arXiv cs.AISaturday · May 23, 2026FREE

Tool-Augmented Agent for Closed-loop Optimization,Simulation,and Modeling Orchestration

COSMO-Agent (Closed-loop Optimization, Simulation, and Modeling Orchestration) is a tool-augmented reinforcement learning (RL) framework that enables LLMs to perform iterative industrial design-simulation optimization. It addresses the CAD-CAE semantic gap by casting CAD generation, CAE solving, result parsing, and geometry revision as an interactive RL environment. The LLM learns to orchestrate external tools and revise parametric geometries until constraints are satisfied. A multi-constraint reward jointly encourages feasibility, toolchain robustness, and structured output validity. The framework includes an industry-aligned dataset covering 25 component categories with executable CAD-CAE tasks. Experiments show that COSMO-Agent training substantially improves small open-source LLMs for constraint-driven design, exceeding large open-source models. The paper is available on arXiv (2605.20190) and was published on May 22, 2026.