Do AI models want to be watched? Measuring monitorability disposition in large reasoning models

Labelbox research examines whether large reasoning models exhibit a measurable disposition toward being monitored, introducing a way to quantify monitorability in frontier AI systems.