Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
PALQO: Physics-informed model for Accelerating Large-scale Quantum Optimization
Authors: Yiming Huang, Yajie Hao, Yuxuan Du, Jing Zhou, Xiao Yuan, Xiaoting Wang
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Through systematic numerical experiments, we demonstrate that our method achieves up to a 30x speedup compared to conventional methods and reduces quantum resource costs by as much as 90% for tasks involving up to 40 qubits, including ground state preparation of different quantum systems, while maintaining competitive accuracy. Our approach complements existing techniques aimed at improving the efficiency of VQAs and further strengthens their potential for practical applications. |
| Researcher Affiliation | Academia | 1Center on Frontiers of Computing Studies, and School of Computer Science, Peking University, Beijing, China 2Institute of Fundamental and Frontier Sciences University of Electronic Science and Technology of China, Chengdu, China 3College of Computing and Data Science Nanyang Technological University, Singapore, Singapore 4School of Physical and Mathematical Sciences Nanyang Technological University, Singapore, Singapore 5Department of Physics, Fudan University, Shanghai, China |
| Pseudocode | No | The paper describes the framework of PALQO in Figure 1, but this is a conceptual diagram and not structured pseudocode or an algorithm block. The methods are described in paragraph text. |
| Open Source Code | Yes | To support the community, we release our source code at [42]. [42] Implementatio of physics-informed model for accelerating large-scale quantum optimization. https://github.com/Yajie-Hao/PALQO. |
| Open Datasets | Yes | We generate these molecule Hamiltonians with Openfermion [83]. Refer to Appendix E for more details about the molecule experiments. [83] Jarrod R Mc Clean, Nicholas C Rubin, Kevin J Sung, Ian D Kivlichan, Xavier Bonet-Monroig, Yudong Cao, Chengyu Dai, E Schuyler Fried, Craig Gidney, Brendan Gimby, et al. Open Fermion: the electronic structure package for quantum computers. https: //github.com/quantumlib/Open Fermion, 2025-02-12. |
| Dataset Splits | Yes | The impact of the number of training samples on model prediction. We performed experiments on 12-qubit TFIMs with HEA ansatz using various sample sizes to explore how the number of training samples affects the performance.In Fig. 6, the results validate that PALQO can achieve satisfactory performance even with a limited number of training samples. |
| Hardware Specification | Yes | The training of all models were conducted on a single Nvidia RTX 4090 GPU, with 24GB VRAM and 24 CPU cores (Intel i9-13900K). |
| Software Dependencies | Yes | We employ PennyLane (v0.32.0) for the quantum circuit simulation and PyTorch (v2.0.1) for the classical machine learning models. |
| Experiment Setup | Yes | For all ansatzes, their initial parameters (0) are uniformly sampled from [0, 1], following the strategy adopted in Qu ACK [36]. The gradient descent is set as the default optimizer. For implementing PALQO, we randomly initialize the parameters w of the neural network fw from [ 1, 1] and employ the Adam as the optimizer... To improve the training stability and convergence, we utilize a linear decay strategy to adaptively adjust the learning rate during training. Besides, the weight hyperparameters in loss function λP1, λP2, λD, are set as 1.0, 1.0 and 10 4, respectively. ...Specifically, we set a 10 3. |