Classical State Preparation for Variational Quantum Algorithms via Reinforcement Learning
Abstract
Variational Quantum Algorithms (VQAs) potentially offer a pathway to practical quantum advantage, but their optimization is heavily hindered by barren plateaus and numerous local minima. While classically simulable Clifford circuits can warm-start VQAs to accelerate convergence, existing heuristic-based initialization methods struggle to scale within vast combinatorial search spaces. We address this bottleneck with CRiSP (a Clifford Reinforcement Learning agent for State Preparation), which casts Clifford prefix selection as a sequential decision-making problem. CRiSP utilizes Neural-Guided Monte Carlo Tree Search, driven by a Transformer-based policy trained via self-play, to insert learned Clifford gates before fixed parameterized rotations. This enables the construction of high-quality initial states entirely through polynomial-time classical stabilizer simulation. By integrating a curriculum learning strategy that progressively expands the search horizon, the agent efficiently scales to deep circuits. Evaluated on QAOA benchmarks of up to $22$ qubits and 1,370 parameters, CRiSP outperforms state-of-the-art Clifford initialization methods by a mean of 3.19× (max 45.02×) in average energy accuracy and 2.41× (max 16.01×) in best-achieved energy accuracy. Assessments on VQE tasks further demonstrate the framework's robustness and generalizability.
Related
- Quantum Hierarchical Reinforcement Learning via Variational Quantum Circuits
- Adaptive Dissipative State Preparation through Reinforcement Learning
- A comparative study of variational quantum classifiers and classical machine learning for tW−tt̅ discrimination at the LHC
- Imaginary-time-enhanced feedback-based quantum algorithms for ground-state preparation
- Accelerating noisy variational quantum algorithms with physics-informed denoising networks
BibTeX
@misc{kwun2026classical,
title = {{Classical State Preparation for Variational Quantum Algorithms via Reinforcement Learning}},
author = {Gino Kwun and Dhanvi Bharadwaj and Gokul Subramanian Ravi},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2605.23138},
eprint = {2605.23138},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.23138},
}
RIS
TY - UNPB
TI - Classical State Preparation for Variational Quantum Algorithms via Reinforcement Learning
AU - Gino Kwun
AU - Dhanvi Bharadwaj
AU - Gokul Subramanian Ravi
PY - 2026
DA - 2026/11/02
PB - arXiv
DO - 10.48550/arxiv.2605.23138
UR - https://arxiv.org/abs/2605.23138
AB - Variational Quantum Algorithms (VQAs) potentially offer a pathway to practical quantum advantage, but their optimization is heavily hindered by barren plateaus and numerous local minima. While classically simulable Clifford circuits can warm-start VQAs to accelerate convergence, existing heuristic-based initialization methods struggle to scale within vast combinatorial search spaces. We address this bottleneck with CRiSP (a Clifford Reinforcement Learning agent for State Preparation), which casts Clifford prefix selection as a sequential decision-making problem. CRiSP utilizes Neural-Guided Monte Carlo Tree Search, driven by a Transformer-based policy trained via self-play, to insert learned Clifford gates before fixed parameterized rotations. This enables the construction of high-quality initial states entirely through polynomial-time classical stabilizer simulation. By integrating a curriculum learning strategy that progressively expands the search horizon, the agent efficiently scales to deep circuits. Evaluated on QAOA benchmarks of up to $22$ qubits and 1,370 parameters, CRiSP outperforms state-of-the-art Clifford initialization methods by a mean of 3.19× (max 45.02×) in average energy accuracy and 2.41× (max 16.01×) in best-achieved energy accuracy. Assessments on VQE tasks further demonstrate the framework's robustness and generalizability.
ER -