The Token-to-Qubit Semantic Translation Bottleneck: Bridging Large Language Models and Quantum Optimization Hamiltonians

Token generation in Large Language Models (LLMs) are designed on multi-head attention transformers. They capture long-range sequences inherently, which might span transformer full attention pass, and results in escalation of the model cost quadratically. This cost inflation trigger higher inference latency, limiting its deployment in resource-constrained scenarios. Quantum computing (QC) can address these limitations, owing to its advantages of entanglement in high-dimensional Hilbert spaces, and allow parallel sequence generation by LLMs. However, pure quantum transformers design are challenged by high noise fidelity, coherence, and lack of sufficient qubits for operation. Motivated by this, in this paper, we design a framework, QuantLLM, a hybrid classical-quantum transformer architecture that reduces latency via a token-wise, and semantically-aware adaptive routing mechanism. In the routing design, tokens are assigned to either classical or quantum attention paths computed by a differentiable complexity score, that minimizes the LLM inference latency. Results are evaluated on the benchmark WikiText-2 dataset. The framework reduces inference latency by a maximum of 30.06% compared to quantum-enhanced attention baselines. At a balanced setting, it achieves 84.1% accuracy with 290 ms latency, versus 85.7% with 418 ms against baseline scheme. For long sequences, up to 48.7% of tokens are routed through the quantum path, with token-level fidelity scores (0.872 for complex nouns and 0.122 for punctuation), which asserts optimized semantic-aware context routing. These results indicate the potential viability of the framework design in practical model settings.Token generation in Large Language Models (LLMs) are designed on multi-head attention transformers. They capture long-range sequences inherently, which might span transformer full attention pass, and results in escalation of the model cost quadratically. This cost inflation trigger higher inference latency, limiting its deployment in resource-constrained scenarios. Quantum computing (QC) can address these limitations, owing to its advantages of entanglement in high-dimensional Hilbert spaces, and allow parallel sequence generation by LLMs. However, pure quantum transformers design are challenged by high noise fidelity, coherence, and lack of sufficient qubits for operation. Motivated by this, in this paper, we design a framework, QuantLLM, a hybrid classical-quantum transformer architecture that reduces latency via a token-wise, and semantically-aware adaptive routing mechanism. In the routing design, tokens are assigned to either classical or quantum attention paths computed by a differentiable complexity score, that minimizes the LLM inference latency. Results are evaluated on the benchmark WikiText-2 dataset. The framework reduces inference latency by a maximum of 30.06% compared to quantum-enhanced attention baselines. At a balanced setting, it achieves 84.1% accuracy with 290 ms latency, versus 85.7% with 418 ms against baseline scheme. For long sequences, up to 48.7% of tokens are routed through the quantum path, with token-level fidelity scores (0.872 for complex nouns and 0.122 for punctuation), which asserts optimized semantic-aware context routing. These results indicate the potential viability of the framework design in practical model settings.

International Conference on Software, Telecommunications and Computer Networks (SoftCOM)Sep 18, 20253 citations

