Skip to content CVSP — Paper guide | Fabio Furini Skip to content

PAPER GUIDE · MATHEMATICAL PROGRAMMING COMPUTATION · 2019

How do you build a benchmark for quadratic optimization?

QPLIB: from thousands of heterogeneous candidate instances to a structured, challenging and reproducible quadratic programming library.

F. Furini · E. Traversi · P. Belotti · A. Frangioni · A. Gleixner · N. Gould · L. Liberti · A. Lodi · R. Misener · H. Mittelmann · N. Sahinidis · S. Vigerske and A. Wiegele

Candidate instances8,000+multiple source librariesClassificationQP taxonomychallenge filteringQPLIBreusable benchmark
Conceptual illustration of the paper's methodology; drawn for this page, not reproduced from the published article.
THE CENTRAL INSIGHT

A benchmark must do more than collect files. QPLIB defines a taxonomy, a selection principle based on computational challenge, and a common data format for very different kinds of quadratic programs.

01 / RESEARCH QUESTION

What makes this problem challenging?

Quadratic programming includes continuous, mixed-integer, convex and nonconvex models with quadratic objectives and/or constraints. Existing benchmarks covered parts of this landscape, making comparisons across algorithmic communities difficult.

02 / THE IDEA

From problem structure to an algorithm

01Gather

Collect candidate models from community submissions and existing optimization libraries.

02Classify

Use a structured taxonomy distinguishing the mathematical properties of instances.

03Select and publish

Retain a manageable and varied set of challenging examples in a documented exchange format.

03 / ORIGINAL CONTRIBUTIONS

What does this work add?

A broad QP taxonomy

A consistent classification scheme that makes very different quadratic programs comparable.

Benchmark-selection methodology

A principled way to screen out broadly easy models without discarding instances that remain difficult for many solver families.

Reusable dataset and format

A documented library of optimization instances with a common file representation.

Community resource

A cross-disciplinary foundation for future solver benchmarking rather than a claim that one algorithm wins.

04 / REPORTED FINDINGS

What does the study demonstrate?

The authors report collecting more than 8,000 candidate instances before the filtering process. Their stated objective is to retain a diverse and challenging library; the article explicitly does not aim to rank quadratic-programming solvers.

Based on the original QPLIB manuscript, including its introduction, benchmark-construction section, and file-format appendix. No unsupported numerical comparison has been added.

05 / READ MORE

The details are worth exploring.

For precise formulations, theorem statements, original figures and numerical tables, consult the published article.

06 / SCIENTIFIC IMPACT

How QPLIB has enabled later research

Selected articles in peer-reviewed research journals. QPLIB is used primarily as a public benchmark, not as an algorithmic component. Each entry explains the concrete experimental use.

Mathematical Programming (2025)

A Practical and Optimal First-Order Method for Large-Scale Convex Quadratic Programming ↗

Uses 33 convex QPLIB problems alongside classical QP benchmarks to evaluate a proposed first-order method.

INFORMS Journal on Computing (2026)

Local Search for Integer Quadratic Programming ↗

Tests a new local-search framework on QPLIB and MINLPLIB, reporting new best-known solutions on open benchmark instances.

Mathematical Programming Computation (2026)

A Frank-Wolfe-based primal heuristic for quadratic mixed-integer optimization ↗

Uses QPLIB as a comparative testbed and reports eight new best-known QPLIB solutions.

Computational Optimization and Applications (2024)

Enhancements of discretization approaches for non-convex mixed-integer quadratically constrained quadratic programming: Part I ↗

Includes QPLIB instances in a 60-instance collection to evaluate strengthened discretization methods for nonconvex MIQCQP.

Computational Optimization and Applications (2024)

Enhancements of discretization approaches for non-convex mixed-integer quadratically constrained quadratic programming: Part II ↗

Continues the computational study of discretization techniques, again using QPLIB instances within its benchmark set.

4OR (2023)

Inductive linearization for binary quadratic programs with linear constraints: a computational study ↗

Benchmarks inductive linearization on quadratic combinatorial optimization problems, including QPLIB and MINLPLIB instances.

Selection checked 9 October 2026. No citation counts are displayed.