OpenWAM

Open framework for world-action model pretraining research

Visit OpenWAM ↗ link: rel="ugc noopener"

OpenWAM is an official research repository providing an open, modular framework for exploring world-action model pretraining. It supports systematic experimentation with foundation models that learn to predict and reason about world dynamics and agent actions.

Designed for researchers and developers working on embodied AI and predictive modeling, OpenWAM enables reproducible exploration of pretraining approaches across diverse environments and tasks.

Discussion