Task Horizon

A concept in AI research used to measure the boundary of an agent's autonomous capability, defined as the number of operational steps a model can complete independently without seeking human feedback. It is used to evaluate an agent's sustained reasoning ability in multi-step complex tasks.