Researchers · Interviewed at COLM 2026
Carnegie Mellon University

COLM 2026 · October 16, 2026 · 12 min
Sticking chopsticks upright in a bowl of rice is offensive in China because it echoes funeral offerings. Chopsticks resting across the bowl mean nothing at all. Can today's vision-language models tell the difference? Mostly, no.
In this COLM 2026 poster conversation, Akhila Yerukola (PhD student at Carnegie Mellon University, advised by Maarten Sap) presents a new benchmark for culturally grounded visual reasoning. It is built from contrastive image pairs: the same scene, with one small change in behavior that flips its cultural meaning.