Honda to Present Multi-Agent AI Model at ICLR AgenticAI Workshop
TOKYO JAPAN, 17TH APRIL 2025 — Honda’s landmark research paper on multi-agent AI systems has been accepted for presentation at the AgenticAI Workshop as part of the prestigious ICLR 2025 conference, to be held in Singapore between April 24 and April 28, 2025.
The paper, “Dynamic Knowledge Integration in Multi-Agent Systems for Content Inference”, investigates how LLM-based AI agents can mimic collaborative processes of real-world multi-expert interactions with Honda’s unique Waigaya culture of open, dynamic discussions as an inspiration.
This research has been carried out by Honda’s Advanced AI Strategy Planning Department associates.
Recognizing the limitations of single-agent LLM systems in addressing interdisciplinary challenges, the Honda team proposed a multi-agent approach where agents of distinct areas of expertise engage in structured discussion to solve complex problems.
The study tested four interaction styles: Decentralized, Centralized, Layered, and Shared Pool. The Waigaya-inspired Decentralized method outperformed the others in idea diversity and solution quality.
The research highlighted the way in which multi-agent systems would greatly enhance content generation through accuracy and stability over single-agent systems, an important prerequisite in fields like mobility development, where consensus-building across differing experts is essential.
Honda intends to take this AI system forward for real-life application throughout its product development process and, in so doing, promote consensus-based collaboration as an enabler for innovation and acceleration.
Recognition by the ICLR cements Honda’s position at the forefront of AI research and its ongoing commitment to integrating human-inspired problem-solving approaches into advanced technologies.
Accelya Launches FLX ONE Cargo to Revolutionize Air Cargo Tech
Infosys and Spark Join Forces for AI-Driven Transformation
NVIDIA’s Blackwell Ultra: The AI Leap That Changes Everything
