Researcher-in-the-Loop Evolutionary Discovery of Entanglement-Assisted Quantum Code Families

Jae-Hyun Baek; Byung-Sun Won; Jon-Lark Kim

AI summary · arXiv: 2610.41183v1 · cs.AI

Researcher-in-the-Loop Evolutionary Discovery of Entanglement-Assisted Quantum Code Families

Jae-Hyun Baek, Byung-Sun Won, Jon-Lark Kim

Original paper Submitted to arXiv on 30 Oct. 2026 25 pages, 6 figures. Code, data and Lean proofs: https://github.com/LeGenAI/llm_guided_EAQECC_search License: not stated (all rights reserved) Also in: fr

How to read this summary? Illustrated guide

The essentials

Researcher-guided program evolution discovered finite entanglement-assisted quantum codes that researchers generalized into a proved family, closing 54 gaps in archived code tables.

Based on the abstract only: check the details in the original paper.

Question

How can LLM-guided program evolution produce mathematical constructions supported by distinct evidence of validity, generality, and reproducibility? The paper investigates this question through entanglement-assisted quantum error-correcting codes.

Method

The pipeline combines program evolution, exact evaluation, researcher-led mathematical generalization, certified nonexistence checks, and formal proof checking. Researchers generalized evolved finite qubit constructions into a code family whose underlying algebraic certificate is checked in Lean. A controlled cohort of eight paired generation blocks compared validation-selected iterative programs with independently generated programs at unseen lengths n = 13 and 15.

Findings

Result Researchers generalized the evolved constructions to the family [[n,1,n−1;n−q−1]]_q for prime powers q ≥ 3 with n ≥ 2q + 1, and for q = 2 with odd n ≥ 5.
See the passage in the paper
Starting from tasks that did not specify a family formula, evolved programs produced finite qubit constructions whose structure researchers generalized to the family $[[n,1,n-1;n-q-1]]_q$ for prime powers $q\ge3$ at $n\ge2q+1$ and for $q=2$ at odd $n\ge5$.
Result Lean checks the family's underlying algebraic certificate, and the family together with entanglement lifting closes 54 gaps in archived EAQECC tables.
See the passage in the paper
Lean checks the algebraic certificate underlying this family. Together with entanglement lifting, the family closes 54 gaps in archived EAQECC tables.
Result The released evidence contains 115 verified code records and ten nonexistence results supported by checked DRAT proofs.
See the passage in the paper
The released evidence includes 115 verified code records, ten nonexistence results with checked DRAT proofs, family implementations, and preserved search programs for exact checks and seeded replay.
Result At unseen lengths n = 13 and 15, validation-selected iterative programs achieved success rates of 46.1% and 31.6%, versus 19.5% and 3.1% for independently generated programs.
See the passage in the paper
In a controlled cohort of eight paired generation blocks, validation-selected iterative programs achieve success in 46.1% and 31.6% of held-out executions at unseen lengths $n=13,15$, compared with 19.5% and 3.1% for independently generated programs.

Limitations

The reported controlled comparison is limited to eight paired generation blocks and held-out executions at two unseen lengths.
See the passage in the paper
In a controlled cohort of eight paired generation blocks, validation-selected iterative programs achieve success in 46.1% and 31.6% of held-out executions at unseen lengths $n=13,15$, compared with 19.5% and 3.1% for independently generated programs.

Why it matters

The authors present a route from finite computational discoveries to a proved quantum-code family with independently checkable evidence. Their contribution combines program evolution with researcher analysis rather than treating generated constructions alone as evidence of generality.

Written by AI from the paper. Passages are checked word for word against the text we analysed; the rest may contain errors.

AI-generated keywords: Entanglement-assisted quantum codes; Evolutionary program search; Researcher-in-the-loop discovery; Formal verification; Mathematical generalization

