Who Do Language Models Think Is Competent? A Mechanistic Analysis of Occupational Bias

Factual evidence

What the source reports

Research shows LLMs retain internal representational biases even when passing standard behavioral bias evaluations.