Language Model General Utility on General Utility Suite IFB ARC HS GSM8K (test)
19.8IFB ScoreBaseline
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| BaselineBackbone=Gemma-2-9B-IT2026.05 | 19.8 | 59 | 70.2 | 79 | — | |
| Angular SteerBackbone=Gemma-2-9B-IT2026.05 | 18.1 | 55.3 | 51.4 | 58.2 | 4.5 | |
| SKOPBackbone=Gemma-2-9B-IT2026.05 | 18 | 56.5 | 59 | 66.4 | 2.44 | |
| CASTBackbone=Gemma-2-9B-IT2026.05 | 16.5 | 38.2 | 55.9 | 39.7 | 5.25 | |
| LoRABackbone=Gemma-2-9B-IT2026.05 | 15.2 | 35.6 | 49.3 | 35.5 | — | |
| SADIBackbone=Gemma-2-9B-IT2026.05 | 13.8 | 45.6 | 52.6 | 55.1 | 3.69 | |
| CAABackbone=Gemma-2-9B-IT2026.05 | 12.6 | 22.2 | 35.2 | 16.8 | 4.38 | |
| DISCO-QBackbone=Gemma-2-9B-IT2026.05 | 10.5 | 31.8 | 36.6 | 20.4 | 5.13 | |
| ITIBackbone=Gemma-2-9B-IT2026.05 | 9.8 | 27.4 | 43.1 | 18.2 | 6 | |
| Comm SteerBackbone=Gemma-2-9B-IT2026.05 | 5.2 | 14.3 | 33.8 | 9.6 | 4.63 |