Collaborative Memory Architectures for Multi-Agent VLM Systems

This framework addresses distributed perception in multi-agent vision-language model (VLM) systems through shared visual memory. It focuses on reconciling differing interpretations of image regions and video frames to maintain consistency across agents' reasoning chains.