MapSatisfyBench Evaluates LLM Agents on Implicit User Needs in Map Queries

MapSatisfyBench targets the gap between underspecified everyday map queries and the implicit decision factors that determine user satisfaction, evaluating whether agents can recover those factors without requiring clarification. The benchmark addresses the challenge that user satisfaction cannot be captured by a single reference answer.