Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2610.22367

Discovering state-of-the-art quantum circuits with large language models

Andi Gu, Alek H. Kemeny, Susanne F. Yelin, Mikhail D. Lukin

quant-phcs.AISubmitted Oct 15, 2026• Announced Oct 16, 2026• 10 citations
arXivHTMLPDF

Abstract

Significant efforts are currently being directed toward the implementation of large-scale quantum information processors, but realizing useful quantum advantage also requires the development of efficient quantum algorithms. While for many practical problems the dominant resource costs are associated with a small set of computational primitives, optimizing them in the quantum setting presents unique challenges, including a reversibility constraint absent from ordinary classical computation. Here we introduce an evolutionary framework in which large language models (LLMs) propose circuit designs and an independent evaluator verifies and scores them. We apply this framework to tasks spanning numerical computation, chemistry, optimization, quantum simulation, and number theory. In seven of the eight tasks, the discovered circuits significantly reduce both Toffoli and qubit counts relative to the state of the art. In the eighth, the search discovers a new reversible architecture for prime-field modular inversion. By handling reversibility differently, the architecture makes divsteps -- a classical algorithm previously impractical for reversible integer arithmetic -- viable and substantially improves the Toffoli-qubit trade-off for modular inversion. These results open the door for evaluator-guided LLM co-design of new algorithmic structures for quantum computation.

AI Summary

The paper asks whether evaluator-guided LLM search can redesign reversible computational primitives beyond local gate rewrites. It searches over Python circuit-generating programs, evolving multiple algorithm families in parallel and minimizing $TQ$, the product of Toffoli count and peak logical-qubit count. An independent evaluator simulates computational-basis trajectories, checking outputs, phases, and ancilla cleanup, including measurement-based uncomputation. Across eight benchmarks, seven circuits reduce both reported resources relative to their baselines. Examples include Reed–Solomon DQI decoding, from $5.1\times10^6$ Toffolis and 1,938 qubits to $4.4\times10^5$ and 1,837, and Gaussian evaluation, with a $39\times$ reduction in $TQ$. The Hermitian-decoding comparison uses a naïve reversible baseline rather than a prior compiled circuit. Recurring constructions share arithmetic through conditional routing, recover branch history from arithmetic invariants, and omit resources needed only by rare inputs. A prime-field divsteps architecture introduces branch decoders that trade computation against history storage. Its 256-bit elliptic-curve point-addition benchmark uses $4.0\times10^6$ Toffolis and 935 qubits, versus a prior construction’s $1.9\times10^6$ and 1,446; its stated failure union bound is $1.599\times10^{-4}$. Correctness combines algebraic arguments with sampled circuit tests, not exhaustive verification. For a fixed candidate, trajectory failure probability $\varepsilon$ implies coherent-state error at most $2\sqrt{\varepsilon}$ under the tested input distribution. Limitations include distribution-dependent failure budgets, fixed task interfaces and encodings, and a resource objective that omits hardware-specific depth, routing, and execution costs.

No sign of AI-generated textAI use disclosedStrong assumptions
Starting from a blank program, it chooses the algorithm, representation, and circuit structure and produces the discovered circuits without substantive human intervention.

Error found: The abstract claims seven improvements over the state of the art, although the Hermitian-decoding comparison uses a newly constructed naïve reversible baseline rather than a prior reversible circuit. The abstract omits that some headline decoding and inversion savings rely on allowing rare failures under specified input distributions, rather than preserving exact correctness for every input.

Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.

