#AI agent evaluation

LangChain, Conviva, CoreWeave: evaluating AI agents on single conversations is not enough, cohort comparison needed

A single exchange with an AI agent can look flawless when scored in isolation and still point to a broken product. This...