ResearchBenchmarksBias Evaluation on BOLD (test)Follow0.0502Bias Score (Gender)Vanilla0.0469080.0691290.091350.113571Jun 30, 2026Evaluation ResultsMethodMethodLinksBias Score (Gender)Bias Score (Race)Bias Score (Religious Ideology)Bias Score (Political Ideology)Bias Score (Profession)Average Bias ScoreVanillaBackbone=Llama 3.1 8BBackbone=Llama 3.1 8B2026.060.05020.19851.36140.9240.84150.6751Fair-GCGBackbone=Llama 3.1 8BBackbone=Llama 3.1 8B2026.060.06850.15191.42450.89240.72810.6531VanillaBackbone=Qwen 2.5 7BBackbone=Qwen 2.5 7B2026.060.12820.10661.43960.91320.87340.6922Fair-GCGBackbone=Qwen 2.5 7BBackbone=Qwen 2.5 7B2026.060.13250.13621.47410.88640.77590.681