Comments
46 pages, 10 figures, 6 tables

Related papers

Innovative design and optimization of arithmetic circuits in quantum computing

Siyi Wang

In recent decades, quantum computing has witnessed remarkable progress, achieved through various advancements across the technology stack, such as quantum error-correcting codes and quantum algorithms. It is noteworthy that several promising algorithms incorporate quantum arithmetic circuits as fundamental building blocks. However, the inherent properties of quantum circuits challenge the efficiency of quantum arithmetic circuits, prompting creative solutions. While several quantum arithmetic circuits have been proposed, there is still a substantial gap in quantum arithmetic circuits optimization. For example, methods for integrating classical techniques such as high radix strategy into quantum designs still remain underexplored, highlighting a significant space for further advancement. This thesis aims to improve the performance of quantum arithmetic designs by integrating classical arithmetic circuit design techniques. Specifically, this thesis contains the following research: (1) The utilization of classical higher radix strategy to enhance quantum addition, reducing overall cost by lowering T-depth and T-count. The proposed design outperforms existing quantum carry-lookahead adders. (2) The application of Ling structure, achieving a reduced Toffoli-depth of 4log(n/2). (3) The exploration of the Kogge-Stone architecture, resulting in the lowest Toffoli-depth of log(n) for quantum adders. (4) The development of an efficient modular addition framework using carry-save architecture, demonstrating a 33.3% improvement in Toffoli depth compared to existing designs. (5) The design of an optimized quantum non-restoring divider framework. (6) A comprehensive overview of current advancements in quantum arithmetic designs, offering insights into potential research directions. These proposed designs are mapped to real-world quantum hardware architectures, including Noisy Intermediate-Scale Quantum systems, to validate their functionality and evaluate efficiency. Their applications in quantum machine learning and cryptography further underscore the broader impact of these improvements. Overall, this thesis makes significant advancements in reducing quantum circuit depth and improving resource efficiency, contributing to the development of scalable and efficient large-scale quantum circuits. By adapting classical design strategies to quantum circuits, it not only enhances various quantum circuit designs but also fosters further exploration at the intersection of classical and quantum computing.

•Jan 1, 2025•0 citations

Scalable quantum state preparation via large-language-model-driven discovery

Qing-Hong Cao, Zong-Yue Hou, Yingying Li +5 more

Efficient quantum state preparation remains a central challenge in first-principles quantum simulations of dynamics in quantum field theories, where the Hilbert space is intrinsically infinite-dimensional. Here, we introduce a large language model (LLM)-assisted framework for quantum-circuit design that systematically scales state-preparation circuits to large lattice volumes. Applied to a 1+1d XY spin chain, the LLM autonomously discovers a compact 4-parameter circuit that captures boundary-induced symmetry breaking with sub-percent energy deviation, enabling successful validation on the \texttt{Zuchongzhi} quantum processor. Guided by this insight, we extend the framework to 2+1d quantum field theories, where scalable variational ansätze have remained elusive. For a scalar field theory, the search yields a symmetry-preserving, 3-parameter shallow-depth ansatz whose optimized parameters converge to size-independent constants for lattices $n \ge 4$, providing, to our knowledge, the first scalable ansatz for this class of 2+1d models. Our results establish a practical route toward AI-assisted, human-guided discovery in quantum simulation.

arXiv:2505.06347•Physical review. D/Physical review. D.•Nov 5, 2026•11 citations

Bridging Classical and Quantum Computing for Next-Generation Language Models

Yi Pan, Hanqi Jiang, Junhao Chen +6 more

The remarkable success of Transformer architectures in Large Language Models (LLMs) has revolutionized natural language processing, yet the transition to quantum computing for next-generation language models remains an open challenge. While quantum computing promises exponential advantages, a fundamental gap exists between classical deep learning and quantum computing paradigms, particularly given the severe constraints of Noisy Intermediate-Scale Quantum (NISQ) devices, including barren plateaus, limited qubit coherence, and circuit depth restrictions. We present Adaptive Quantum-Classical Fusion (AQCF), the first framework to bridge classical and quantum computing for language models by reimagining Transformer architectures through quantum-classical co-design. Our key insight is that effective bridging requires dynamic adaptation rather than static translation—the framework analyzes input complexity in real-time to orchestrate seamless transitions between classical and quantum processing. AQCF introduces entropy-driven adaptive circuits that circumvent barren plateaus, quantum memory banks that unify classical attention with quantum state-based similarity retrieval, and intelligent fusion controllers that ensure each computational paradigm handles tasks where it naturally excels. This bridging architecture maintains full compatibility with existing classical Transformers while progressively incorporating quantum advantages as they become accessible. Experiments on sentiment analysis demonstrate that AQCF achieves competitive performance while significantly improving quantum resource efficiency, operating successfully within typical NISQ constraints. By establishing a seamless integration pathway from today's classical LLMs to tomorrow's quantum-enhanced models, our framework provides both immediate practical value on current quantum hardware and a clear evolution path toward full Quantum LLMs as technology matures.

