REVIEW 3 major objections 4 minor 1 cited by
The Liouville model in the $L^1$ phase: coupling and extreme values
T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash
Pith's one-line read On the two-dimensional torus, in the subcritical phase, the Liouville field can be coupled to the Gaussian free field so that their difference is a Hölder continuous function.
desk verdict Claims a major LQG–GFF coupling with Hölder difference, but the only proof text I have is unreadable; the sign/FKG control of the Polchinski flow is the load-bearing premise that needs referee scrutiny. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The machinery is a renormalisation group flow that interpolates between the short-distance (ultraviolet) and large-distance (infrared) scales of the field. The flow is run for the Liouville model, and the same flow is run for the Gaussian free field; comparing the two flows at every scale yields a bound on the difference field. The two properties that make the comparison work are that the Liouville flow has a definite sign, so the difference between flows does not oscillate unpredictably, and that a correlation inequality for positively associated random fields keeps correlations under control as the cutoff is removed.
What would settle it
Simulate both fields on a fine grid on the torus with the same underlying noise, and measure the empirical moments of their difference as the grid is refined. If the variance of the difference grows like the logarithm of the number of points rather than staying bounded, then the claimed Hölder coupling fails; conversely, bounded differences at all scales support the theorem.
Extended reading notes
Core claim
The central claim is a strong-coupling theorem. For any inverse temperature in the L1 phase, there exists a coupling of the Liouville field and the Gaussian free field on the torus such that their difference is Hölder continuous with some positive exponent. The theorem is constructive: the Liouville field is obtained by running a renormalisation group flow from short to large scales, and at every step the flow can be compared with the Gaussian flow because the Liouville flow's sign and its monotonicity are known. The comparison survives the removal of the ultraviolet cutoff, giving the Hölder bound. As a direct corollary, the global maximum of the Liouville field converges in distribution to
Load-bearing premise
The proof rests on a sign property of the renormalisation flow and a correlation inequality; if either fails at any scale in the continuum limit, the coupling and Gumbel maximum claims collapse.
Editorial extensions
If this is right
- The Liouville field inherits the full extreme-value statistics of the Gaussian free field in the entire subcritical phase, not just at special parameter values.
- The global maximum of the Liouville field converges to a randomly shifted Gumbel law, so predictions that relied on that law for Gaussian fields transfer to the Liouville model.
- Quantitative field-level comparison via the renormalisation group gives a way to compute finite-scale corrections to the maximum distribution.
- The construction extends the renormalisation-group coupling method, previously applied to other two-dimensional Euclidean field theories, to the Liouville model.
Reading between the lines
- If the coupling is as strong as claimed, other geometric features of the Liouville field—such as level sets, thick points, and spatial correlations of high peaks—should match the Gaussian free field up to the same Hölder error; this follows in spirit but is not established in the paper.
- The random shift in the Gumbel law is presumably determined by the large-scale average of the coupling; identifying it explicitly for the torus could give a universal constant for the Liouville maximum.
- The same coupling strategy may extend to other compact surfaces or to the massive Liouville model, where the large-scale Gaussian comparison would require a mass parameter.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper claims a strong coupling between the Liouville model and the Gaussian free field on the two-dimensional torus for all β∈(0,8π), in the sense that the difference of the two fields is a Hölder continuous function. The stated mechanism is a Polchinski renormalisation group flow with two key observations: the flow has a definite sign and can be controlled by an FKG argument. From this coupling the authors derive that the global maximum of the Liouville field converges in distribution to a randomly shifted Gumbel distribution, transferring the known extreme-value structure of the GFF. The abstract states these results, but the supplied full text is corrupted and unreadable, so the actual derivations, lemmas, and estimates cannot be inspected.
Significance. If correct, the result is a major advance for the Liouville model in the entire L1 phase: a field-level comparison with the GFF via a quantitative Polchinski RG construction, yielding the full extreme-value transfer. The claimed coupling is stronger than distributional convergence of the maximum and would provide a concrete probabilistic bridge between the two fields. The use of a sign property and FKG control for the Polchinski flow is an interesting and potentially powerful technique. However, the significance is conditional on the unverifiable proof content in the submitted text; I cannot currently certify the mathematical claims.
major comments (3)
- [Abstract] The central coupling theorem rests entirely on two assertions: 'the Polchinski flow has a definite sign' and 'can be controlled well thanks to an FKG argument.' No precise statement of either property appears in the abstract, no inequalities or lemmas are given, and no domain of validity in β is specified beyond the interval. Since the supplied full text is unreadable, I cannot verify that the sign is preserved at every scale of the flow or that the FKG control survives removal of the regularization. These are load-bearing premises: if either fails, the Hölder coupling and the Gumbel maximum corollary do not follow. The manuscript must state and prove the sign and FKG estimates explicitly, including their dependence on β and the treatment of the β→8π endpoint.
