Evaluation
Measure whether agents, workflows, and intelligent applications meet the requirements you set before and after release.
Evaluation asks whether a system behaved correctly for a defined requirement. Monitoring tells you what happened in production; evaluation tells you whether that behaviour was good enough. You need both.
Evaluate more than the final text. Depending on the job, assess groundedness, tool selection, action correctness, policy compliance, latency, cost, and whether the system stopped when it should have. The evidence should be attached to the version you promote.
2026년 8월 31일 업데이트