Research

Models that predict how systems change under action.

Training-free confidence for visual grounding.

Agents that use browsers to act and gather evidence.

Video models predicting physical outcomes.