Benchmark control problems in nonequilibrium statistical mechanics

TL;DR

This paper introduces NESTbench25, a set of computational benchmarks for testing optimization methods in nonequilibrium statistical mechanics, featuring simple yet challenging stochastic control problems with implementations in C++ and Python.

Contribution

It provides a standardized, accessible benchmark suite for evaluating optimization techniques in nonequilibrium systems, facilitating progress in the field.

Findings

Five benchmark problems included.

Implementations available in C++ and Python.

Designed to challenge modern optimization methods.

Abstract

We present a set of computer codes designed to test methods for optimizing time-dependent control protocols in fluctuating nonequilibrium systems. Each problem consists of a stochastic model, an optimization objective, and C++ and Python implementations that can be run on Unix-like systems. These benchmark systems are simple enough to run on a laptop, but challenging enough to test the capabilities of modern optimization methods. This release includes five problems and a worked example. The problem set is called NESTbench25, for NonEquilibrium STatistical mechanics benchmarks (2025).