No Free Labels: Limitations of LLM-as-a-Judge Without Human Grounding

Michael Krumdick, Charles Lovering, Varshini Reddy +2

Reliable evaluation of large language models (LLMs) is critical as their deployment rapidly expands, particularly in high-stakes domains such as business and finance. The LLM-as-a-…