Compiling quantum algorithms for real devices involves hard optimization problems that can decide whether an algorithm runs at scale. We present a general method for solving such problems with AI agents, with every result certified. It builds on autoresearch loops, in which agents iteratively improve programs against a fixed evaluator and which have recently improved quantum cryptanalysis circuits, and adds two elements: a frozen, exact verifier certifies every result, and this verifier, together with the challenge around it, is built by an agent that formulates the problem with a researcher and validates the verifier against a solved case, deliberately broken solutions and published results before any solution exists. The method applies to any optimization problem whose solutions can be verified exactly and efficiently. We apply it to three compilation problems in quantum simulation, namely fermion-to-qubit encoding, Hamiltonian layout and unitary synthesis. In each, the agents improved substantially on published or standard constructions, reducing the cost being optimized by up to 73% at the largest sizes tested and outperforming a state-of-the-art optimizer for fermion-to-qubit encodings. Every submission is a program accompanied by its agent's notes, so the ideas behind each improvement are recorded in human-readable form; later agents built on them, and we describe them here. These results open the way to automated, certified discovery across quantum circuit design and beyond.
AI Summary
The paper studies whether coding agents can optimize quantum compilations under fixed, exact correctness checks, including agent-assisted construction of the evaluators. Its workflow specifies a problem contract, implements verification and scoring, tests these against solved cases, deliberately invalid artifacts and published results, then freezes the evaluator before search. Nineteen accepted submissions address fermion-to-qubit encodings, Hamiltonian layout and CNOT-phase synthesis.
For ancilla-free fermionic encodings on a $15\times15$ lattice, the best total Pauli weight is 10,749: 37% below Jordan–Wigner, 23% below the published optimized comparison, and 11% below the strongest standard encoding evaluated internally. Clifford-transvection searches preserve Majorana anticommutation; correctness is checked by binary symplectic algebra. A separate encoding reaches maximum weight nine versus the published eleven. For routing random interaction graphs on an $8\times8$ grid, schedule replay verifies a mean 552.4 SWAPs versus 2,016 for a snake-path swap network. Eight-qubit synthesis yields 23 CNOTs versus 46 for naive diagonal synthesis, and 25 versus 73 with a linear map. Exhaustive searches establish optimality for the six synthesis targets through four qubits.
The comparisons have stated qualifications: the routing reference ignores graph structure, synthesis references are not state-of-the-art optimizers, and discrepancies in published fermionic benchmarks complicate comparisons. Routing depth worsens, and larger-instance optima remain unknown. Synthesis provides stored circuits for fourteen fixed targets, not a submitted general search algorithm; verification enumerates $2^n$ inputs. Evaluator validation is testing rather than formal proof. The framework requires exact or rigorously certified classical verification and deterministic costs.
Little sign of AI-generated textAI use disclosedUnclear noveltyMissing comparisonOverclaims
Claude (Anthropic) was used to write the code of the evaluation harnesses and of the challenge-construction system, and to assist in writing this paper.
Error found: The claim that pre-search validation makes every later evaluator–submission disagreement attributable to the submission is unjustified, because finite validation tests cannot exclude remaining verifier defects.
The unitary-synthesis verifier checks all 2^n basis states and therefore does not demonstrate the scalable efficient verification emphasized by the proposed framework.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
Compiling quantum algorithms for real devices involves hard optimization problems that can decide whether an algorithm runs at scale. We present a general method for solving such problems with AI agents, with every result certified. It builds on autoresearch loops, in which agents iteratively improve programs against a fixed evaluator and which have recently improved quantum cryptanalysis circuits, and adds two elements: a frozen, exact verifier certifies every result, and this verifier, together with the challenge around it, is built by an agent that formulates the problem with a researcher and validates the verifier against a solved case, deliberately broken solutions and published results before any solution exists. The method applies to any optimization problem whose solutions can be verified exactly and efficiently. We apply it to three compilation problems in quantum simulation, namely fermion-to-qubit encoding, Hamiltonian layout and unitary synthesis. In each, the agents improved substantially on published or standard constructions, reducing the cost being optimized by up to 73% at the largest sizes tested and outperforming a state-of-the-art optimizer for fermion-to-qubit encodings. Every submission is a program accompanied by its agent's notes, so the ideas behind each improvement are recorded in human-readable form; later agents built on them, and we describe them here. These results open the way to automated, certified discovery across quantum circuit design and beyond.
The paper studies whether coding agents can optimize quantum compilations under fixed, exact correctness checks, including agent-assisted construction of the evaluators. Its workflow specifies a problem contract, implements verification and scoring, tests these against solved cases, deliberately invalid artifacts and published results, then freezes the evaluator before search. Nineteen accepted submissions address fermion-to-qubit encodings, Hamiltonian layout and CNOT-phase synthesis.
