LMMs-Lab: Multimodality Is a Precondition for Next-Level Intelligence

liuziwei7 · x · 2026-11-17

Open-source multimodal team LMMs-Lab pushes back on claims that multimodality is useless or harmful, arguing frontier models like Astra/Fable already show impressive environment interaction, and that multimodality is a precondition for next-level intelligence. Their site lists work including NEO-VLA, LLaVA-OneVision-2, OneVision Encoder, and LongVT.

Original post →

More from Multimodal

Multimodal channel →