Paper Accepted to Transactions on Machine Learning Research (TMLR)

Our paper, which presents a comparative analysis of robot foundation models from the perspective of the Embodiment Gap, has been accepted for publication in Transactions on Machine Learning Research (TMLR). The study analyzes the limitations of current robot foundation models and proposes key requirements for enabling robust operation in real-world environments.

Yukiyasu Domae, Keisuke Shirai, Hanbit Oh, Ryoichi Nakajo, Tomohiro Motoda, Koshi Makihara, Masaki Murooka, Takuma Yagi, Yoshisaki Bando, and Ryo Hanai, “The Embodiment Gap in Robot Foundation Models”
https://openreview.net/pdf?id=D0XcH9Cso4

This work is an outcome of the JST CREST project Real-World Intelligent Systems, “MORAL: Morphoception-Oriented Reasoning and Action with Language” (JPMJCR2553), and serves as a position paper that outlines the project’s research scope and vision. In addition, the paper includes a discussion of the differing approaches to foundation models pursued under the FRONTia program and the AIRoA program (JPNP25015), highlighting these distinctions as one of its key contributions.

@article{
domae2026the,
title={The Embodiment Gap in Robot Foundation Models},
author={Yukiyasu Domae and Keisuke Shirai and Hanbit Oh and Ryoichi Nakajo and Tomohiro Motoda and Koshi Makihara and Masaki Murooka and Takuma Yagi and Yoshiaki Bando and Ryo Hanai},
journal={Transactions on Machine Learning Research},
issn={2835-8856},
year={2026},
url={https://openreview.net/forum?id=D0XcH9Cso4},
note={}
}