Some technical terms become so familiar that we eventually forget that they had an origin.
Quantum Architecture Search (QAS) is now one of those terms.
A strict exact-match search for the phrase “Quantum Architecture Search” on Google Scholar now returns more than 800 academic papers. The phrase appears in papers on variational quantum algorithms, quantum machine learning, quantum compiling, reinforcement learning, hardware-aware circuit design, and many other topics.
But the term itself has a rather precise starting point.
Where the name came from
The phrase Quantum Architecture Search was introduced in our work “Differentiable Quantum Architecture Search,” first posted on arXiv on October 16, 2020 https://arxiv.org/abs/2010.08561. The paper was subsequently published in Quantum Science and Technology in 2022.
The idea behind the name was straightforward.
At the time, there were already discussions of searching for quantum ansätze and automatically designing variational circuits. The word ansatz, however, naturally suggests a particular variational formulation: one chooses a parameterized family of quantum states or circuits and then optimizes its parameters.
We wanted to describe something more general.
The inspiration was partly the distinction that had already become familiar in machine learning between parameter optimization and architecture search. Neural Architecture Search does not merely optimize the weights of a neural network; it searches for the computational structure itself.
We asked a similar question for quantum circuits:
Instead of optimizing a predefined quantum ansatz, can we search for the quantum architecture itself?
That was the reason for choosing the word architecture.
It deliberately expanded the scope beyond ansatz search. The object being searched could include the choice and arrangement of gates, circuit connectivity, depth, parameter sharing, and other structural degrees of freedom. In other words, the problem was not restricted to finding a better ansatz for one particular VQA.
The term Quantum Architecture Search was intended to name this broader problem.
From a name to a research direction
Our first work introduced Differentiable Quantum Architecture Search (DQAS), treating the architecture search problem in an end-to-end differentiable framework. The paper demonstrated the idea through several different circuit-design problems, including unitary decomposition, noise-aware circuit redesign, and automatically discovering circuit layouts for QAOA.
The important point was not any particular search algorithm.
It was the abstraction.
Once quantum circuit design is viewed as an architecture search problem, many different optimization techniques can naturally enter the picture: reinforcement learning, evolutionary algorithms, gradient-based methods, predictors, meta-learning, and others. The search strategy can change without changing the underlying problem definition.
That abstraction turned out to be useful.
Subsequent work adopted the terminology for a variety of settings which introduced differentiable, reinforcement-learning, predictor-based, distributed, hardware-aware, and other variants of QAS.
A 2024 survey was eventually devoted specifically to the subject: “Quantum Architecture Search: A Survey.”
At that point, QAS was no longer simply the name of one method. It had become a recognizable research direction.
A useful historical distinction
There is an interesting difference between inventing an algorithm and introducing a research vocabulary.
An algorithm can be independently rediscovered. A useful vocabulary is different: once it captures a real conceptual boundary, it gives many subsequent works a common language.
This is why the history of the phrase matters.
There were certainly earlier studies on automatically optimizing quantum circuits, variational circuit structures, evolutionary circuit design, and related problems. But “Quantum Architecture Search” as a named research concept was introduced in the 2020 DQAS work. The paper explicitly defined QAS as the automation of quantum-circuit architecture engineering.
The choice of architecture was therefore intentional. It was not simply another name for optimizing a variational ansatz.
It was meant to define a broader abstraction:
from searching for an ansatz to searching for a quantum architecture.
How far has the term traveled?
The scale of adoption is perhaps the most interesting part of the story.
A phrase that did not exist as an established research label in 2020 is now used across hundreds of papers and across multiple subfields of quantum computing. Recent work describes QAS as a “prominent paradigm” for quantum circuit design, and dedicated surveys and specialized QAS methods have appeared as the field has expanded.
The original paper itself has also received substantial recognition. In 2025, Differentiable Quantum Architecture Search was selected for the IOP Publishing China Top Cited Paper Award. IOP's 2025 award covered research published by China-based corresponding authors across several disciplines; the award list included only eight papers in Physics.
That trajectory is a useful reminder that terminology is not merely cosmetic.
Sometimes a new name identifies a new way of organizing a problem. If the abstraction is useful enough, other researchers begin building on it, the terminology becomes shared vocabulary, and eventually the vocabulary itself becomes part of the structure of the field.
Top comments (0)