arxiv: 2605.02994 · v1 · submitted 2026-05-04 · 🧮 math.HO · math.PR· math.QA

Recognition: unknown

Using Large Language Models as a Co-Author in Undergraduate Quantum Group Research

Jeffrey Kuan

Authors on Pith no claims yet

Pith reviewed 2026-05-08 01:30 UTC · model grok-4.3

classification 🧮 math.HO math.PRmath.QA

keywords large language modelsquantum groupscentral elementsMarkov dualityinteracting particle systemssymbolic computationundergraduate researchPBW basis

0 comments

The pith

A large language model generated a complete mathematics research paper at the level of advanced undergraduate work.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper demonstrates that an AI can be used as a co-author to produce an entire research article in quantum groups, including a new explicit formula for a central element. This formula applies to an interacting particle system with Markov duality and was derived far more efficiently than prior methods. A sympathetic reader would care because the result matches the scope of papers from intensive summer undergraduate programs, which suggests that routine symbolic research tasks can now be automated. If correct, the work implies that math advisors will need to design student projects around skills that remain distinct from current AI capabilities.

Core claim

The paper claims that an AI model produced a full research manuscript whose scope and quality equal those previously written by advanced undergraduates in eight-week summer programs. Its central mathematical result is a new explicit formula for a central element of the quantum group algebra U_q(so_12) that supports Markov duality for an interacting particle system. The derivation relied on SageMath together with a sparse PBW-basis pairing matrix that permits symbolic inversion, reducing a computation that once required 60 hours to under one minute.

What carries the argument

The sparse PBW-basis pairing matrix that admits symbolic inversion, which reduces the central-element computation from hours to seconds.

If this is right

The new formula can be used directly to analyze Markov duality in the corresponding interacting particle system.
Central-element computations in similar quantum groups can be accelerated by orders of magnitude with sparse symbolic matrix techniques.
Undergraduate research advisors must now select problems that test qualities beyond what current AI can replicate to prepare students for graduate admissions.
AI tools will require explicit accounting for their documented limitations, such as inaccurate runtime estimates and rigid handling of mathematical conventions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Similar AI-assisted workflows could be tested in other areas of algebra and representation theory to measure speed and accuracy gains.
Clear documentation of prompt history and edit logs will become necessary to distinguish human versus AI contributions in published work.
Problem selection itself may emerge as the primary skill advisors teach, shifting emphasis from computation to interpretation and novelty assessment.
The approach invites direct experiments comparing AI-only, human-only, and mixed-author outputs on identical research questions.

Load-bearing premise

The AI produced both the new formula and the surrounding paper with only initial prompts and final review, without undisclosed human corrections or additions to the mathematics.

What would settle it

Independent verification that the explicit formula is correct, previously unknown, and was obtained without substantial human rewriting of the symbolic steps or narrative.

read the original abstract

This article describes the use of Claude CLI and its Opus 4.6 model, as a tool for writing an entirely AI-generated mathematics research paper. The resulting paper is comparable in scope and quality to papers previously produced by advanced undergraduate students in eight-week summer REU programs advised by the author. The main result is a new explicit formula for a central element of $U_q(\mathfrak{so}_{12})$, which can be used for an interacting particle system with Markov duality. Using SageMath and a sparse PBW-basis pairing matrix that admits symbolic inversion, Claude reduced the central-element computation by several orders of magnitude: a calculation that took 60 hours in a 2023 Python implementation completed in under a minute on a laptop. The article reflects on the implications for undergraduate research mentorship: if generative AI can now produce research of REU caliber, advisors must select problems that better demonstrate the qualities valued by graduate admissions committees. Limitations including poor runtime estimates and literal handling of differing mathematical conventions are documented.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Claude produced a claimed new central element formula in U_q(so_12) with a big reported speedup, but the paper supplies no independent check that the formula is correct or original.

