MIPLIB 2024: Call for Submissions

Do you have challenging mixed-integer programming instances and want to contribute to the future of optimization research? If so, please consider submitting your instances to MIPLIB, which is now accepting submissions through the end of November.

What Is MIPLIB? 

The Mixed-Integer Programming Library (MIPLIB) is a publicly available collection of mixed-integer programming model instances. The first version was released in 1992 by Gurobi co-founder Robert E. Bixby and co-authors E.A. Boyd and R.R. Indovina.   

MIPLIB provides an important testbed of real-world MIP instances for hundreds of academic researchers and commercial developers, including Gurobi. This collection of instances has played a significant role in advancing the field as a whole, as many innovative algorithmic ideas have been discovered by researchers from academia and industry while investigating MIPLIB problem instances. 

Over time, advances in the field have created the need for larger and more challenging test sets; updated versions of MIPLIB were released in 1996, 2003, 2010, and 2017.

The most recent iteration, MIPLIB 2017, was compiled by a group of academic researchers and commercial software developers (representing nine solvers, including Gurobi). In the years since this version, new applications of mixed-integer programming have arisen, and the number of instances solved in practice has increased alongside the power of state-of-the-art software.  

In keeping with the established seven-year cadence, instance collection is now underway for MIPLIB 2024. 

Why Should I Submit an Instance to MIPLIB? 

To develop the next generation of optimization algorithms, it is necessary to have access to realistic and challenging problem instances. Gurobi’s own R&D efforts are guided by a large internal collection of customer models; however, since these instances are treated with confidentiality, they are not available to the wider research community.  

By submitting your challenging real-world instances to MIPLIB, you help to support innovation in the MIP research community. This helps the entire field move forward, and may lead to better solver performance for models from your application domain. This test set also provides a means to compare and benchmark different solvers and quantify progress over time. Representing the most industries and applications possible is critical for the continued relevance and success of MIPLIB.  

How Do I Submit an Instance? 

The instance submission form and additional details can be found here.

Any and all mixed-integer linear program instances are welcome, including pure binary and integer programs, as well as instances containing indicator and SOS constraints.  

MIPLIB 2017 had submissions derived from 66 unique submitters (including bulk submissions from online optimization servers and optimization competitions), and we hope to see even greater success for MIPLIB 2024.