Gauging the variational optimization of projected entangled-pair states

Jutho Haegeman; Laurens Vanderstraeten; Wei Tang

REVIEW

Gauging the variational optimization of projected entangled-pair states

Wei Tang, Laurens Vanderstraeten, Jutho Haegeman

Strongly Correlated Electrons · Submitted 14 Aug 2025

Pith's verdict: Plausible and potentially important diagnosis of a gauge-induced failure mode in PEPS optimization, but the abstract alone doesn't let me check the math or the numerics.

T0 · AI review

How this review was made: An AI referee (deepseek-v4-flash) read the paper on August 5, 2026 and answered a fixed list of questions about it. T0 is the starting level every paper gets. The levels describe how deeply Pith has checked the paper, never the authors.

T0 AI review An AI read the paper and answered a fixed list of questions.
T1 Audit Automatic checks that the references and files the paper relies on exist and match.
T2 Compute The paper's code or calculations re-run with the same results.
T3 Formal statement The claims written out in a formal math language a computer can read.
T4 Proof Those claims proved, and the proof checked by computer.

How Pith checks papers →

· 3 major objections · 3 minor · Ask about this paper · The compiled record · Read the full review →

Challenge this review Comment (0) Read on arXiv ↗

Cite ▾

arXiv ID2508.10822 (version 1)

Pith IDVJPNRYLV, Pith's permanent record of this paper and its review

CitationJutho Haegeman, Laurens Vanderstraeten, Wei Tang. "Gauging the variational optimization of projected entangled-pair states." arXiv:2508.10822 (2025). https://pith.science/paper/VJPNRYLV

Cite this reviewBibTeX for Pith's review of this paper

Claim paper

Why Pith says this

Highlight any text to comment on it.

Editor's note

read the letter

Here's the off-record read. The abstract describes a subtle and plausible mechanism by which variational PEPS optimization can go off the rails: the gradient can exploit gauge freedom to make the approximate contraction progressively worse while pushing the variational energy down. If true, that's a real pitfall, and the proposed gauge-fixed strategy is an obvious way to try to fix it. The novelty is in the diagnosis and the practical remedy, not in new machinery; U(1)-symmetric PEPS with point-group symmetry are a well-studied setting, and gauge freedom in tensor networks is a known issue. The analysis of reducing the gauge degrees of freedom to a single class is a useful step, and if correct it makes the problem tractable and gives a clean statement.

The soft spot is that I cannot verify any of this from the abstract. The central claim requires two pieces of evidence: a derivation or clear argument that gauge transformations change the accuracy of the approximate contraction, and numerical results showing that the gauge-fixed strategy suppresses the spurious energy lowering without just shifting the bias elsewhere. The abstract gives neither. The 'single class' reduction is the load-bearing assumption; if there are additional independent gauge directions, the failure mode and the fix may be incomplete. The final line about general settings is an extrapolation, and it goes beyond what the analysis of a specific symmetry class can support.

The paper could be wrong, but it is not obviously wrong. The reasoning in the abstract is coherent, and the problem is worth solving. On an abstract-only review it would be unfair to demand the full derivation here; in a real referee report I would want to see the math and the numerics.

For you: this is a plausible candidate for a reading-group discussion if you work on tensor networks or variational optimization in 2D. I would not cite it in my own work until I have seen the full text, but I would send it to a serious referee without hesitation—the abstract alone deserves the scrutiny.

Referee report

Major objections (3)

[Abstract] The central claim that U(1)-symmetric PEPS with point group symmetry have a single class of gauge degrees of freedom is asserted without derivation. This reduction is load-bearing: if additional independent gauge directions exist, the analysis of the failure mode and the proposed gauge-fixed remedy may be incomplete. The authors should provide a derivation or a precise characterization of the gauge orbit in the full text.
[Abstract] The manuscript claims that gradient-based optimization exploits the gauge freedom to make the tensor network contraction increasingly inaccurate and produce artificially low variational energies. No mechanism, equation, or numerical demonstration is given in the abstract. The paper needs to specify how gauge transformations interact with the approximate contraction (e.g., boundary MPS bond dimension or truncation error) and to show concrete examples where the variational energy drops below the true energy.
[Abstract] The claim that the gauge-fixed optimization 'largely suppresses' the effect is not quantified. The abstract provides no benchmark, no comparison against unfixed optimization on the same models, and no discussion of whether gauge fixing introduces other biases (e.g., slower convergence or bias toward a particular gauge). Without numerical evidence, the robustness of the proposed strategy is unverified.

