LLMs for Constraint Modeling
IJCAI 2026 Tutorial
Date: Monday Morning, August 17, 2026
Location: GW2 B3009
IJCAI 2026 Tutorial
Date: Monday Morning, August 17, 2026
Location: GW2 B3009
Constraint Programming (CP) and Operations Research (OR) provide powerful paradigms for solving complex decision-making and optimization problems. However, their practical adoption is often limited by the difficulty of modeling: translating a real-world problem into a formal specification of variables, constraints, and objectives requires significant expertise. This modeling step is widely recognized as a bottleneck, even with the availability of high-level modeling languages and solver technologies. Recent advances in LLMs open up a new direction toward alleviating this barrier by acting as modeling assistants that can transform natural‑language descriptions into formal models. Over the past three years, LLMs have demonstrated substantial improvements in structured reasoning, code generation, and multi‑step problem decomposition. At the same time, new methodologies and benchmarks such as Nl4Opt, ComplexOr, Dcp, IndustryOr, Optimus, Text2Zinc, Orlm, OptiChat, and Cp‑Bench now enable systematic evaluation of LLM‑based modeling pipelines. In this tutorial, we bring these developments together into a coherent overview of this rapidly emerging area.
This tutorial provides a structured overview of the field, covering both discrete and continuous optimization settings. We present different approaches for leveraging LLMs for model generation, including pipelines with intermediate representations such as entity extraction, knowledge‑graph compilation, and blueprint models; prompt engineering and retrieval‑augmented in‑context learning; self‑reflection; and agentic approaches. We also discuss fine‑tuning and domain‑adaptation strategies, along with practical demonstrations within existing modeling frameworks. A key component of the tutorial is evaluation: we examine methodologies and benchmarks for assessing LLM‑generated models with a focus on correctness, solution quality, and robustness. The tutorial highlights these challenges and provides guidance on responsible and effective use of LLM‑based modeling tools, offering participants both conceptual understanding and practical insights into the rapidly evolving intersection of LLMs and constraint modeling.
1. Ramamonjison et al. ‘NL4Opt Competition: Formulating Optimization Problems Based on Their Natural Language Descriptions’ (2022)
2. AhmadiTeshnizi et al. ‘OptiMUS: Scalable Optimization Modeling with (MI)LP Solvers and Large Language Models’ (2024)
3. Mostajabdaveh et al. ‘Optimization Modeling and Verification from Problem Specifications Using a Multi-Agent Multi-Stage LLM Framework’ (2024)
4. Xiao et al. ‘Chain-of-Experts: When LLMs Meet Complex Operations Research Problems’ (2024)
5. Huang et al. ‘LLMs for Mathematical Modeling: Towards Bridging the Gap Between Natural and Mathematical Languages’ (2025)
6. Huang et al. ‘ORLM: A Customizable Framework in Training Large Models for Automated Optimization Modeling’ (2025)
7. Yang et al. ‘OptiBench Meets ReSocratic: Measure and Improve LLMs for Optimization Modeling’ (2025)
8. Zhou et al. ‘Auto-Formulating Dynamic Programming Problems with Large Language Models’ (2025)
9. Michailidis et al. ‘CP-Bench: Evaluating Large Language Models for Constraint Modelling’ (2025)
10. Singirikonda et al. ‘Text2Zinc: A Cross-Domain Dataset for Modeling Optimization and Satisfaction Problems in MiniZinc’ (2025)
11. Wang et al. ‘ORGEval: Graph-Theoretic Evaluation of LLMs in Optimization Modeling’ (2025)
12. Lu et al. ‘OptMATH: A Scalable Bidirectional Data Synthesis Framework for LLM-Based Optimization Modeling’ (2025)
13. Michailidis, Tsouros, and Guns. ‘DCP-Bench-Open: Evaluating LLMs for Constraint Modelling of Discrete Combinatorial Problems’ (2026)
14. Song and Cohen. ‘Do LLMs Understand Constraint Programming? Zero-Shot Constraint Programming Model Generation Using LLMs’ (2025)
15. Wang et al. ‘Formalize, Don’t Optimize: The Heuristic Trap in LLM-Generated Combinatorial Solvers’ (2026)
Also used information from:
Xiao et al., “A Survey of Optimization Modeling Meets LLMs: Progress and Future Directions”, IJCAI 2025
Xiu, Li, Fan, and Liu, “Large Language Models for Operations Research: A Comprehensive Survey,” arXiv:2605.20849 (2026),
Wasserkrug et al. Decision Optimization CoPilot Manifesto, 2024
Lawless et. al. "It Was a Magical Box": Understanding Practitioner Workflows and Needs in Optimization, 2025
Simchi-Levi et al. Democratizing Optimization with Generative AI, 2025
Tsouros et al. Holy Grail 2.0: From Natural Language to Constraint Models 2023
Ramamonjison et al. LaTeX2Solver: Hierarchical Document Parsing, ACL’23
Kadioglu et. al. Ner4Opt: Named Entity Recognition for Optimization, Constraints, 2024
Michailidis et. al., Constraint Modelling with LLMs Using In-Context Learning, CP'24
Kadioglu et. al., Text2Model: Modeling Copilots for Text-to-Model Translation, 2025
Huang et. al. ORLM: A Customizable Framework in Training Large Models for Automated Optimization Modeling, 2024
Lima et. al. Toward a Trustworthy Optimization Modeling Agent via Verifiable Synthetic Data Generation, 2025
Kadioglu et. al. Learn2Zinc: Fine-tuning Small Language Models for Text-to-Model Translation in MiniZinc, 2026
Zhang et al. OptiMind: Teaching LLMs to Think Like Optimization Experts, 2026
Xiao et. al. Chain-of-Experts: When LLMs Meet Complex Operations Research Problems, 2024
Liang et. al. Large-Scale Optimization Model Auto-Formulation: Harnessing LLM Flexibility via Structured Workflow, 2026
AhmadiTeshnizi et. al. OptiMUS-0.3: Using Large Language Models to Model and Solve Optimization Problems at Scale, 2026
Cai et. al. Gala: Global LLM Agents for Text-to-Model Translation, 2025
Astorga et. al. Autoformulation of Mathematical Optimization Models Using LLMs, 2024
Jiang et. al. LLMOPT: Learning to Define and Solve General Optimization Problems from Scratch, 2025
Rossi et. al. Grammar-Aware Literate Generative Mathematical Programming with Compiler-in-the-Loop, 2026
Lawless et. al., "I Want It That Way": Enabling Interactive Decision Support Using Large Language Models and Constraint Programming, 2024
Chen et. al. OptiChat: Bridging Optimization Models and Practitioners with Large Language Models, 2025
This is a vibrant area of research so please let us know if we missed anything .