Interview Guides
You are evaluating an LLM-based agent and need a framework that reflects real-world quality, not just whether the final answer matches a reference. The team wants to know how you would measure agent performance when tool calls, reasoning quality, and failure modes all matter.
Explain your approach to evaluating LLM-based agents, what metrics do you prioritize beyond standard accuracy?
