When to Ask for Help: A Decision-Theoretic Account and Measurement Standard for Slow-Module Scheduling in Dual-System Vision-Language-Action Models
The paper formalizes when to use a slow, expensive module in AI models by treating it as a calibrated decision problem. It introduces the Value of Intervention (VoI) and cost ratios ($ ho$) to turn scheduling from an engineering guess into a measurable economic optimization. This allows researchers to design triggers based on quantifiable metrics rather than intuition, turning invocation into a data-driven decision.