arXiv subjects

Nickholas Gutierrez

Publications and source records attributed to Nickholas Gutierrez.

2 recordsLinked to original sources

arXiv2025

Enhanced Squeezing and Faster Metrology from Layered Quantum Neural Networks

Spin squeezing is a powerful resource for quantum metrology, and recent hardware platforms based on interacting qubits provide multiple possible architectures to generate and reverse squeezing during a sensing protocol. In this work, we compare the sensing performance of three such architectures: quantum reservoir computers (QRCs), quantum perceptrons, and multi-layer quantum neural networks (QNNs), when used as squeezing-based field sensors. For all models, we consider a standard metrological sequence consisting of coherent-spin preparation, one-axis-twisting dynamics, field encoding via a weak rotation, time-reversal, and collective readout. We show that a single quantum perceptron generates the same optimal sensitivity as a QRC, but in the perturbative regime it benefits from accelerated squeezing due to steering by the output qubit. Stacking perceptrons into a QNN further amplifies this effect: in a 2-layer QNN with N_in input and N_out output qubits, the optimal squeezing time is reduced by a factor of N_out, while the achievable phase sensitivity remains Heisenberg-limited, Delta phi ~ 1/(N_in + N_out). Moreover, if the layers are used sequentially, first using the outputs to squeeze the inputs and then the inputs to squeeze the outputs, the two contributions to the response add constructively. This yields a sqrt(2) enhancement in sensitivity over a QRC when N_in = N_out and requires shorter total squeezing time. Generalizing to L layers, we show that the metrological gain scales as sqrt(L) while the required squeezing time decreases as 1/N_l, where N_l is the number of qubits per layer. Our results demonstrate that the structure of quantum neural networks can be exploited not only for computation, but also to engineer faster and more sensitive squeezing-based quantum sensors.

quant-ph↗

arXiv2025

HARDMath2: A Benchmark for Applied Mathematics Built by Students as Part of a Graduate Class

Large language models (LLMs) have shown remarkable progress in mathematical problem-solving, but evaluation has largely focused on problems that have exact analytical solutions or involve formal proofs, often overlooking approximation-based problems ubiquitous in applied science and engineering. To fill this gap, we build on prior work and present HARDMath2, a dataset of 211 original problems covering the core topics in an introductory graduate applied math class, including boundary-layer analysis, WKB methods, asymptotic solutions of nonlinear partial differential equations, and the asymptotics of oscillatory integrals. This dataset was designed and verified by the students and instructors of a core graduate applied mathematics course at Harvard. We build the dataset through a novel collaborative environment that challenges students to write and refine difficult problems consistent with the class syllabus, peer-validate solutions, test different models, and automatically check LLM-generated solutions against their own answers and numerical ground truths. Evaluation results show that leading frontier models still struggle with many of the problems in the dataset, highlighting a gap in the mathematical reasoning skills of current LLMs. Importantly, students identified strategies to create increasingly difficult problems by interacting with the models and exploiting common failure modes. This back-and-forth with the models not only resulted in a richer and more challenging benchmark but also led to qualitative improvements in the students' understanding of the course material, which is increasingly important as we enter an age where state-of-the-art language models can solve many challenging problems across a wide domain of fields.

cs.LG↗