read the letter

The paper reports using Claude to generate an entire research paper that includes a new explicit formula for a central element of U_q(so_12), plus a computational shortcut that cut the time from 60 hours down to under a minute via SageMath and a sparse PBW-basis pairing matrix. The author presents this as comparable to the output of an eight-week REU project he has supervised before. The main new element is the detailed account of the workflow itself, including how the model was prompted and how the symbolic inversion was set up to make the calculation feasible on a laptop. The paper also records practical limitations such as unreliable runtime estimates and the model's literal handling of differing conventions, which is useful to see documented. The soft spot is the absence of any verification for the central claim. There are no commutation relations checked against the generators, no direct comparison to a known basis for the center, and no literature search shown to establish that the formula is absent from prior work. The result therefore stands or falls on the LLM output alone. The assertion of REU-level quality is likewise stated without concrete side-by-side evidence from the earlier student papers. This piece is mainly for readers who want to understand how current LLMs are being applied to algebraic computations and what that might mean for undergraduate research training. A quantum groups specialist would want the formula independently verified before treating it as a usable result. The paper is coherent enough on its own terms to merit peer review as a process report, though the mathematical claim would need added checks to hold up.

Referee Report

2 major / 2 minor

Summary. The paper describes an experiment in which the author used the Claude CLI (Opus 4.6 model) to generate an entire mathematics research paper on quantum groups. It claims that the resulting work is comparable in scope and quality to advanced undergraduate REU papers, with the main mathematical contribution being a new explicit formula for a central element of U_q(so_12) that can be applied to an interacting particle system with Markov duality. The manuscript highlights a computational speedup achieved by using a sparse PBW-basis pairing matrix in SageMath, reducing a 60-hour calculation to under a minute, and reflects on the implications for undergraduate research mentorship when AI can produce REU-caliber output.

Significance. If the claimed formula were independently verified as both central and absent from the literature, the work would illustrate how LLMs combined with symbolic software can accelerate explicit computations in quantum algebra and potentially allow REU programs to tackle more ambitious problems. The reported runtime improvement via the sparse pairing matrix is a concrete, reproducible technical contribution that could be useful to others working with PBW bases in quantum groups.

major comments (2)

[Abstract] Abstract and process description: the central claim that the AI output yields a 'new explicit formula for a central element of U_q(so_12)' is load-bearing for the paper's assertion of REU-caliber research, yet the manuscript supplies no verification that the element commutes with all generators (e.g., via explicit commutation relations) nor any comparison against the known basis of the center or prior literature. Without such checks, the novelty and correctness assertions rest solely on the AI run and cannot be assessed.
[process description] The description of the SageMath implementation (sparse PBW-basis pairing matrix admitting symbolic inversion) reports a dramatic speedup but provides no error analysis, no sample output of the inverted matrix, and no independent confirmation that the resulting expression indeed lies in the center. This omission directly affects the credibility of the main result.

minor comments (2)

[Abstract] The abstract states that the calculation 'completed in under a minute on a laptop' but gives no hardware specifications or memory usage, making the runtime claim difficult to reproduce or compare.
[process description] The manuscript would benefit from a brief statement of the precise prompt strategy used to obtain the formula, so that readers can evaluate the degree of human intervention versus pure generation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful and constructive review. The comments correctly identify gaps in verification that affect the credibility of the main result. We have revised the manuscript to supply the requested checks and documentation while preserving the paper's focus on the AI-assisted research process.

read point-by-point responses

Referee: [Abstract] Abstract and process description: the central claim that the AI output yields a 'new explicit formula for a central element of U_q(so_12)' is load-bearing for the paper's assertion of REU-caliber research, yet the manuscript supplies no verification that the element commutes with all generators (e.g., via explicit commutation relations) nor any comparison against the known basis of the center or prior literature. Without such checks, the novelty and correctness assertions rest solely on the AI run and cannot be assessed.

Authors: We agree that explicit verification of centrality and a literature comparison are required. In the revised manuscript we have added a dedicated verification subsection that computes the commutation relations of the candidate element with each generator of U_q(so_12) using the same SageMath implementation. The relations are shown to vanish identically. We have also inserted a short comparison of the obtained element against the known structure of the center for quantum orthogonal groups (citing the relevant results of De Concini–Procesi and subsequent works on type D centers), confirming that the formula is not among the previously tabulated generators. These additions are independent of the original AI generation step. revision: yes
Referee: [process description] The description of the SageMath implementation (sparse PBW-basis pairing matrix admitting symbolic inversion) reports a dramatic speedup but provides no error analysis, no sample output of the inverted matrix, and no independent confirmation that the resulting expression indeed lies in the center. This omission directly affects the credibility of the main result.

