Q-Bridge: Code Translation for Quantum Machine Learning via LLMs
Organizations: Rochester Institute of Technology, USA · Kent State University, USA · Rutgers University, USA
Abstract
Large language models have recently shown potential in bridging the gap between classical machine learning and quantum machine learning. However, the lack of standardized, high-quality datasets and robust translation frameworks limits progress in this domain. We introduce Q-Bridge, an LLM-guided code translation framework that systematically converts CML implementations into executable QML variants. Our approach builds on a self-involving pipeline that iteratively expands a verified seed codebase into a large-scale dataset, CML-2-QML, integrating verifiable and unverifiable code pairs. The Q-Bridge model is fine-tuned using supervised LoRA adaptation for scalable and memory-efficient training, achieving faithful and interpretable quantum code generation across diverse architectures. Empirical analysis confirms the feasibility of direct CML-to-QML translation and reveals consistent structural alignment between classical and quantum paradigms. Case studies further demonstrate that Q-Bridge can maintain deterministic correctness and also enable creative architectural exploration. This work establishes the first reproducible framework and dataset for LLM-driven quantum code translation, offering a foundation for scalable quantum AI development.
Figures & tables
| QML Model, Description, Applications, Ansatz, and Feature Map | ||||
| QML Model | Description | Applications | Ansatz | Feature Map |
|---|---|---|---|---|
| Continuous-Variable QNN ( Killoran et al., 2019 ) | Layered QNN using Gaussian (affine) + non-Gaussian (activation) gates | Fraud detection; autoencoding | Interferometers + squeezers + displacements + non-Gaussian gates | Displacement encoding of input data on vacuum modes |
| Quanvolutional NN ( Henderson et al., 2020 ) | Local quantum circuits on image patches to augment CNNs | Image recognition/classification | Random small quantum circuit per patch | Binary threshold encoding |
| Deep QNN ( Beer et al., 2020 ) | Parametrized unitaries with fidelity-based training; depth-robust | Learning an unknown unitary | Stacked “quantum perceptron” unitaries on qubits | Direct quantum input ( e . g .|0…0⟩); no separate classical feature map |
| Quantum Autoencoder ( Bondarenko and Feldmann, 2020 ) | Compress/denoise entangled states with an autoencoder circuit | Restore clean states; metrology | Feedforward QNN (quantum perceptron layers) | Quantum input states directly (no classical encoding) |
| Quantum GNN ( Beer et al., 2020 ) | QML for graph structured quantum data with self-supervised loss | Graph-data learning; noise-aware BO | Universal QNN ansatz (arbitrary PQC on all graph vertices) | Quantum input states at vertices (encoded by graph edges), no classical mapping |
| QML Model, Description, Applications, Ansatz, and Feature Map | ||||
| QML Model | Description | Applications | Ansatz | Feature Map |
|---|---|---|---|---|
| Quantum SVM (Kernel) ( Wang et al., 2022b ) | Compute quantum kernel via circuit; train classical SVM | Separability and classification | Kernel PQC ( e . g .ZZ-feature map circuit with entanglement) | Angle embedding of features ( e . g .Z-feature map) |
| Quanvolution (TorchQuantum) ( Wang et al., 2022b ) | Quantum circuits as convolution-like filters | Image recognition and classification | Fixed PQC filters (random or structured rotations+CX on patches) | Patch pixels encoded into qubit angles ( e . g .via RY gates) |
| QuantumNAT ( Wang et al., 2022a ) | Noise-aware PQC training (injection, normalization, quantization) | Classification on NISQ devices | Layered QNN (PQC classifier) with trainable weights | Classical inputs encoded via encoder layer (angle rotations) |
| QNN Regression ( Wang et al., 2022b ) | PQC for predicting continuous outputs | Regression and prediction | PQC similar to classification QNN (rotation+entangler layers) | Angle encoding of input features |
| QCNN ( Cong et al., 2019 ; Vatan and Williams, 2004 ) | Quantum convolution + pooling layers for patterns | Image classification | Parametrized two-qubit unitaries (convolution filters); pooling discards qubits | ZFeatureMap (phase rotation map on input pixels) |
Quantum Machine Learning
Quantum machine learning uses quantum circuits and quantum data to build learning algorithms and study their computational and statistical limits.
See all 256 papers