Read further

Comprehensive Summary
Layman's Summary
Blog article
Original abstract

“Researcher-in-the-Loop Evolutionary Discovery of Entanglement-Assisted Quantum Code Families,” by Jae-Hyun Baek, Byung-Sun Won, and Jon-Lark Kim, examines how large language model–guided program evolution can support reliable mathematical discovery. Its central question is not merely whether AI-assisted search can produce useful examples, but how finite computational findings can lead to general constructions supported by reproducible evidence and formal verification. The application is entanglement-assisted quantum error-correcting codes (EAQECCs), which use shared entanglement to protect quantum information. This summary draws exclusively on the supplied title and abstract. The full text is unavailable under the stated license, so the detailed methods and limitations cannot be independently assessed. The authors propose a researcher-in-the-loop pipeline that distinguishes several types of evidence. Program evolution searches for constructions, exact evaluation checks individual examples, and researchers analyze their structure to formulate a general family. Certified nonexistence checks and formal proof checking provide further validation. These distinctions matter: a program’s ability to generate valid examples does not establish that its construction works generally. The pipeline therefore connects computational discovery, mathematical generalization, and independently checkable verification without treating them as interchangeable. Starting from tasks that did not supply a family formula, evolved programs generated finite qubit constructions that researchers subsequently generalized. The resulting family has length n, encodes one logical qubit, has distance n − 1, and requires n − q − 1 shared entangled pairs. According to the abstract, the construction applies when q is a prime power of at least 3 and n ≥ 2q + 1. It also applies when q = 2 and n is odd and at least 5. Lean, a formal proof assistant, checks the algebraic certificate underlying the family, although the abstract does not provide the certificate or explain the proof’s organization in detail. Together with a procedure called entanglement lifting, the family closes 54 gaps in archived EAQECC tables. The released evidence includes 115 verified code records, ten nonexistence results supported by checked DRAT proofs, implementations of the family, and preserved search programs. These programs enable exact checks and replay with specified random seeds, allowing examination of both the results and the discovery process. The paper also reports a controlled comparison across eight paired generation blocks. Programs selected through validation during iterative evolution succeeded in 46.1% of held-out executions at the previously unseen length 13, compared with 19.5% for independently generated programs. At the unseen length 15, the corresponding success rates were 31.6% and 3.1%. These findings support the value of iterative program evolution for generalization within the tested setting. However, the abstract does not specify the success criterion, execution counts, or statistical uncertainty. Overall, the paper presents AI-assisted mathematical discovery as a collaborative process: evolved programs produce promising finite examples, researchers identify general structure, and exact and formal methods provide stronger guarantees. The reported code family, table improvements, and released verification artifacts support this approach. Nevertheless, the empirical comparison covers a particular task and a small cohort of generation blocks. Without the full text, the approach’s broader applicability, computational costs, experimental controls, and precise scope of formal verification remain unclear.

From AI-Generated Examples to Verified Quantum Code Families

Can artificial intelligence help discover new mathematics without confusing a promising pattern with a proven result? “Researcher-in-the-Loop Evolutionary Discovery of Entanglement-Assisted Quantum Code Families,” by Jae-Hyun Baek, Byung-Sun Won, and Jon-Lark Kim, explores that question through quantum error correction. Its central contribution is not simply a collection of useful codes, but a workflow connecting computational search, human insight, and independently checkable evidence.

This article draws exclusively on the supplied title and abstract. The full text is unavailable under the stated license, so the detailed methods, experimental controls, and limitations cannot be independently assessed. That boundary matters especially for a paper whose central theme is distinguishing different levels of evidence.

Why Quantum Codes Need More Than Good Examples

Quantum error-correcting codes protect quantum information against errors. Entanglement-assisted quantum error-correcting codes, or EAQECCs, use shared entanglement as an additional resource. Researchers seek constructions with useful combinations of code length, information capacity, error protection, and entanglement requirements.

Finding a valid code at one length is valuable, but it does not establish a construction that works across an entire family. Even a program that repeatedly produces successful examples may fail outside the cases tested. The paper addresses this gap directly: how can finite computational discoveries become general mathematical constructions with stronger guarantees?

A Discovery Pipeline With Distinct Responsibilities

The proposed researcher-in-the-loop pipeline assigns different jobs to computational search, researchers, and verification tools. Large language model–guided program evolution searches for constructions. Exact evaluation checks individual examples. Researchers then examine the resulting structures and formulate a general family.

The workflow also incorporates certified nonexistence checks and formal proof checking. These are not interchangeable forms of validation. Checking that a particular code works establishes something different from proving that an algebraic construction applies throughout a specified range.

Program evolution produces candidate construction methods.
Exact evaluation checks finite code instances.
Researcher analysis turns observed structure into a proposed general family.
Formal and certificate-based checks provide additional, independently checkable support.

