When robot data hits VLM pipelines, VLA vs WAM distinction will vanish

ChongZzZhang · x · 2026-09-06

ChongZzZhang predicts that once robot data enters VLM training pipelines and video becomes a native modality (in and out), the distinction between pi-style VLA and WAM will disappear — and code-as-policy might turn out to be the stronger approach.

Original post →

More from Embodied

Embodied channel →