Research · 6 November 2026
ResearchNEO-VLA proposes direct visual computation for vision‑language‑action policies
NEO-VLA challenges the standard encoder‑first pipeline by allowing action policies to access evolving visual computation directly, targeting tasks that depend on fine‑grained geometry, depth, and subtle state changes.
Why it matters
NEO-VLA suggests building action models that ingest raw visual streams rather than static embeddings; prototype such pipelines.
For product managers
For your product
What does this mean for your product?
Try it on your product More from this issueAll AI news from 6 Nov →
Get this every morning, written for your product.
Deltaday reads the whole AI news cycle and tells you what changed, whether it's real yet, and what to do about it.