@misc {pub6458,
	title = {Actionable intelligence: Tool use at scale for embodied agents},
	author = {Felix Ocker},
	year = {2025},
	month = {December},
	abstract = {In recent years, the field of artificial intelligence has witnessed significant progress, particularly with the advent of powerful language models. This evolution has catalyzed a growing interest in autonomous agents capable of complex reasoning and decision-making. Central to their practical deployment is tool use - a mechanism that enables agents to interact with and act upon the world - and the planning capabilities required to orchestrate such interactions effectively. However, designing agents that can robustly utilize a wide range of tools presents key challenges. These include scalability concerns around the number of tools made available, as well as the computational and financial costs associated with managing extensive tool repertoires. Tulip Agent addresses these issues by combining recursive task decomposition with a modular, extensible, and scalable tool library. This talk provides an overview of the Tulip Agent architecture and demonstrates its capabilities through real-world applications in robotics and the automotive domain. The presentation highlights how Tulip balances generality with efficiency and opens up new directions for actionable, tool-using AI agents in dynamic environments.},
	publisher = {Honda internal},
	booktitle = {Honda Technical Forum}
}