Artifact for VeriQR: A Robustness Verification Tool for Quantum Machine Learning Models

We provide the artifact for the tool VeriQR, a Qt application implemented in C++ for robustness verification of quantum machine learning (QML) models, presented in our accepted paper. This artifact consists of the software, data sets, models, test suites, and benchmarks, expanding upon the content of the paper to facilitate the experimental evaluation discussed in our study. VeriQR is a graphical user interface (GUI) tool for the robustness verification of QML models in the current NISQ era, where noise is unavoidable. VeriQR offers exact, under-approximate, and tensor network-based algorithms for local and global robustness verification of real-world QML models in the presence of quantum noise. Throughout the verification process, VeriQR can identify quantum adversarial examples (states), which can be utilized by users for adversarial training to improve the local robustness as the same as classical machine learning techniques. Additionally, users can use VeriQR to apply specific quantum noise to enhance global robustness. Furthermore, VeriQR is capable of accommodating any quantum model in the OpenQASM 2.0 format and can convert QML models into this format to establish a unified benchmark framework for robustness verification.