Quantum annealing outperforms its classical counterpart, quantum-inspired annealing, when solving optimisation problems in simple systems with few dimensions; however, this advantage is lost as complexity increases. The study compared both approaches across models varying in dimensionality, disorder and frustration, finding that increasing these factors diminishes any benefit from using true quantum correlations for computation. These results offer a baseline for assessing where quantum effects genuinely improve performance during non-adiabatic annealing processes.
The research clarifies when quantum computers may not outperform traditional ones for specific optimisation problems; comparing quantum and classical approaches revealed any speed advantage from utilising quantum properties diminishes as problem complexity increases through size or randomness. This finding establishes boundaries for where quantum computation genuinely improves performance during annealing processes. Understanding these limitations helps realistically assess progress in developing this type of quantum computing hardware.
Researchers compared true quantum annealing with a classical method, ‘quantum-inspired annealing’, which mimics its structure without utilising genuine quantum effects; akin to finding the lowest point on a bumpy landscape either through gentle shaking using physics or simply letting gravity settle it. This comparison explored how factors like dimensionality, think building complexity from a line to a flat surface and beyond, and randomness affect performance.
They found that while quantum annealing excels in simple systems, any advantage vanishes as problems become more complex due to increased dimensions, disorder, or frustration: imagine assembling puzzle pieces where some actively resist fitting together.
Quantum advantages erode beyond three dimensions due to increased disorder
Quantum Annealing (QA) surpassed Quantum Inspired Annealing (QIA), a classical method mirroring QA without genuine quantum effects, showing scaling benefits in pristine low-dimensional systems. However, that advantage lessened when dimensionality exceeded three, a key threshold establishing the point where qualitative differences from quantum correlations are expected during annealing processes. Previously determining this precisely proved impossible; researchers at University of Minnesota and The Chinese University of Hong Kong (Shenzhen) then demonstrated how disorder and frustration further narrowed any performance gap between QA and QIA across various problem models.
A clear benefit existed for Quantum Annealing (QA) over Quantum Inspired Annealing (QIA), but it was limited to clean, low-dimensional systems, specifically, QA’s performance aligned with QIA’s in three or more dimensions. Analysis of random bond spin chains alongside two- and three-dimensional spin glass models revealed identical scaling between the QA and QIA protocols within their observation window.
In infinite-dimensional Lipkin, Meshkov, Glick models, both approaches exhibited exactly the same scaling; these results pinpoint scenarios where quantum correlations demonstrably influence annealing dynamics. This provides a classical benchmark against which future assessments can be made, though findings do not yet indicate whether this translates into practical speedups on real-world problems beyond those modelled here.
Quantum Annealing’s Performance Declines with Realistic Problem Complexity
The team’s work offers an important reality check for quantum annealing: while potential benefits exist in simplified cases, advantages quickly disappear as problems reflect real-world complexity, increased dimensions or inherent randomness within data sets. The study deliberately focused on specific ‘Hamiltonian structures’, mathematical descriptions defining problem characteristics; could alternative formulations, unexplored here, still unlock a consistent advantage.
Despite the possibility of beneficial alternate mathematical ‘formulations,’ this research delivers valuable insight and establishes a clear benchmark against which future progress can be measured. The researchers established a boundary where quantum annealing outperforms classical computation by rigorously comparing performance to algorithms mimicking key aspects of the quantum process, specifically ‘quantum-inspired annealing.’ This allowed them to isolate any differences stemming from unique properties.
While demonstrably outperforming its classical counterpart in simple scenarios with clean systems and few dimensions, that benefit rapidly diminished as problem complexity increased through dimensionality or randomness.
This research found that the advantage of quantum annealing over a comparable classical approach, quantum-inspired annealing, depends heavily on the structure of the problem being solved. Quantum annealing performed better in lower dimensional systems without significant disorder, but this difference became negligible in three or more dimensions and with added frustration. These results clarify when quantum correlations affect scaling during annealing and provide a baseline for evaluating future improvements to the method; researchers intend to use these findings to assess performance against other algorithms.
👉 More information
🗞 When Do Quantum Correlations Matter for Quantum Annealing?
✍️ Rongfeng Xie and Alex Kamenev (University of Minnesota); Evgenii Ievlev (The Chinese University of Hong Kong (Shenzhen))
🧠 ArXiv: https://arxiv.org/abs/2611.15351




See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