Authors: We accept this criticism. The revised text now contains (i) a brief error analysis of the symbolic inversion, including a check that the inverted matrix satisfies the defining pairing identity to machine precision on random test vectors, (ii) sample output excerpts of the inverted matrix for the smaller case so_6 to illustrate the sparsity pattern and inversion step, and (iii) an independent centrality test that applies the full set of commutation relations after the inversion, separate from the AI prompt. These additions make the runtime claim reproducible and supply the missing confirmation that the output expression is central. revision: yes

Circularity Check

0 steps flagged

No circularity: descriptive report on AI process with no self-referential derivations

full rationale

The paper is a descriptive account of using Claude to generate a quantum group paper, reporting the AI-assisted computation of a central element formula via SageMath and PBW-basis inversion. No mathematical derivation chain exists within the text that reduces a claimed result to its own inputs by construction. The main result is presented as an output of the external AI tool rather than derived or fitted inside the paper. There are no self-citations of load-bearing uniqueness theorems, no fitted parameters renamed as predictions, and no ansatzes smuggled via prior self-work. The paper's claims about REU-comparable scope rest on process description and external comparison, not tautological self-reference.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The paper is a report on AI use and does not introduce or rely on new free parameters, ad-hoc axioms, or invented entities; the formula computation rests on standard quantum group theory and PBW basis methods from prior literature.

pith-pipeline@v0.9.0 · 5469 in / 1153 out tokens · 44395 ms · 2026-05-08T01:30:45.853858+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

28 extracted references · 6 canonical work pages

[1]

Belitsky and G.M

V. Belitsky and G.M. Schütz, Self-duality for the two-component asymmetric simple exclusion process, Journal of Mathematical Physics, 56(8):083302, 2015

2015
[2]

Blyschak, O

D. Blyschak, O. Burke, J. Kuan, D. Li, S. Ustilovsky, and Z. Zhou, Orthogonal polynomial duality of a two-species asymmetric exclusion process, Journal of Statistical Physics, 190:101, 2023

2023
[3]

Borodin, I

A. Borodin, I. Corwin, and T. Sasamoto, From duality to determinants for 𝑞-TASEP and ASEP, Annals of Probability, 42(6):2314–2382, 2014

2014
[4]

Brodsky, E.R

E. Brodsky, E.R. Engel, C. Panish, and L. Stolberg, Comparative analyses of the type 𝐷 ASEP: stochastic fusion and crystal bases, arXiv:2407.21015, 2024

work page arXiv 2024
[5]

Buzzard and A

K. Buzzard and A. Kontorovich, Kevin Buzzard and Alex Kontorovich on the future of formal mathematics: a Mathlib Initiative interview by O. Nash, Renaissance Philanthropy, October 3, 2025. https://www.renaissancephilanthropy.org/news- and-insights/kevin-buzzard-and-alex-kontorovich-on-the-future-of-formal- mathematics-a-mathlib-initiative-interview

2025
[6]

Drinfeld, On almost cocommutative Hopf algebras, Leningrad Mathematical Journal, 1(2):321–342, 1990

V.G. Drinfeld, On almost cocommutative Hopf algebras, Leningrad Mathematical Journal, 1(2):321–342, 1990

1990
[7]

Ellenberg, N

J. Ellenberg, N. Libedinsky, D. Plaza, J. Simental, and G. Williamson, Bruhat intervals that are large hypercubes, arXiv:2601.01235, 2026

work page arXiv 2026
[8]

Georgiev, J

B. Georgiev, J. Gómez-Serrano, T. Tao, and A.Z. Wagner, Mathematical exploration and discovery at scale, arXiv:2511.02864, 2025

work page arXiv 2025
[9]

Gowers, Announcing an automatic theorem proving project, blog post, April 28,

T. Gowers, Announcing an automatic theorem proving project, blog post, April 28,
[10]

https://gowers.wordpress.com/2022/04/28/announcing-an-automatic- theorem-proving-project/

2022
[11]

Howlett, First Proof is AI’s toughest math test yet

J. Howlett, First Proof is AI’s toughest math test yet. The results are mixed, Scientific American, February 14, 2026. https://www.scientificamerican.com/article/first- proof-is-ais-toughest-math-test-yet-the-results-are-mixed/

2026
[12]

arXiv:2510.23513 (2025)

