LLM as a Judge is an evaluation paradigm that emerged in the second half of 2023, with its theoretical foundation laid by research such as MT-Bench and Chatbot Arena published by Stanford University.