COOPA: A Modular LLM Agent Architecture for Operations Research Problems

2026-06-25

AI Summary

Current large language models face significant challenges in operations research modeling, including low accuracy, limited interpretability, and insufficient solver support. This work proposes COOPA, an architecture that iteratively generates and evaluates multiple candidate mathematical models through confidence-aware modeling. COOPA integrates element-wise provenance tracing for interpretability and a multi-solver intelligent routing mechanism to enable scalable, explainable decision support. The framework employs modular LLM agents that jointly perform confidence assessment, traceable natural language-to-mathematical programming mapping, and specialized solver dispatching tailored to problem types such as linear and integer programming. Experimental results across three operations research benchmarks and eight LLMs demonstrate that COOPA achieves the highest macro-average accuracy among six evaluated models, outperforming the strongest baseline by up to 6.7 percentage points.

Abstract

Operations Research (OR) provides a rigorous framework for high-stakes decision-making, but effective OR modeling requires substantial domain knowledge, mathematical abstraction, and solver expertise. Recent LLM-based systems automate parts of this pipeline, yet remain limited by low accuracy on complex problems, opaque outputs, and narrow solver support. We propose COOPA (COoperative OPerations Agent), a modular LLM-agent architecture for interpretable and scalable OR decision support. It combines three components: iterative confidence-based modeling, which generates multiple candidate formulations, self-evaluates them across modeling dimensions, and selects one using a max-min confidence criterion; element-level provenance and confidence explanations, which link variables, parameters, constraints, and objectives to quoted source text and provide an audit trail for human verification; and multi-solver routing to specialized optimizer agents for different OR problem classes. Across three OR benchmarks, eight LLM backbones, and four baselines under identical conditions, COOPA achieves the best macro-average accuracy on six of eight backbones and improves over the strongest baseline by up to 6.7 percentage points. A within-system ablation isolates the contribution of iterative confidence-based modeling, while additional analyses and case studies illustrate the value of source traceability and multi-solver dispatch.

Problem

• Operations Research
• Large Language Models
• Decision Support
• Interpretability
• Solver Integration

Innovation

• modular LLM agent
• confidence-based modeling
• element-level provenance
• multi-solver routing
• interpretable optimization