arXiv:2508.07026•Proceedings of the AAAI Symposium Series•Nov 23, 2025•7 citations

Quantum Circuit Generation via test-time learning with large language models

Adriano Macarone-Palmieri, Rosario Lo Franco

Deploying large language models (LLMs) as optimizers for black-box scientific design problems requires efficient test-time refinement under expensive evaluations and without training data. We propose a \emph{memory-augmented test-time optimization} framework that combines episodic memory of high-scoring candidates, score-difference feedback, and restart-from-best sampling to improve iterative search. We evaluate the approach on quantum circuit synthesis, where the objective is to maximize the Meyer--Wallach (MW) global entanglement measure under an exponentially expensive black-box oracle. On 20-qubit circuits, the framework achieves $Q(ψ)=0.99$ without feedback. On the more challenging 25-qubit task, feedback and restart mechanisms enable multiple runs to reach $Q(ψ)=1.0$ within 45 oracle calls, while a budget-matched random hill-climbing baseline stalls below $Q(ψ)\approx0.29$. These results demonstrate that memory and evaluator feedback substantially improve the sample efficiency of LLM-based black-box optimization and establish quantum circuit synthesis as a challenging benchmark for test-time optimization.

quant-pharXiv:2602.03466•Feb 4, 2026•5 citations

An LLM System for Autonomous Variational Quantum Circuit Design

Kenya Sakka, Wataru Mizukami, Kosuke Mitarai

The design of high performing quantum circuits remains largely dependent on human expertise. We introduce an autonomous agentic framework that employs large language models (LLMs) to conduct iterative quantum circuit designs under explicit design constraints. Our system integrates seven components: Exploration, Generation, Discussion, Validation, Storage, Evaluation, and Review. These components form a closed-loop workflow that combines web-based knowledge acquisition, literature-grounded critique, executable code generation, and experimental feedback. We evaluate the framework on two tasks: quantum feature map construction for quantum machine learning and ansatz generation for variational quantum eigensolver applications in quantum chemistry. In image classification benchmarks, the best generated feature map outperforms representative quantum feature maps and, when scaled to larger qubit counts, surpasses the classical radial basis function kernel. In molecular ground state estimation across seven molecules, the generated ansatz attains competitive accuracy with widely used chemically inspired and hardware-efficient constructions while satisfying the imposed scaling constraints. These results establish LLM driven agentic system as a viable paradigm for automated quantum circuit design and illustrate how AI systems can participate in iterative scientific optimization workflows across scientific domains.

quant-pharXiv:2606.13380•Jun 12, 2026•6 citations

Resource-Efficient Modular Quantum Arithmetic Logic Unit with Clifford+T Optimization for Quantum Nanoelectronic Systems