For ancilla-free fermionic encodings on a 15×15 lattice, the best total Pauli weight is 10,749: 37% below Jordan–Wigner, 23% below the published optimized comparison, and 11% below the strongest standard encoding evaluated internally. Clifford-transvection searches preserve Majorana anticommutation; correctness is checked by binary symplectic algebra. A separate encoding reaches maximum weight nine versus the published eleven. For routing random interaction graphs on an 8×8 grid, schedule replay verifies a mean 552.4 SWAPs versus 2,016 for a snake-path swap network. Eight-qubit synthesis yields 23 CNOTs versus 46 for naive diagonal synthesis, and 25 versus 73 with a linear map. Exhaustive searches establish optimality for the six synthesis targets through four qubits.
The comparisons have stated qualifications: the routing reference ignores graph structure, synthesis references are not state-of-the-art optimizers, and discrepancies in published fermionic benchmarks complicate comparisons. Routing depth worsens, and larger-instance optima remain unknown. Synthesis provides stored circuits for fourteen fixed targets, not a submitted general search algorithm; verification enumerates 2n inputs. Evaluator validation is testing rather than formal proof. The framework requires exact or rigorously certified classical verification and deterministic costs.
We discuss an effective paradigm of utilizing AI agents to better compile quantum circuits when one can constrain bottleneck subroutines to have some form of classically verifiable structure. For example, when critical subroutines are composed exclusively of reversible classical logic, phasing operations, and measurement based uncomputation, one can efficiently simulate execution for classical inputs. This property facilitates efficient certification of approximate circuit synthesis via a form of fuzz testing, provided suitable assumptions on the input distribution. While this pattern was recently exploited to improve quantum cryptanalysis circuits, here we highlight the flexibility of the broader approach by discussing three important applications. (1) The computation of 1/x that bottlenecks real space simulations of the Coulomb operator such as those important for modeling inertial confinement fusion reactors. (2) The decoding of Reed-Solomon codes which bottlenecks the decoded quantum interferometry algorithm when used to approximately solve the optimal polynomial intersection problem. And (3) the quantum look-up tables and controlled application of Givens rotations that bottleneck the most performant quantum algorithms for simulating chemistry in second quantization. We discuss reference implementations of these algorithms and how they can be constrained and verified in the interest of seeding a Kaggle competition designed to accelerate AI-based research into the optimization of their compilation.
The timeline by which quantum computers can solve breakthrough problems in science depends on improved quantum algorithms as much as on better hardware. Such algorithmic advances are typically achieved through many years of expert human collaboration. In mathematics and computer science, large language models have accelerated progress through automated discovery loops. However, quantum algorithms can only be evaluated by classical computers on restricted or small instances. Here we demonstrate automated discovery of quantum programs beyond these regimes including order-of-magnitude improvements on scientific applications. To achieve this, we build a quantum program library in the Lean theorem prover in which Gemini 4 agents synthesize novel programs against formal problem specifications and verify their correctness and cost. Agent teams collaborate, sharing results and knowledge through wikis and discussing ideas at organized conferences. Our results improve on prior baselines for parallel search (up to 2.55x), quantum arithmetic (290x), and simulations relevant to fusion, photochemistry and drug-metabolism by 11x, 8x and 2.5x respectively. Combining human and artificial intelligence at scale, under the rigor of mathematical proof, lets researchers design algorithms that today's classical and quantum computers cannot test, shortening the path to practical quantum computation.
Variational quantum circuits (VQCs) constitute a central building block of near-term quantum machine learning (QML), yet the principled design of expressive and trainable architectures remains a major open challenge. The VQC design space grows combinatorially with the number of qubits, layers, entanglement structures, and gate parameterizations, rendering manual circuit construction inefficient and often suboptimal. We introduce an autonomous agent-based framework for VQC architecture search that integrates high-level reasoning with a quantum simulation environment. The agent proposes candidate circuit architectures, evaluates them through fully automated training and validation pipelines, and iteratively improves its design strategy via performance-driven feedback. Empirically, we show that the agent autonomously evolves circuit architectures from simple initial ansätze toward increasingly expressive designs, progressively trying to improve task performance. This demonstrates that agentic AI can effectively navigate and refine the VQC design landscape with minimal human intervention, providing a scalable methodology for automated quantum model development in the Noisy Intermediate-Scale Quantum (NISQ) regime.
