Back
A practical scorecard for agent planning experiments separates completion, verification, grounding, repair efficiency, latency, and token cost.