This separation is the paper’s most broadly relevant idea. AI-generated outputs can guide discovery, but their evidential role should remain clear. Successful examples are a starting point for generalization, not a substitute for it.

The Quantum Code Family That Emerged

According to the abstract, the search began with tasks that did not supply a family formula. Evolved programs generated finite qubit constructions, which the researchers subsequently generalized. The resulting family has length n, encodes one logical qubit, has distance n − 1, and requires n − q − 1 shared entangled pairs.

The stated conditions are precise. The construction applies when q is a prime power of at least 3 and n ≥ 2q + 1. It also applies when q = 2 and n is odd and at least 5. These restrictions are part of the result: the abstract does not claim that the family works for arbitrary parameter choices.

Lean, a formal proof assistant, checks the algebraic certificate underlying the family. This adds an important verification layer, although the abstract does not provide the certificate or explain the proof’s organization. It would therefore be too broad to conclude that every component of the discovery pipeline has been formally verified.

Reproducibility Beyond a Results Table

Together with a procedure called entanglement lifting, the family closes 54 gaps in archived EAQECC tables. The released evidence includes 115 verified code records, ten nonexistence results supported by checked DRAT proofs, implementations of the family, and preserved search programs.

Those artifacts make the work more than a report of final outcomes. The programs support exact checks and replay with specified random seeds, allowing examination of both the reported results and aspects of the discovery process. Preserving how candidates were found complements checking whether those candidates are valid.

Does Iterative Evolution Improve Generalization?

The paper also reports a controlled comparison across eight paired generation blocks. Programs selected through validation during iterative evolution were tested at previously unseen lengths and compared with independently generated programs.

At length 13, evolved programs succeeded in 46.1% of held-out executions, compared with 19.5% for independently generated programs.
At length 15, the corresponding success rates were 31.6% and 3.1%.

These results support the usefulness of iterative program evolution within the tested setting. However, the abstract does not specify the success criterion, execution counts, or statistical uncertainty. The comparison therefore offers encouraging evidence, not a general guarantee about AI-assisted mathematical discovery.

A Collaborative Model of Mathematical Discovery

The paper presents a productive division of labor: evolved programs generate promising examples, researchers identify general structure, and exact and formal methods strengthen the resulting claims. Its reported code family, table improvements, and verification artifacts give that approach concrete substance.

Broader applicability, computational costs, and the precise scope of formal verification remain unclear without the full text. Still, the central lesson is compelling: reliable AI-assisted discovery depends not only on finding patterns, but on making the path from examples to mathematical guarantees explicit and checkable.

Created on 02 Nov. 2026

Was this synthesis useful?

Score: 0

Ask questions about this paper to our AI assistant

How to chat with this paper? Illustrated guide

Compare with the PDF

You can also chat with multiple papers at once here.

⚠The license of the paper does not allow us to build upon its content and the AI assistant only knows about the paper metadata rather than the full article.

AI assistant instructions?

Tools for this paper

Explain it differently 3 free rewrites left today

Save to a collection Group papers in collections to compare them, chat with them and export them. Create a free account Log in

Export BibTeX · RIS (Zotero, Mendeley, EndNote) · Obsidian (Markdown) · Notion (Markdown)

Advanced tools Premium

For researchers who follow a field closely: citation alerts, journal-club slides, version comparison and Zotero export. Discover Premium

Zotero

Citation alerts Citation alerts guide

Who works on it? The people and institutions who publish on its topic

Weekly letter Presented in "Artificial intelligence this week" of 2 November 2026. Artificial intelligence this week: when better scores hide the wrong answer

Journal club Premium+ Slides to present it Propose for the journal club

Versions What changed between versions

Talk about it in public Premium+ Press kit: release, questions, hype check Press kit guide

Papers around this one

See them on a map

Ranked by how close their title and abstract are to this paper’s. Select some to read them together.

Closest papers

Close Another approach to the problem
Quantum Error Correction via Noise Guessing Decoding
Diogo Cruz, Francisco A. Monteiro, Bruno C. Coutinho · 2022 · Quantum Physics · cited 23 times · AI summary available

Noise Guessing Decoding constructs and decodes quantum error-correction codes across chosen block lengths, while this paper discovers and proves a family of entanglement-assisted codes.

