Open Science Robotics Foundation Models for Real-world Deployment with MolmoAct

🤖 Vision-language-action (VLA) models are at the forefront of enabling robots to tackle a multitude of tasks without the need for hand-engineered policies. This talk will explore important advancements and challenges in deploying these models in real-world settings.

🚀 Prof. Jiafei Duan from the National University of Singapore will discuss two groundbreaking projects from the Allen Institute for AI and the University of Washington: MolmoAct and MolmoAct2.

• MolmoAct introduces Action Reasoning Models, which transform observations and language instructions into depth-aware perception tokens, editable trajectory traces, and low-level robot actions.
• MolmoAct2 enhances practical deployment by providing open datasets, an embodied reasoning vision-language model, support for a variety of robot platforms, and faster closed-loop control through adaptive-depth reasoning.

🗣️ Prof. Jiafei will also touch upon the significance of openness in advancing robotics foundation models. Robust research depends on not just model weights but also access to datasets, tokenizers, and evaluation protocols.

More About the Speaker

Jiafei Duan is an Assistant Professor under the Presidential Young Professorship scheme and has made significant contributions to scalable data collection, robotic reasoning, and generalization in robot learning. His work has garnered multiple awards in reputable AI and robotics conferences.

More About the Series

Praxis is a focused technical sharing series revolving around advanced AI and applied research discussions.

🔗 Get involved: Learn more about Lorong AI | Speaker Sign-up | WhatsApp Community | LinkedIn | X