You are designing an evaluation for an agent. The test should compare the agent’s free-text summary of a ticket with an expected human summary, rewarding answer...