Question Answering on OfficeQA (test)
55.32AccuracyMetaSkill-Evolve
Evaluation Results
| Method | Links | |
|---|---|---|
| MetaSkill-EvolveBackbone=Gemma-4 31B2026.07 | 55.32 | |
| Single-LevelBackbone=Gemma-4 31B2026.07 | 48.94 | |
| Static skillBackbone=Gemma-4 31B2026.07 | 36.09 | |
| No-SkillBackbone=Gemma-4 31B2026.07 | 31.78 |