Towards Scaling Quantum Fine-Tuning of Foundational Time Series Models for Classification
研究通过在时间序列模型Chronos上使用量子头进行微调，以解决电力网格事件分类问题，并引入wing模块来增加数据带宽，从而提高准确性。
0 citationsRead paper
Selected work
Representative Papers
Breakeven demonstration of quantum low-density parity-check codes
This work addresses the challenge of deploying high-rate quantum error-correcting codes, which are often hindered by hardware constraints such as long-range couplings. On a single trapped-ion quantum computer and without any hardware reconfiguration, the authors demonstrate, for the first time, flexible implementation of nine distinct error-correcting codes—spanning qLDPC, topological, and concatenated families—with markedly different connectivity requirements. Leveraging an optical–metastable–ground (OMG) architecture, the system enables addressable mid-circuit measurement and reset without requiring ion shuttling or dedicated coolant ions. Notably, a qLDPC code encoding four logical qubits into eighteen physical qubits achieves break-even performance, exhibiting a logical error rate nine times lower than comparable superconducting-platform experiments; moreover, certain logical qubits surpass the coherence time of their constituent physical qubits, substantially enhancing resource efficiency.
0 citationsRead paper
Quantum Parity Representations: Learnable Basis Discovery, Encoders, and Shadow Deployment
This work addresses the challenge of modeling high-order feature interactions under binarized or quantized inputs by proposing a classically efficient inference method that requires no quantum resources. Leveraging quantum-inspired ideas during training—through learnable Pauli word selection, projection-based encoding, and an sPQC-Parity architecture—the approach constructs parity representations that rely solely on classical computation at inference time. On native binary tasks with 5–10 bits, the method achieves accuracy improvements of 23.9%–41.7% over logistic regression and SVM, and significantly outperforms baselines such as PCA-bin on textual and discrete datasets. Notably, it even surpasses fully continuous models in certain scenarios, marking the first demonstration of a performance advantage for quantum-inspired parity representations within a classically efficient inference framework.
0 citationsRead paper
Measuring Accuracy and Energy-to-Solution of Quantum Fine-Tuning of Foundational AI Models
This work investigates the efficient evaluation of accuracy and energy efficiency for quantum-classical hybrid AI models on noisy, limited-scale quantum hardware. By directly measuring power consumption on the Forte Enterprise trapped-ion quantum processor, the study implements an end-to-end quantum fine-tuning pipeline and introduces “energy-to-solution” (ETS) as a core evaluation metric. Empirical results on real quantum hardware demonstrate, for the first time, that quantum fine-tuned models achieve approximately 24% lower classification error than the best classical baseline. Furthermore, QPU energy consumption scales nearly linearly with qubit count, whereas classical simulation exhibits exponential growth, leading to an ETS breakeven point at around 34 qubits. These findings establish ETS as a viable benchmark for scalable quantum AI applications.
0 citationsRead paper
Quantum Feature Selection with Higher-Order Binary Optimization on Trapped-Ion Hardware
This work addresses the limitations of traditional feature selection methods in modeling high-order variable dependencies and the restricted expressivity of existing quantum approaches, which are largely confined to quadratic optimization. The authors propose a quantum feature selection framework based on Higher-Order Unconstrained Binary Optimization (HUBO), introducing for the first time a high-order Hamiltonian that incorporates three-body interactions. Feature interactions up to third order are quantified using mutual information, and structured sparsity regularization is integrated to enhance interpretability. Implemented on the IonQ Forte trapped-ion quantum processor using a digital reverse annealing algorithm, the method yields compact yet highly discriminative feature subsets on the Gallstone and Spambase datasets, achieving classification performance that matches or surpasses classical baselines such as SelectKBest and PCA, thereby demonstrating the feasibility and advantage of high-order quantum feature selection.
0 citationsRead paper
Recent publications
Latest Papers
Towards Scaling Quantum Fine-Tuning of Foundational Time Series Models for Classification
研究通过在时间序列模型Chronos上使用量子头进行微调，以解决电力网格事件分类问题，并引入wing模块来增加数据带宽，从而提高准确性。
0 citationsRead paper
Breakeven demonstration of quantum low-density parity-check codes
This work addresses the challenge of deploying high-rate quantum error-correcting codes, which are often hindered by hardware constraints such as long-range couplings. On a single trapped-ion quantum computer and without any hardware reconfiguration, the authors demonstrate, for the first time, flexible implementation of nine distinct error-correcting codes—spanning qLDPC, topological, and concatenated families—with markedly different connectivity requirements. Leveraging an optical–metastable–ground (OMG) architecture, the system enables addressable mid-circuit measurement and reset without requiring ion shuttling or dedicated coolant ions. Notably, a qLDPC code encoding four logical qubits into eighteen physical qubits achieves break-even performance, exhibiting a logical error rate nine times lower than comparable superconducting-platform experiments; moreover, certain logical qubits surpass the coherence time of their constituent physical qubits, substantially enhancing resource efficiency.
0 citationsRead paper
Quantum Parity Representations: Learnable Basis Discovery, Encoders, and Shadow Deployment
This work addresses the challenge of modeling high-order feature interactions under binarized or quantized inputs by proposing a classically efficient inference method that requires no quantum resources. Leveraging quantum-inspired ideas during training—through learnable Pauli word selection, projection-based encoding, and an sPQC-Parity architecture—the approach constructs parity representations that rely solely on classical computation at inference time. On native binary tasks with 5–10 bits, the method achieves accuracy improvements of 23.9%–41.7% over logistic regression and SVM, and significantly outperforms baselines such as PCA-bin on textual and discrete datasets. Notably, it even surpasses fully continuous models in certain scenarios, marking the first demonstration of a performance advantage for quantum-inspired parity representations within a classically efficient inference framework.
0 citationsRead paper
Measuring Accuracy and Energy-to-Solution of Quantum Fine-Tuning of Foundational AI Models
This work investigates the efficient evaluation of accuracy and energy efficiency for quantum-classical hybrid AI models on noisy, limited-scale quantum hardware. By directly measuring power consumption on the Forte Enterprise trapped-ion quantum processor, the study implements an end-to-end quantum fine-tuning pipeline and introduces “energy-to-solution” (ETS) as a core evaluation metric. Empirical results on real quantum hardware demonstrate, for the first time, that quantum fine-tuned models achieve approximately 24% lower classification error than the best classical baseline. Furthermore, QPU energy consumption scales nearly linearly with qubit count, whereas classical simulation exhibits exponential growth, leading to an ETS breakeven point at around 34 qubits. These findings establish ETS as a viable benchmark for scalable quantum AI applications.
0 citationsRead paper
Quantum Feature Selection with Higher-Order Binary Optimization on Trapped-Ion Hardware
This work addresses the limitations of traditional feature selection methods in modeling high-order variable dependencies and the restricted expressivity of existing quantum approaches, which are largely confined to quadratic optimization. The authors propose a quantum feature selection framework based on Higher-Order Unconstrained Binary Optimization (HUBO), introducing for the first time a high-order Hamiltonian that incorporates three-body interactions. Feature interactions up to third order are quantified using mutual information, and structured sparsity regularization is integrated to enhance interpretability. Implemented on the IonQ Forte trapped-ion quantum processor using a digital reverse annealing algorithm, the method yields compact yet highly discriminative feature subsets on the Gallstone and Spambase datasets, achieving classification performance that matches or surpasses classical baselines such as SelectKBest and PCA, thereby demonstrating the feasibility and advantage of high-order quantum feature selection.
0 citationsRead paper