P. Sowmithri, N. Roshitha, K. Santhi Priya, M. Manoj Kumar, K. Harika and M. Prakash

ABSTRACT Quantum arithmetic is a basic computational layer for reversible processing in quantum processors, yet its implementation cost grows rapidly with operand width and multi-qubit gate requirements. This paper presents a modular quantum arithmetic and logic unit (QALU) designed with resource-aware reversible primitives and Clifford+T decomposition. The architecture separates addition, subtraction, multiplication, and division into reusable submodules and evaluates their circuit cost in terms of T-count, T-depth, CNOT-count, CNOT-depth, qubit width, and total gate count. Hermitian LI gates and approximate Toffoli decompositions are incorporated to reduce redundant non-Clifford operations and circuit depth. The circuits are constructed and validated in Qiskit for small operand sizes, followed by symbolic resource analysis for scalable forms. For the demonstrated instances, the optimized adder is reduced from 10 to 7 total gates, the subtractor from 9 to 5, the multiplier from 12 to 8, and the divider from 11 to 6, while preserving the intended arithmetic mappings. The resulting modular framework provides a compact circuit-level approach for quantum arithmetic and is relevant to emerging quantum technology and quantum nanoelectronic systems where qubit count and fault-tolerant gate overhead are critical design constraints. Key words: quantum arithmetic, quantum ALU, reversible logic, Clifford+T, Hermitian LI gate, approximate Toffoli, Qiskit, quantum nanoelectronics.

