3-1B, Selection Budget=10k, Tier=T22026.06 | 67.45 | 43.55 | |
| BaseBase Model=Gemma3-1B, Selection Budget=10k, Tier=T12026.06 | 61.55 | 41.04 | |
| BaseBase Model=Gemma3-1B, Selection Budget=30k, Tier=T12026.06 | 61.55 | 41.04 |