- [Full text] The submitted full text is corrupted: it consists of unreadable mojibake and contains an unrelated cross-reference to arXiv:2508.15679v1 [cs.LG]. It is impossible to locate the main theorem, the Polchinski flow equation, the sign lemma, the FKG estimate, or the proof of Hölder regularity. This is not a minor typographical issue; it blocks verification of every substantive claim. A complete, readable manuscript is required before the results can be evaluated.
- [Abstract (extreme values)] The extreme-value application asserts convergence of the global maximum of the Liouville field to a randomly shifted Gumbel distribution, but the statement is incomplete. The abstract does not specify the normalization of the Liouville field (e.g., subtraction of expectation, rescaling), the random shift's distribution or its dependence on β, or the precise sense of convergence (in distribution with respect to what probability space). These details are needed to make the theorem falsifiable and comparable with the known GFF extremes.
minor comments (4)
- [Abstract] Typo: 'distribtion' should be 'distribution'.
- [Abstract] The phrase 'L1 phase' is used without definition; the interval β∈(0,8π) is the subcritical phase, but the terminology should be explained or referenced.
- [Full text] The full text includes a cross-reference to a cs.LG paper that is unrelated to the Liouville model; this appears to be an accidental artifact and should be removed.
- [Abstract] The Hölder exponent of the difference field is not stated; a theorem of this strength should specify the exponent or at least the class of regularity.
Circularity Check
No circularity identified: the coupling is derived from Polchinski RG and FKG controls, and the Gumbel maximum is transferred from known GFF results.
full rationale
The abstract's derivation chain is: (i) construct a strong coupling between the Liouville model and the GFF via a Polchinski renormalisation group flow; (ii) control that flow using a 'definite sign' and an FKG argument; (iii) transfer known extreme-value results for the GFF to the Liouville field through the coupling. None of these steps reduces to the conclusion by definition. The 'definite sign' and 'FKG argument' are stated as technical controls on the RG flow, not as renamed versions of the target coupling or the Gumbel limit. The random shift in the Gumbel distribution is not fitted from Liouville data; it is a consequence of the coupling with the GFF. The paper cites prior use of the same RG approach for other two-dimensional field theories, but the abstract does not make that citation load-bearing in a way that can be shown to be self-referential. The supplied full text is mostly unreadable mojibake and contains an unrelated cs.LG arXiv header, so the proof details cannot be inspected; this is an evidentiary limitation, not evidence of circularity. The unverified sign/FKG control is a correctness risk, but without an exhibited reduction of a claimed prediction to an input, it does not constitute circularity. Accordingly the honest finding is no significant circularity.
Assumptions & free parameters
assumptions (3)
- domain assumption The Polchinski renormalisation group flow for the Liouville model is sign-definite and can be controlled by an FKG argument.
- domain assumption Existence of the unregularised Liouville field and its renormalised measure on the torus in the subcritical phase beta in (0, 8 pi) is taken as given.
- standard math The Gaussian free field on the torus has a maximum converging to a randomly shifted Gumbel distribution.
Cite this review
Pith. "Pith review of The Liouville model in the $L^1$ phase: coupling and extreme values." pith.science (2026). https://pith.science/paper/QHWLNHKI
@misc{pith2026250815689,
author = {Pith},
title = {Pith review of: The Liouville model in the $L^1$ phase: coupling and extreme values},
year = {2026},
howpublished = {\url{https://pith.science/paper/QHWLNHKI}},
note = {Machine review of arXiv:2508.15689}
}
abstract
We establish a strong coupling between the Liouville model and the Gaussian free field on the two dimensional torus in the $L^1$ phase $\beta \in (0, 8\pi)$, such that the difference of the two fields is a H\"older continuous function. The coupling originates from a Polchinski renormalisation group approach, which was previously used to prove analogous results for other Euclidean field theories in dimension two. Our main observations for the Liouville model are that the Polchinski flow has a definite sign and can be controlled well thanks to an FKG argument. The coupling allows to relate extreme values of the Liouville model and the Gaussian free field, and as an application we show that the global maximum of the Liouville field converges in distribution to a randomly shifted Gumbel distribtion.
Forward citations
Cited by 1 Pith paper
-
Extrema of cooling branching Brownian motion and related Gaussian fields
Introduces slowed-down Gaussian fields (including 1D branching Brownian motions in cooling environments) and proves tightness of maxima with growth T^{1-α} and phase transition at α=1/3.
Reference graph
Works this paper leans on
-
[1]
� � � �������� � ��� � ���� � ���������� ��� � ���� �� ���� � ����� � ������� � � ������� ���� ���� ��� ����� ������� ������� ����� �� ����� ������ �� �������� ������� ��� ������������ ���������� �� ���������� ��������������������������������������� ��������� �� ���� � ������� ���� ���������� ������ ������ �� ������ ��� �� �������� �� ���������� ���������...
arXiv 2025
Reviewed August 5, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.