One Hundred Real-World Industrial Optimization Problems

IndustryOR provides 100 real-world operations research problems across five modeling types: linear programming, integer programming, mixed integer programming, non-linear programming, and others. CardinalOperations created this benchmark to train large language models for optimization modeling, with the dataset last updated in October 2025.

Use Cases

• Training language models to generate linear programming formulations from textual problem descriptions.
• Benchmarking AI systems on integer programming problems across three difficulty levels.
• Evaluating model performance on mixed integer programming and non-linear programming tasks.
• Fine-tuning models to classify optimization problems by type (e.g., LP, IP, MIP, NLP) based on their textual structure.

Strengths

• Contains 100 distinct real-world optimization problems.
• Covers 5 specific problem types: LP, IP, MIP, NLP, and others.
• Problems are categorized into 3 levels of difficulty.

Limitations

• Specific row count, column names, and sample data are unavailable.
• The dataset's geographic and temporal coverage for the source problems is unspecified.
• License terms are not explicitly stated in the provided description.

Provenance

Source

CardinalOperations

Collection Method

Curated as a benchmark from real-world operations research problems.

Freshness

Last updated on 2025-10-11.