🏢 City University of Hong Kong

Crowd-based comparative evaluation significantly boosts LLM-as-a-judge accuracy by using crowd responses to expose deeper details, resulting in more reliable and efficient auto-evaluation.