Abstract
Quantum error correction codes (QECCs) play a central role both in quantum communications and in quantum computation, given how error-prone quantum technologies are. Practical quantum error correction codes, such as stabilizer codes, are generally structured to suit a specific use, and present rigid code lengths and code rates, limiting their adaptability to changing requirements. This paper shows that it is possible to both construct and decode QECCs that can attain the maximum performance of the finite blocklength regime, for any chosen code length and when the code rate is sufficiently high. A recently proposed strategy for decoding classical codes called GRAND (guessing random additive n…
Close Broader context
Implementation of quantum imaginary-time evolution method on NISQ devices: Nonlocal approximation
Hirofumi Nishi, Taichi Kosugi, Yu-ichiro Matsushita · 2020 · npj Quantum Inf 7, 85 (2021) · cited 103 times · AI summary available

Nonlocal approximation reduces circuit depth for quantum imaginary-time evolution on NISQ devices, rather than constructing entanglement-assisted error-correcting codes.

Abstract
The imaginary-time evolution method is widely known to be efficient for obtaining the ground state in quantum many-body problems on a classical computer. A recently proposed quantum imaginary-time evolution method (QITE) faces problems of deep circuit depth and difficulty in the implementation on noisy intermediate-scale quantum (NISQ) devices. In this study, a nonlocal approximation is developed to tackle this difficulty. We found that by removing the locality condition or local approximation (LA), which was imposed when the imaginary-time evolution operator is converted to a unitary operator, the quantum circuit depth is significantly reduced. We propose two-step approximation methods base…
Close Broader context
Logical Error Rates for a [[4,2,2]]-Encoded Variational Quantum Eigensolver Ansatz
Meenambika Gowrishankar, Daniel Claudino, Jerimiah Wright et al. · 2024 · Quantum Physics · cited 9 times · Not summarized yet

This paper measures how a [[4,2,2]] code affects logical error rates in a VQE application, whereas the main paper constructs a family of entanglement-assisted codes.

Abstract
Application benchmarks that run on noisy, intermediate-scale quantum (NISQ) computing devices require techniques for mitigating errors to improve accuracy and precision. Quantum error detection codes offer a framework by which to encode quantum computations and identify when errors occur. However, the subsequent logical error rate depends on the encoded application circuit as well as the underlying noise. Here, we quantify how the [[4,2,2]] quantum error detection code improves the logical error rate, accuracy, and precision of an encoded variational quantum eigensolver (VQE) application. We benchmark the performance of the encoded VQE for estimating the energy of the hydrogen molecule with …
Close Same method, another problem
Code Evolution Graphs: Understanding Large Language Model Driven Design of Algorithms
Niki van Stein, Anna V. Kononova, Lars Kotthoff et al. · 2025 · Neural and Evolutionary Computing · cited 21 times · AI summary available

Code Evolution Graphs analyzes how LLM-generated algorithms change during evolutionary prompting, a process also used in the main paper to discover code constructions.

Abstract
Large Language Models (LLMs) have demonstrated great promise in generating code, especially when used inside an evolutionary computation framework to iteratively optimize the generated algorithms. However, in some cases they fail to generate competitive algorithms or the code optimization stalls, and we are left with no recourse because of a lack of understanding of the generation process and generated codes. We present a novel approach to mitigate this problem by enabling users to analyze the generated codes inside the evolutionary process and how they evolve over repeated prompting of the LLM. We show results for three benchmark problem classes and demonstrate novel insights. In particular…
Close Broader context
Bridging Entanglement and Magic Resources through Operator Space
Neil Dowling, Kavan Modi, Gregory A. L. White · 2025 · Quantum Physics · cited 35 times · AI summary available

This paper relates operator entanglement to magic resources in quantum circuits, not to entanglement-assisted error-correcting code construction.

Abstract
Local operator entanglement (LOE) dictates the complexity of simulating Heisenberg evolution using tensor network methods, and serves as strong dynamical signature of quantum chaos. We show that LOE is also sensitive to how non-Clifford a unitary is: its magic resources. In particular, we prove that LOE is always upper-bound by three distinct magic monotones: $T$-count, unitary nullity, and operator stabilizer Rényi entropy. Moreover, in the average case for large, random circuits, LOE and magic monotones approximately coincide. Our results imply that an operator evolution that is expensive to simulate using tensor network methods must also be inefficient using both stabilizer and Pauli trun…
Close Broader context
Determining eigenstates and thermal states on a quantum computer using quantum imaginary time evolution
Mario Motta, Chong Sun, Adrian Teck Keng Tan et al. · 2019 · Nature Physics 16, 205-210 (2020) · cited 866 times · AI summary available

This paper develops quantum imaginary-time evolution for finding eigenstates and thermal states, rather than evolving programs to discover quantum codes.