An adaptive contextual quantum language model

Jingfei Li, Peng Zhang, Dawei Song +1 more

Physica A Statistical Mechanics and its ApplicationsMar 28, 201623 citations

Hybrid Classical Quantum Embeddings for NLP and AI Using Hamiltonians

Damir Ćavar, James Bryan Graves, Shane A. Sparks +1 more

We present experimental results from three approaches to encoding classical text, word, or token embeddings, i.e., dense vector representations, as quantum states from state-of-the-art large language models. The experiments focus on identifying the best encoding strategies that achieve the highest compression of classical dense vectors in the number of required qubits, while maintaining the semantic properties of the embeddings trained and reflected in classical AI models. We identify the most optimal interface and data conversion approaches in hybrid classical and quantum computing systems. We demonstrate that the conversion of dense embedding vectors to Hamiltonians is a feasible strategy for a hybrid classicalquantum computing environment and storage of quantum states on classical computers. We report the experimental results of reducing the full embedding-based Hamiltonians using Principal Component Analysis (PCA).We present experimental results from three approaches to encoding classical text, word, or token embeddings, i.e., dense vector representations, as quantum states from state-of-the-art large language models. The experiments focus on identifying the best encoding strategies that achieve the highest compression of classical dense vectors in the number of required qubits, while maintaining the semantic properties of the embeddings trained and reflected in classical AI models. We identify the most optimal interface and data conversion approaches in hybrid classical and quantum computing systems. We demonstrate that the conversion of dense embedding vectors to Hamiltonians is a feasible strategy for a hybrid classicalquantum computing environment and storage of quantum states on classical computers. We report the experimental results of reducing the full embedding-based Hamiltonians using Principal Component Analysis (PCA).

IEEE International Conference on Quantum Computing and Engineering (QCE)Aug 30, 20250 citations

Trustworthy Natural Language Interfaces for Quantum Optimization: A Secure and Resilient Translation Framework

Kensay Sato, Mehtaab Singh, Rishabh Dhadda +4 more

As quantum computing emerges as a powerful tool for solving combinatorial optimization problems, the absence of intuitive programming interfaces remains a major barrier to practical adoption. This work presents a secure and resilient natural language interface that translates high-level optimization prompts into QUBO (Quadratic Unconstrained Binary Optimization) formulations suitable for quantum solvers such as D-Wave. The translation pipeline leverages large language models (LLMs) and incorporates multiple verification layers, including schema validation, constraint compliance checks, and reverse translation for semantic fidelity auditing. Fidelity is quantified using cosine similarity between sentence embeddings of the original and reconstructed problem descriptions, enabling intent alignment analysis. We evaluate the system across 100 optimization instances spanning five canonical problem types: Knapsack, Traveling Salesman Problem, Graph Coloring, Seating Assignment, and Team Assignment. Results show that GPT-4 significantly outperforms GPT-3.5 in complex cases such as TSP, achieving up to $90\%$ success rates where earlier models fail entirely. Latency measurements confirm interactive performance, with most translations completing in under 6 seconds. These results demonstrate the feasibility of trustworthy natural language interfaces for quantum optimization and provide a framework for verifiable, interpretable, and efficient translation from user intent to quantum-executable code.As quantum computing emerges as a powerful tool for solving combinatorial optimization problems, the absence of intuitive programming interfaces remains a major barrier to practical adoption. This work presents a secure and resilient natural language interface that translates high-level optimization prompts into QUBO (Quadratic Unconstrained Binary Optimization) formulations suitable for quantum solvers such as D-Wave. The translation pipeline leverages large language models (LLMs) and incorporates multiple verification layers, including schema validation, constraint compliance checks, and reverse translation for semantic fidelity auditing. Fidelity is quantified using cosine similarity between sentence embeddings of the original and reconstructed problem descriptions, enabling intent alignment analysis. We evaluate the system across 100 optimization instances spanning five canonical problem types: Knapsack, Traveling Salesman Problem, Graph Coloring, Seating Assignment, and Team Assignment. Results show that GPT-4 significantly outperforms GPT-3.5 in complex cases such as TSP, achieving up to$90\%$success rates where earlier models fail entirely. Latency measurements confirm interactive performance, with most translations completing in under 6 seconds. These results demonstrate the feasibility of trustworthy natural language interfaces for quantum optimization and provide a framework for verifiable, interpretable, and efficient translation from user intent to quantum-executable code.

IEEE International Conference on Quantum Computing and Engineering (QCE)Aug 30, 20250 citations

Compact Quantum Modules for Scalable Fine-Tuning in Large Language Models

Emily Jimin Roh, Joongheon Kim

Communications in computer and information scienceJan 1, 20260 citations

90\%

success rates where earlier models fail entirely. Latency measurements confirm interactive performance, with most translations completing in under 6 seconds. These results demonstrate the feasibility of trustworthy natural language interfaces for quantum optimization and provide a framework for verifiable, interpretable, and efficient translation from user intent to quantum-executable code.