Minor objections (3)

[Abstract] The phrase 'single class' of gauge degrees of freedom is undefined. Clarify whether this refers to a single continuous parameter, a single equivalence class under point group symmetries, or a single independent generator.
[Abstract] The abstract does not identify the models or lattice geometries considered. A sentence naming the benchmark systems would help the reader judge the scope of the claim.
[Abstract] No references are provided in the abstract; since this is an abstract-only review, I cannot check whether prior work on gauge fixing in PEPS (e.g., in the context of fermionic PEPS or symmetric tensor networks) is properly acknowledged.

How Pith checks papers → · Full audit of this paper →

The paper

The paper claims

Gradient-based optimization of U(1)-symmetric PEPS can exploit gauge freedom to drive tensor-network contraction off target and produce artificially low variational energies; gauge-fixed optimization suppresses this.

The reading

This paper tries to establish that gauge degrees of freedom in projected entangled-pair states are not harmless during variational optimization. For a U(1)-symmetric PEPS with point group symmetry, the gauge freedom reduces to a single class, and gradient-based optimizers learn to exploit it. The result is that tensor-network contraction becomes increasingly inaccurate and the reported variational energy drifts artificially low. The paper proposes a gauge-fixed optimization that suppresses this effect and makes the energy minimization more robust. A sympathetic reader would care because variational PEPS is a standard tool for two-dimensional quantum ground states, and artificially low energies would silently undermine its reliability.

What carries the argument

The central object is the gauge degree of freedom of a U(1)-symmetric PEPS with point group symmetry, reduced to a single class of gauge transformations. In standard gradient-based optimization this gauge direction is left free, and the optimizer can drift along it, degrading the quality of the tensor-network contraction while the energy appears to drop. The proposed gauge-fixed strategy pins down this direction during optimization, which is what suppresses the artifact.

Extended reading notes

Core claim

The paper argues that for a U(1)-symmetric projected entangled-pair state with point group symmetry, gauge transformations—local redefinitions of the tensors that leave the physical state invariant—reduce to a single class. Gradient-based optimization exploits this gauge freedom: instead of improving the variational state, the optimizer can use the gauge direction to lower the energy estimate while the underlying tensor-network contraction grows inaccurate. The paper identifies this as a mechanism for artificially low variational energies in PEPS. It then develops a gauge-fixed optimization strategy that removes the exploitable direction, yielding more accurate contractions and more robust e

Load-bearing premise

The analysis rests on the claim that for a U(1)-symmetric PEPS with point group symmetry the gauge degrees of freedom reduce to a single class; if other independent gauge directions exist, the identified failure mode and the proposed fix could be incomplete.

Editorial extensions

If this is right

Reported variational energies from gradient-based PEPS optimizations of symmetric systems may systematically sit below the true ground-state energy, so such results need a gauge-aware check.
Gauge-fixed optimization should produce more accurate tensor-network contractions and energies that are more trustworthy as upper bounds.
Monitoring only the variational energy is insufficient; contraction accuracy or environment information must be tracked alongside it.
Gauge-aware optimization is likely required for reliable variational PEPS beyond the specific U(1) symmetric case studied here.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The failure mode probably generalizes to other continuous symmetries in PEPS, since gauge freedom is generic to tensor network representations and any symmetry can leave exploitable gauge directions.
Some existing PEPS ground-state energies in the literature may be artificially low for exactly this reason; re-running those optimizations with gauge fixing would provide a concrete test.
Comparing gauge-fixed and unfixed optimizations on the same system could serve as a practical diagnostic: a large energy gap between the two signals that the unfixed result is unreliable.

Original abstract

Projected entangled-pair states (PEPS) constitute a powerful variational ansatz for capturing ground state physics of two-dimensional quantum systems. However, accurately computing and minimizing the energy expectation value remains challenging, in part because the impact of the gauge degrees of freedom that are present in the tensor network representation is poorly understood. We analyze the role of gauge transformations for the case of a U(1)-symmetric PEPS with point group symmetry, thereby reducing the gauge degrees of freedom to a single class. We show how gradient-based optimization strategies exploit the gauge freedom, causing the tensor network contraction to become increasingly inaccurate and to produce artificially low variational energies. Furthermore, we develop a gauge-fixed optimization strategy that largely suppresses this effect, resulting in a more robust optimization. Our study underscores the need for gauge-aware optimization strategies to guarantee reliability of variational PEPS in general settings.

The compiled record

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 4,921 input tokens.

