Review for NeurIPS paper: Multi-agent active perception with prediction rewards