International Journal of Emerging Trends in Engineering and Development•Nov 6, 2026•0 citations
Significant efforts are currently being directed toward the implementation of large-scale quantum information processors, but realizing useful quantum advantage also requires the development of efficient quantum algorithms. While for many practical problems the dominant resource costs are associated with a small set of computational primitives, optimizing them in the quantum setting presents unique challenges, including a reversibility constraint absent from ordinary classical computation. Here we introduce an evolutionary framework in which large language models (LLMs) propose circuit designs and an independent evaluator verifies and scores them. We apply this framework to tasks spanning numerical computation, chemistry, optimization, quantum simulation, and number theory. In seven of the eight tasks, the discovered circuits significantly reduce both Toffoli and qubit counts relative to the state of the art. In the eighth, the search discovers a new reversible architecture for prime-field modular inversion. By handling reversibility differently, the architecture makes divsteps -- a classical algorithm previously impractical for reversible integer arithmetic -- viable and substantially improves the Toffoli-qubit trade-off for modular inversion. These results open the door for evaluator-guided LLM co-design of new algorithmic structures for quantum computation.
The paper asks whether evaluator-guided LLM search can redesign reversible computational primitives beyond local gate rewrites. It searches over Python circuit-generating programs, evolving multiple algorithm families in parallel and minimizing TQTQTQ, the product of Toffoli count and peak logical-qubit count. An independent evaluator simulates computational-basis trajectories, checking outputs, phases, and ancilla cleanup, including measurement-based uncomputation.
Across eight benchmarks, seven circuits reduce both reported resources relative to their baselines. Examples include Reed–Solomon DQI decoding, from 5.1×1065.1\times10^65.1×106 Toffolis and 1,938 qubits to 4.4×1054.4\times10^54.4×105 and 1,837, and Gaussian evaluation, with a 39×39\times39× reduction in TQTQTQ. The Hermitian-decoding comparison uses a naïve reversible baseline rather than a prior compiled circuit. Recurring constructions share arithmetic through conditional routing, recover branch history from arithmetic invariants, and omit resources needed only by rare inputs.
A prime-field divsteps architecture introduces branch decoders that trade computation against history storage. Its 256-bit elliptic-curve point-addition benchmark uses 4.0×1064.0\times10^64.0×106 Toffolis and 935 qubits, versus a prior construction’s 1.9×1061.9\times10^61.9×106 and 1,446; its stated failure union bound is 1.599×10−41.599\times10^{-4}1.599×10−4.
Correctness combines algebraic arguments with sampled circuit tests, not exhaustive verification. For a fixed candidate, trajectory failure probability ε\varepsilonε implies coherent-state error at most 2ε2\sqrt{\varepsilon}2ε​ under the tested input distribution. Limitations include distribution-dependent failure budgets, fixed task interfaces and encodings, and a resource objective that omits hardware-specific depth, routing, and execution costs.
In recent decades, quantum computing has witnessed remarkable progress, achieved through various advancements across the technology stack, such as quantum error-correcting codes and quantum algorithms. It is noteworthy that several promising algorithms incorporate quantum arithmetic circuits as fundamental building blocks. However, the inherent properties of quantum circuits challenge the efficiency of quantum arithmetic circuits, prompting creative solutions. While several quantum arithmetic circuits have been proposed, there is still a substantial gap in quantum arithmetic circuits optimization. For example, methods for integrating classical techniques such as high radix strategy into quantum designs still remain underexplored, highlighting a significant space for further advancement. This thesis aims to improve the performance of quantum arithmetic designs by integrating classical arithmetic circuit design techniques. Specifically, this thesis contains the following research: (1) The utilization of classical higher radix strategy to enhance quantum addition, reducing overall cost by lowering T-depth and T-count. The proposed design outperforms existing quantum carry-lookahead adders. (2) The application of Ling structure, achieving a reduced Toffoli-depth of 4log(n/2). (3) The exploration of the Kogge-Stone architecture, resulting in the lowest Toffoli-depth of log(n) for quantum adders. (4) The development of an efficient modular addition framework using carry-save architecture, demonstrating a 33.3% improvement in Toffoli depth compared to existing designs. (5) The design of an optimized quantum non-restoring divider framework. (6) A comprehensive overview of current advancements in quantum arithmetic designs, offering insights into potential research directions. These proposed designs are mapped to real-world quantum hardware architectures, including Noisy Intermediate-Scale Quantum systems, to validate their functionality and evaluate efficiency. Their applications in quantum machine learning and cryptography further underscore the broader impact of these improvements. Overall, this thesis makes significant advancements in reducing quantum circuit depth and improving resource efficiency, contributing to the development of scalable and efficient large-scale quantum circuits. By adapting classical design strategies to quantum circuits, it not only enhances various quantum circuit designs but also fosters further exploration at the intersection of classical and quantum computing.
Efficient quantum state preparation remains a central challenge in first-principles quantum simulations of dynamics in quantum field theories, where the Hilbert space is intrinsically infinite-dimensional. Here, we introduce a large language model (LLM)-assisted framework for quantum-circuit design that systematically scales state-preparation circuits to large lattice volumes. Applied to a 1+1d XY spin chain, the LLM autonomously discovers a compact 4-parameter circuit that captures boundary-induced symmetry breaking with sub-percent energy deviation, enabling successful validation on the \texttt{Zuchongzhi} quantum processor. Guided by this insight, we extend the framework to 2+1d quantum field theories, where scalable variational ansätze have remained elusive. For a scalar field theory, the search yields a symmetry-preserving, 3-parameter shallow-depth ansatz whose optimized parameters converge to size-independent constants for lattices n≥4n \ge 4n≥4, providing, to our knowledge, the first scalable ansatz for this class of 2+1d models. Our results establish a practical route toward AI-assisted, human-guided discovery in quantum simulation.
