ASIA: LLM-Powered Autonomous Agent for System Identification
A new framework called ASIA (Autonomous System Identification Agent) has been developed by researchers to utilize large language models for automating the iterative system identification process. Typically, the selection of model classes, training algorithms, and hyperparameters involves considerable expert trial-and-error. ASIA, which is constructed on existing agentic platforms, effectively links hypothesis, implementation, and evaluation autonomously, needing only a straightforward English description of the issue. An empirical investigation on two benchmarks examines the search behavior of the agent and the architectures it identifies.
Key facts
- ASIA stands for Autonomous System Identification Agent.
- It uses a large language model as an autonomous coding agent.
- The framework requires only a plain-English description of the identification problem.
- It closes the loop between hypothesis, implementation, and evaluation without human intervention.
- An empirical study was conducted on two system identification benchmarks.
- The study analyzed the agent's search behavior and discovered architectures and training strategies.
- The research is published on arXiv with ID 2605.10480.
- The work is motivated by advances in agentic artificial intelligence.
Entities
Institutions
- arXiv