PlanContext: Benchmarking In-Context Driving Decision-Making

This paper introduces PlanContext, a benchmark for evaluating in-context driving decision-making. It constructs paired evaluations by varying high-level driving intentions while keeping the observed scene fixed to assess context-dependent decision-making. The study defines six representative driving contexts and evaluates contextual decision correctness, road compliance, and dynamic safety using class-balanced metrics. Experiments on 9,561 instances from NAVSIM and 23 public end-to-end driving models show that high conventional driving-quality scores do not consistently correspond to strong route-level in-context driving decision-making, highlighting the importance of this dimension for evaluating end-to-end driving systems.