List your company

Sequential Model Defense on Gemma-S Case II: Base -> Unbias -> CAT

2.98Fairness ScoreUnbias
2.5045.7178.9312.143May 14, 2026

Evaluation results

MethodLinks
UnbiasDefense=+D12026.052.9866.1
Conflict-Guided Layer FreezingDefense=+D1+D2 (frozen L9)2026.056.5629.8
Unbias -> CATDefense=+D1+D22026.058.0831.9
BaseDefense=Base2026.0514.8855.9