On the Clock: Towards Punctual and Productive Time-Budgeted AI Agents

We study whether small LLM agents can operate effectively under explicit wall-clock time budgets by both respecting the allocated runtime and using available time productively.