Multimodal AI

A public release of SCOUT++, a derivative multimodal human-robot instruction dataset built from the ARL SCOUT corpus — 12,000 model input rows, 12,003 aligned Navigator-view images, plus schema, sample data, and reproducible processing artifacts.