Qwen's Justin Lin: spatial reasoning is a clear gap holding back physical agents

JustinLin610 · x · 2026-11-18

Justin Lin (Qwen team) notes that spatial reasoning has been a clear weakness of vision-language models this year, especially for open-weight models versus frontier closed ones, and argues this gap makes it hard to build high-quality physical agents with current VLMs.

Original post →

More from AGI Musings

AGI Musings channel →