The practical implementation of quantum algorithms on noisy intermediate-scale quantum devices encounters operational limitations due to decoherence and other sources of noise inherent in real hardware. To mitigate these errors while preserving the original functionality of the algorithm, shorter quantum circuits are therefore preferred. This motivates the development of effective quantum circuit optimization algorithms. Learning-based approaches have emerged as a leading candidate, yet existing autonomous agents remain inefficient, spending most of their training capacity rediscovering elementary reductions that deterministic rule-based methods already handle reliably. To address this challenge, we propose a reinforcement learning framework that embeds a deterministic Commutation-and-Reduction (CR) algorithm directly into the training environment. After every agent action, the CR algorithm automatically resolves elementary commutations and cancellations, enabling the agent to focus its learning capacity on the non-trivial optimizations where reinforcement learning adds real value. Empirical evaluation across two gate sets, the universal Clifford+T basis and the CNOT+Pauli basis, shows that RL+CR produces shorter circuits than a standard RL agent at all tested scales. We demonstrate that RL trained on smaller quantum circuits can be applied to larger quantum circuits. On 20-qubit Clifford+T circuits, five times larger than the training circuits, RL+CR removes twice as many gates as standard RL. This work provides a robust approach that could accelerate the compilation and optimization processes for future fault-tolerant and utility-scale quantum systems.
Optimizing and simulating quantum circuits at scale are critical bottlenecks in realizing the full potential of quantum computing. This thesis delivers a comprehensive suite of systems to overcome these limitations, spanning both physical compilation and classical simulation. For quantum circuit optimization, we first automate the discovery and verification of transformation rules by introducing Equivalence Circuit Class (ECC) sets and an efficient generation algorithm for arbitrary gate sets. We deploy these rules within the superoptimizer Quartz, which utilizes a cost-based backtracking search to escape local minima that trap traditional rule-based methods. To eliminate the exponential runtime penalties inherent to pure search, we introduce QALM, a hybrid optimization framework. By unifying search-based exploration with greedy rule-based exploitation in an interleaved control loop, QALM dynamically escapes local minima without exhaustive backtracking. Evaluated across a large benchmark suite of 248 diverse circuits, QALM outperforms existing search-based optimizers in final circuit quality while operating with the computational efficiency of rule-based systems. While the prior two approaches aim for global optimization, this problem is intrinsically QMA-hard, creating a bottleneck for large programs. To circumvent this issue and scale up, we introduce OAC, a cut-and-meld circuit optimization algorithm. OAC cuts a circuit into subcircuits, applies an existing oracle optimizer independently, and seamlessly melds the results. This algorithm operates with a linear number of oracle calls while attaining local optimality. Empirical evaluation shows that OAC significantly improves the efficiency of state-of-the-art optimizers while enhancing overall quality. Beyond physical execution, the scalable simulation of quantum circuits on classical hardware presents another major challenge. To address this, we present Atlas, a distributed GPU-based simulator that hierarchically partitions circuits to exploit data parallelism while minimizing communication overhead. By leveraging integer linear programming to allocate structurally related gates to nearby GPUs and dynamic programming for kernel scheduling, Atlas runs significantly faster than prior state-of-the-art GPU simulators. Together, these frameworks provide a robust toolchain, improving both the execution of quantum circuits on physical devices and their scalable classical simulation.
We propose Open Autoresearch, a paradigm in which humans and AI agents publish evaluator-verified improvements to a public leaderboard. We instantiate it in ECDSA.Fail, optimizing reversible secp256k1 point-addition circuits, a bottleneck in Shor's algorithm for elliptic-curve cryptography. The benchmark minimizes the spacetime-inspired score S=Q×T, where Q is peak logical qubit width and T is average executed Toffoli count. Participants reduced S by 86.1%. At the data cutoff (26 July 2026), the best-scoring circuit uses 1,151 qubits and 1,299,453 average executed Toffoli gates, giving Q×T≈1.496 billion. This is more than 50% below Google's published point-addition score thresholds (arXiv:2603.28846), under different accounting conventions. Because the benchmark supplies one addend classically, we construct a coherent windowed-addition-compatible variant implementing the single-call interface required by windowed Shor. It uses 1,162 qubits and 1,684,161 average executed Toffoli gates. On 100,000 random inputs, its empirical success probability is p^=0.99809, giving Q×T/p^≈1.961 billion under an independently rerunnable per-call sensitivity model, not a full-Shor success estimate. Its qubit and Toffoli counts lie below Google's published thresholds and Schrottenloher's reported operating points (arXiv:2606.02235), although differing interfaces, accounting conventions, and validation scope preclude formal dominance. After the cutoff, the score was further reduced to 1.259 billion, while a separate low-width circuit reached 813 qubits. The public record shows AI agents complementing human judgment, providing evidence for open autoresearch on efficiently evaluable, machine-checkable objectives.