Reproduce “ASIA: an Autonomous System Identification Agent” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Dario Piga, Marco Forgione
The Dalle Molle Institute for Artificial Intelligence introduced ASIA, an Autonomous System Identification Agent powered by a Large Language Model, which autonomously navigates the system identification pipeline to discover effective models and training strategies. The agent achieved competitive performance on both a cascaded two-tank system and a nanodrone dynamics benchmark, reducing the need for extensive human expert intervention.
Key insights
- LLM-based autonomous agents can effectively explore a rich and potentially unbounded hypothesis space in system identification, allowing for the generation of conceptually diverse modeling solutions.
- The agent demonstrated the ability to autonomously discover non-trivial and often physics-informed architectural designs and effective training strategies, such as LTC/CfC networks and scheduled sampling.
- Agentic AI can serve as a complementary tool to human expertise, automating iterative tasks and freeing researchers to focus on higher-level problem formulation and validation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.