LLM-as-judge
A separate language model that evaluates matched male-female responses for bias in wording, tone, structure, content and stereotypes.
Source: The Indian Financial Automated Advice Bias (IFAAB) Dataset: Auditing Socioeconomic and Demographic Disparities in LLM-Generated Financial Advice in India (Academic research) · p.4
Have a question about LLM-as-judge? Ask the AI, answers cite the source.