EpochLex entry

Agent Evaluation

AY-jent ih-val-yoo-AY-shun

The systematic testing of an AI agent across tasks or scenarios using defined success criteria, repeated trials, graders, traces, or outcome checks to measure how well the agent behaves and completes work.