MA-VLCM: A Vision Language Critic Model for Value Estimation of Policies in Multi-Agent Team Settings