hard20 pts · first answer countsllm-evaluationagent-eval

Agent results vary by run order

An agent eval runs 100 tasks against one shared test database. Results change depending on which tasks run first. What is the fix?

A.Always run tasks in alphabetical orderB.Run each task only once to avoid changesC.Ignore tasks whose results varyD.Give each task a fresh, identical environment state

Discussion 0

Join free to post in the discussion.

Agent results vary by run order

Discussion 0

Related questions