What the abstract itself states, as probabilities: claims first 0.04 · states limitation 0.28 · human participants 0.01 · controlled comparison 0.18 · randomized assignment 0.01 · states quantitative result 0.05 · compares with prior results 0.30 · mentions code or data availability 0.02.

The abstract, sentence by sentence

arXiv listing abstract v1. Results and the main claim are underlined; show every role.

Projected entangled-pair states (PEPS) constitute a powerful variational ansatz for capturing ground state physics of two-dimensional quantum systems. However, accurately computing and minimizing the energy expectation value remains challenging, in part because the impact of the gauge degrees of freedom that are present in the tensor network representation is poorly understood. We analyze the role of gauge transformations for the case of a U(1)-symmetric PEPS with point group symmetry, thereby reducing the gauge degrees of freedom to a single class. We show how gradient-based optimization strategies exploit the gauge freedom, causing the tensor network contraction to become increasingly inaccurate and to produce artificially low variational energies. Furthermore, we develop a gauge-fixed optimization strategy that largely suppresses this effect, resulting in a more robust optimization. Our study underscores the need for gauge-aware optimization strategies to guarantee reliability of variational PEPS in general settings.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#	sentence	role	the authors' own work	states the main conclusion	strength
s0	Projected entangled-pair states (PEPS) constitute a powerful variational ansatz for capturing ground state physics of two-dimensional quantum systems.	background 1.00	0.05	0.03	asserted 1.00
s1	However, accurately computing and minimizing the energy expectation value remains challenging, in part because the impact of the gauge degrees of freedom that are present in the tensor network representation is poorly understood.	gap 0.82	0.06	0.04	asserted 0.97
s2	We analyze the role of gauge transformations for the case of a U(1)-symmetric PEPS with point group symmetry, thereby reducing the gauge degrees of freedom to a single class.	method 0.65	0.93	0.05	asserted 0.99
s3	We show how gradient-based optimization strategies exploit the gauge freedom, causing the tensor network contraction to become increasingly inaccurate and to produce artificially low variational energies.	result 1.00	0.91	0.24	asserted 1.00
s4	Furthermore, we develop a gauge-fixed optimization strategy that largely suppresses this effect, resulting in a more robust optimization.	contribution 0.54	0.96	0.25	asserted 1.00
s5	Our study underscores the need for gauge-aware optimization strategies to guarantee reliability of variational PEPS in general settings.	interpretation 0.96	0.41	0.86	asserted 0.90

What kind of paper the classifier reports this is: method or tool 0.97 · evidence from simulation or computation 0.76.

What was examined, and what was not

sources read

arXiv listing abstract v1

not examined

full text
figures and tables
references and citing sentences
relations to other papers

levels run

P abstract

not run

S sections
T body sentences
C citations
R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

References

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Matrix Product Belief Propagation
quant-ph 2026-09 accept novelty 7.0 of 10

Matrix product belief propagation provides a controlled approximation for 2D tensor network contraction with quadratically reduced error and gauge invariance.
Unraveling square and kagome lattice frustrated magnetism by scalable tensor networks
cond-mat.str-el 2026-11 conditional novelty 6.0 of 10

High-precision infinite tensor-network simulations indicate a stable gapless spin liquid on the kagome Heisenberg lattice and plaquette valence-bond order with a gapless intermediate regime in the square-lattice J1-J2 model.

How the literature uses it

2 arXiv papers cite this in 3 sentences

Each citing sentence is taken from the citing paper's LaTeX source and typed by Pith's classifier. A label appears at probability 0.8 or more (0.95 for support), a contest at 0.7 or more; the quoted words are the citing authors'. How each label was checked · JSON

Papers citing it · 2 sentences without a confident label

Uses its method or tool 1 paper

When successive energy values are sufficiently converged — at least below $ΔE < 10^-7$, depending on the bond dimension — we re-gauge the iPEPS by performing trivial simple update steps [ref] to rule out that the variational energy was undershot due to the optimizer artificially exploiting virtual gauge degrees of freedom [ref] .

PEPSKit.jl: A Julia package for projected entangled-pair state simulations · Variational optimization for the square lattice J1-J2 model · 0.92

What it cites · 1 sentence without a confident label

Cites as background 1 paper

The class of projected entangled-pair states (PEPS) [ref] , a natural extension of matrix product states (MPS) to two dimensions, has proven powerful but remains significantly more challenging to handle numerically.

Renormalization algorithms for Quantum-Many Body Systems in two and higher dimensions · Introduction · 0.97

Discussion

Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.