1 paper
Hantao Zhang, Jinru Sui, Ed Li +2
Recent benchmarks for VLMs largely assess single- or limited-view perception, leaving untested the core cognitive ability to integrate observations across viewpoints into a coheren…