Abstract
The accurate computation of Hamiltonian ground, excited, and thermal states on quantum computers stands to impact many problems in the physical and computer sciences, from quantum simulation to machine learning. Given the challenges posed in constructing large-scale quantum computers, these tasks should be carried out in a resource-efficient way. In this regard, existing techniques based on phase estimation or variational algorithms display potential disadvantages; phase estimation requires deep circuits with ancillae, that are hard to execute reliably without error correction, while variational algorithms, while flexible with respect to circuit depth, entail additional high-dimensional clas…
Close Broader context
Quantum eigenvalue processing
Guang Hao Low, Yuan Su · 2024 · Quantum Physics · cited 60 times · AI summary available

Quantum eigenvalue processing transforms eigenvalues of block-encoded operators, a separate quantum algorithmic problem from code-family discovery.

Abstract
Many problems in linear algebra -- such as those arising from non-Hermitian physics and differential equations -- can be solved on a quantum computer by processing eigenvalues of the non-normal input matrices. However, the existing Quantum Singular Value Transformation (QSVT) framework is ill-suited to this task, as eigenvalues and singular values are different in general. We present a Quantum EigenValue Transformation (QEVT) framework for applying arbitrary polynomial transformations on eigenvalues of block-encoded non-normal operators, and a related Quantum EigenValue Estimation (QEVE) algorithm for operators with real spectra. QEVT has query complexity to the block encoding nearly recover…
Close Broader context
Benchmarking highly entangled states on a 60-atom analog quantum simulator
Adam L. Shaw, Zhuo Chen, Joonhee Choi et al. · 2023 · Quantum Physics · cited 107 times · AI summary available

This paper benchmarks entanglement in a 60-atom simulator, whereas the main paper studies algebraic constructions of entanglement-assisted quantum error-correcting codes.

Abstract
Quantum systems have entered a competitive regime where classical computers must make approximations to represent highly entangled quantum states. However, in this beyond-classically-exact regime, fidelity comparisons between quantum and classical systems have so far been limited to digital quantum devices, and it remains unsolved how to estimate the actual entanglement content of experiments. Here we perform fidelity benchmarking and mixed-state entanglement estimation with a 60-atom analog Rydberg quantum simulator, reaching a high entanglement entropy regime where exact classical simulation becomes impractical. Our benchmarking protocol involves extrapolation from comparisons against many…

Go further

How are these papers found?

We compare the title and abstract of this paper with those of every paper summarized on this site, and with the papers Semantic Scholar recommends, those it cites and those citing it (for a much-cited paper, its most recent citations and, from OpenAlex, the most cited ones). Among the prior work and the follow-ups, the papers that use each other in depth, a method it uses and the much-cited follow-ups come a little higher. The levels (very close, close, related) come from the similarity of the texts, not from a judgment of quality; * marks a paper without an abstract, compared by its title. The sentence under a paper is written by an AI from the two abstracts: check it in the papers. A paper marked not relevant by three readers leaves the list.

Updated on 2 November 2026.

Disclaimer: The AI-based summarization tool and virtual assistant provided on this website may not always provide accurate and complete summaries or responses. We encourage you to carefully review and evaluate the generated content to ensure its quality and relevance to your needs.

Researcher-in-the-Loop Evolutionary Discovery of Entanglement-Assisted Quantum Code Families

The essentials

Question

Method

Findings

Limitations

Why it matters

Read further

Definitions

From AI-Generated Examples to Verified Quantum Code Families

Why Quantum Codes Need More Than Good Examples

A Discovery Pipeline With Distinct Responsibilities

The Quantum Code Family That Emerged

Reproducibility Beyond a Results Table

Does Iterative Evolution Improve Generalization?

A Collaborative Model of Mathematical Discovery

Ask questions about this paper to our AI assistant

Tools for this paper

Advanced tools Premium

Papers around this one

Closest papers

Go further

Cite this paper (BibTeX)