The remarkable success of Transformer architectures in Large Language Models (LLMs) has revolutionized natural language processing, yet the transition to quantum computing for next-generation language models remains an open challenge. While quantum computing promises exponential advantages, a fundamental gap exists between classical deep learning and quantum computing paradigms, particularly given the severe constraints of Noisy Intermediate-Scale Quantum (NISQ) devices, including barren plateaus, limited qubit coherence, and circuit depth restrictions. We present Adaptive Quantum-Classical Fusion (AQCF), the first framework to bridge classical and quantum computing for language models by reimagining Transformer architectures through quantum-classical co-design. Our key insight is that effective bridging requires dynamic adaptation rather than static translation—the framework analyzes input complexity in real-time to orchestrate seamless transitions between classical and quantum processing. AQCF introduces entropy-driven adaptive circuits that circumvent barren plateaus, quantum memory banks that unify classical attention with quantum state-based similarity retrieval, and intelligent fusion controllers that ensure each computational paradigm handles tasks where it naturally excels. This bridging architecture maintains full compatibility with existing classical Transformers while progressively incorporating quantum advantages as they become accessible. Experiments on sentiment analysis demonstrate that AQCF achieves competitive performance while significantly improving quantum resource efficiency, operating successfully within typical NISQ constraints. By establishing a seamless integration pathway from today's classical LLMs to tomorrow's quantum-enhanced models, our framework provides both immediate practical value on current quantum hardware and a clear evolution path toward full Quantum LLMs as technology matures.
Deploying large language models (LLMs) as optimizers for black-box scientific design problems requires efficient test-time refinement under expensive evaluations and without training data. We propose a \emph{memory-augmented test-time optimization} framework that combines episodic memory of high-scoring candidates, score-difference feedback, and restart-from-best sampling to improve iterative search. We evaluate the approach on quantum circuit synthesis, where the objective is to maximize the Meyer--Wallach (MW) global entanglement measure under an exponentially expensive black-box oracle. On 20-qubit circuits, the framework achieves Q(ψ)=0.99Q(ψ)=0.99Q(ψ)=0.99 without feedback. On the more challenging 25-qubit task, feedback and restart mechanisms enable multiple runs to reach Q(ψ)=1.0Q(ψ)=1.0Q(ψ)=1.0 within 45 oracle calls, while a budget-matched random hill-climbing baseline stalls below Q(ψ)≈0.29Q(ψ)\approx0.29Q(ψ)≈0.29. These results demonstrate that memory and evaluator feedback substantially improve the sample efficiency of LLM-based black-box optimization and establish quantum circuit synthesis as a challenging benchmark for test-time optimization.
The design of high performing quantum circuits remains largely dependent on human expertise. We introduce an autonomous agentic framework that employs large language models (LLMs) to conduct iterative quantum circuit designs under explicit design constraints. Our system integrates seven components: Exploration, Generation, Discussion, Validation, Storage, Evaluation, and Review. These components form a closed-loop workflow that combines web-based knowledge acquisition, literature-grounded critique, executable code generation, and experimental feedback. We evaluate the framework on two tasks: quantum feature map construction for quantum machine learning and ansatz generation for variational quantum eigensolver applications in quantum chemistry. In image classification benchmarks, the best generated feature map outperforms representative quantum feature maps and, when scaled to larger qubit counts, surpasses the classical radial basis function kernel. In molecular ground state estimation across seven molecules, the generated ansatz attains competitive accuracy with widely used chemically inspired and hardware-efficient constructions while satisfying the imposed scaling constraints. These results establish LLM driven agentic system as a viable paradigm for automated quantum circuit design and illustrate how AI systems can participate in iterative scientific optimization workflows across scientific domains.
ABSTRACT Quantum arithmetic is a basic computational layer for reversible processing in quantum processors, yet its implementation cost grows rapidly with operand width and multi-qubit gate requirements. This paper presents a modular quantum arithmetic and logic unit (QALU) designed with resource-aware reversible primitives and Clifford+T decomposition. The architecture separates addition, subtraction, multiplication, and division into reusable submodules and evaluates their circuit cost in terms of T-count, T-depth, CNOT-count, CNOT-depth, qubit width, and total gate count. Hermitian LI gates and approximate Toffoli decompositions are incorporated to reduce redundant non-Clifford operations and circuit depth. The circuits are constructed and validated in Qiskit for small operand sizes, followed by symbolic resource analysis for scalable forms. For the demonstrated instances, the optimized adder is reduced from 10 to 7 total gates, the subtractor from 9 to 5, the multiplier from 12 to 8, and the divider from 11 to 6, while preserving the intended arithmetic mappings. The resulting modular framework provides a compact circuit-level approach for quantum arithmetic and is relevant to emerging quantum technology and quantum nanoelectronic systems where qubit count and fault-tolerant gate overhead are critical design constraints. Key words: quantum arithmetic, quantum ALU, reversible logic, Clifford+T, Hermitian LI gate, approximate Toffoli, Qiskit, quantum nanoelectronics.