“judging”

LLM-as-a-Judge for Agent Outputs: Strengths and Pitfalls

How to use LLMs as evaluators for agent outputs — the strengths, systematic biases, and practical debiasing strategies for production evaluation.