Tag
This paper introduces ChildAgentEval, a psychometrically grounded benchmark for assessing cognitive age alignment in MLLM-based agents, comparing their reasoning against human developmental stages.