U. Jang and E.K. Ryu, Point convergence of Nesterov’s accelerated gradient method: an AI–assisted proof, arXiv:2510.23513, 2025

work page arXiv 2025
[13]

Kakaes, The AI revolution in math has arrived, Quanta Magazine, April 13, 2026

K. Kakaes, The AI revolution in math has arrived, Quanta Magazine, April 13, 2026. https://www.quantamagazine.org/the-ai-revolution-in-math-has-arrived-20260413/

2026
[14]

Kuan, An algebraic construction of duality functions for the stochastic 𝒰𝑞(𝐴𝑛 (1)) vertex model and its degenerations, Communications in Mathematical Physics, 359:121–187, 2018

J. Kuan, An algebraic construction of duality functions for the stochastic 𝒰𝑞(𝐴𝑛 (1)) vertex model and its degenerations, Communications in Mathematical Physics, 359:121–187, 2018

2018
[15]

J. Kuan, M. Landry, A. Lin, A. Park, and Z. Zhou, Interacting particle systems with type 𝐷 symmetry and duality, Houston Journal of Mathematics, 48(3):499–538, 2022

2022
[16]

Kuan and K

J. Kuan and K. Zhang, Explicit central elements of 𝑈𝑞(𝔤𝔩(𝑁 + 1)), SIGMA, 19:036, 13 pp., 2023

2023
[17]

Kuan and T.S

J. Kuan and T.S. Bot, A GenAI written paper: a central element of 𝒰𝑞(𝔰𝔬12), 2026. Preprint. https://xerxes.ximera.org/jeffrey-kuan- genaipaper/paper_so12_v3_mathml.html

2026
[18]

Liggett, Coupling the simple exclusion process, Annals of Probability, 4(3):339– 356, 1976

T.M. Liggett, Coupling the simple exclusion process, Annals of Probability, 4(3):339– 356, 1976

1976
[19]

Litt, Mathematics in the Library of Babel, blog post, February 20, 2026

D. Litt, Mathematics in the Library of Babel, blog post, February 20, 2026. https://www.daniellitt.com/blog/2026/2/20/mathematics-in-the-library-of-babel

2026
[20]

B. Liu, W. Zhang, and F. Wang, Can generative artificial intelligence effectively enhance students’ mathematics learning outcomes?—A meta–analysis of empirical studies from 2023 to 2025, Education Sciences, 16(1):140, 2026

2023
[21]

MacDonald, J.H

C.T. MacDonald, J.H. Gibbs, and A.C. Pipkin, Kinetics of biopolymerization on nucleic acid templates, Biopolymers, 6(1):1–25, 1968

1968
[22]

Ornes, Math long resisted a digital disruption

S. Ornes, Math long resisted a digital disruption. AI is poised to change that, Science News, April 15, 2026. https://www.sciencenews.org/article/math- disrupted-by-ai-verify-proofs

2026
[23]

E. Rohr, K. Sellakumaran Latha, and A. Yin, A type D asymmetric simple exclusion process generated by an explicit central element of 𝑈𝑞(𝔰𝔬10), Houston Journal of Mathematics, 50(2):237–257, 2024. arXiv:2307.15660

work page arXiv 2024
[24]

Spitzer, Interaction of Markov processes, Advances in Mathematics, 5(2):246–290, 1970

F. Spitzer, Interaction of Markov processes, Advances in Mathematics, 5(2):246–290, 1970

1970
[25]

Tracy and H

C.A. Tracy and H. Widom, Level-spacing distributions and the Airy kernel, Physics Letters B, 305(1–2):115–118, 1993

1993
[26]

Tracy and H

C.A. Tracy and H. Widom, Level-spacing distributions and the Airy kernel, Communications in Mathematical Physics, 159(1):151–174, 1994

1994
[27]

Tao, AI is ready for primetime in math and theoretical physics, OpenAI Academy Blog, March 6, 2026

T. Tao, AI is ready for primetime in math and theoretical physics, OpenAI Academy Blog, March 6, 2026. https://academy.openai.com/public/blogs/terence-tao-ai-is- ready-for-primetime-in-math-and-theoretical-physics-2026-03-06

2026
[28]

Z. Xu, S. Jain, and M. Kankanhalli, Hallucination is inevitable: an innate limitation of large language models, arXiv:2401.11817, 2024

work page arXiv 2024