Quantum Large Language Model Fine-Tuning

2025-04-11

AI Summary

This work addresses the limited accuracy of large language models (LLMs) on complex semantic tasks—such as sentiment prediction—during fine-tuning. We propose the first scalable hybrid quantum-classical architecture, co-designing a sentence-level Transformer with a parameterized quantum circuit (PQC) featuring long-range entanglement, augmented by a quantum state reloading mechanism and a noise-robust training strategy. Our method pioneers systematic integration of quantum modules into LLM fine-tuning. We rigorously validate component contributions via finite-sampling noise simulation, hyperparameter sensitivity analysis, and ablation studies. Experiments demonstrate that, at comparable parameter counts, our architecture achieves up to a 3.14% absolute accuracy gain over classical baselines; moreover, increasing qubit count yields consistent performance improvements, confirming both the efficacy and scalability of quantum enhancement.

Technology Category

• Machine Learning: Quantum Machine Learning
• Natural Language Processing: (Large) Language Models
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

We introduce a hybrid quantum-classical deep learning architecture for large language model fine-tuning. The classical portion of the architecture is a sentence transformer that is powerful enough to display significant accuracy for complex tasks such as sentiment prediction. The quantum portion of the architecture consists of parameterized quantum circuits that utilize long-range connections between qubits. We analyze the performance of the hybrid models for various settings of hyperparameters, including the number of qubits, the depth of the quantum circuits, learning rate, number of re-uploading steps, etc. Based on a screening study of main effects, we show an overall improvement in prediction accuracy over a comparable classical baseline, with a trend of increasing accuracy with number of qubits. We observe up to $3.14%$ improvements in accuracy over classical architectures of comparable model size, within the set of hyperparameters probed in this study. We demonstrate the contribution of each module in our architecture through ablation studies. Our studies are based on finite shot-counts and include simulations based on noisy quantum gates.

Problem

• Hybrid quantum-classical architecture for fine-tuning large language models
• Improving prediction accuracy using parameterized quantum circuits
• Analyzing performance under various hyperparameters and noise conditions

Innovation

• Hybrid quantum-classical deep learning architecture
• Parameterized quantum circuits with long-range qubit connections
• Improved accuracy over classical baselines via quantum components