{"id":"e1f8dd81-2d40-49d5-940f-868be013bef7","arxiv_id":"1908.05573","paper_version":3,"verdict":"UNVERDICTED","confidence":"HIGH","novelty_score":0.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"An expository overview of the definition of Liouville quantum gravity surfaces, the three senses in which random planar maps converge to them, and the major open problems.","lead":"This paper is an expository survey of Liouville quantum gravity (LQG), the canonical model of random fractal surfaces. It explains how random planar maps converge to LQG and lists the main open problems in the field.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; the survey's claims track proven theorems and its open-problem section explicitly qualifies the limits of the convergence statements.","rationale":"The reader classified this as an expository survey and assigned UNVERDICTED because it proves no new results. My stress-test pass did not uncover a load-bearing concern that would change that classification. The central claim is a synthesis of known results, and the paper is explicit about which convergence statements are proven and which remain open. The weakest point is the abstract's sweeping 'canonical models' language, but the body of the paper immediately qualifies it: Section 3.1 states that Gromov-Hausdorff convergence is only known for uniform planar maps, and Problem 2 lists the convergence for other models as a major open problem. This is not an internal inconsistency; it is appropriate scholarly framing. The cited external theorems are the natural locus of risk, but I have no concrete evidence of misstatement. A verification step comparing the survey's summaries with the published versions of [GM21] and [DMS14] is a worthwhile check and would settle the only real vulnerability. Since no specific error or omission was identified, the reader's UNVERDICTED verdict should remain unchanged.","tokens_in":12999,"tokens_out":3359,"duration_ms":34728,"concrete_test":"Pull the published versions of [GM21] and [DMS14] and verify the two key sentences in Sections 2.5 and 3.3: (i) [GM21] proves convergence in probability of the renormalized metrics and an axiomatic uniqueness result for all gamma in (0,2); (ii) [DMS14] constructs a gamma-LQG surface from correlated Brownian motion with correlation -cos(pi gamma^2 / 4). If either summary mismatches the source theorem, the survey's state-of-the-art report would need a correction; otherwise the central exposition stands.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I read this as an expository survey whose central claim is not a new theorem but a synthesis of the current state of the art. The three senses of convergence described in Section 3 are presented accurately: Gromov-Hausdorff convergence is restricted to uniform planar maps and sqrt(8/3)-LQG, embedding convergence is established only for specific families, and mating-of-trees convergence is the broadest but explicitly less geometric mode. The abstract's phrase 'continuum limits of discrete random surfaces' is a motivating summary rather than a theorem, and Section 3 plus Problem 2 carefully delineate what is and is not proven. The most load-bearing external citations, [GM21] for existence and uniqueness of the gamma-LQG metric and [DMS14] for the mating-of-trees construction, are correctly summarized and are published peer-reviewed results. I could not identify an internal inconsistency, a misstatement of a cited theorem, or a circular step that would undermine the survey's usefulness. The main vulnerability is generic to any survey: correctness depends on the accuracy of its citations, but nothing in the text suggests a mismatch with the cited sources.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This expository article introduces random planar maps and Liouville quantum gravity (LQG) surfaces, explains how the latter arise as continuum limits of random planar maps, and surveys the state of the art. Sections 1 and 2 define planar maps and the Gaussian free field and describe the gamma-LQG area measure and metric. Section 3 distinguishes three modes of convergence (Gromov-Hausdorff, embedding, and mating-of-trees) and records which random map models have been proved to converge in each mode. Section 4 lists open problems, chiefly the value of the Hausdorff dimension d_gamma for gamma not equal to sqrt(8/3), and the convergence of weighted random planar maps for general gamma. The paper is aimed at readers with roughly second-year graduate-level background and does not prove new theorems.","tokens_in":13347,"tokens_out":8664,"duration_ms":75855,"significance":"The paper is a concise and generally reliable survey of a large and technical subject. Its main value is pedagogical: it isolates three genuinely different notions of convergence to LQG and carefully qualifies what is known for each, which is exactly the kind of precision an introductory text needs. The statements about the LQG measure, the LQG metric, and the Brownian map equivalence are consistent with the cited literature, and the open-problem section draws a clear line between established results and conjectures. The author is also careful to attribute results to original sources and to mark open problems as open. If the survey's account is accurate, it will be a useful entry point for graduate students and non-experts.","major_comments":[],"minor_comments":[{"comment":"In the paragraph containing equation (2.4), the Hilbert space H(U) is described as the completion of smooth compactly supported functions on D, but the domain under discussion is U; this should be corrected to 'functions on U'.","section":"Section 2.2"},{"comment":"The word 'estalbished' in the sentence about Duplantier and Sheffield should be 'established'.","section":"Section 2.4"},{"comment":"The term 'baycentric embedding' should be 'barycentric embedding'.","section":"Section 3.2"},{"comment":"The first fundamental form is conventionally written E dx^2 + 2F dx dy + G dy^2; as written with F dx dy, the factor of 2 is omitted. Since the cross term disappears in isothermal coordinates, this does not affect the main argument, but the notation should either follow the standard convention or explicitly note the convention being used.","section":"Section 2.1"},{"comment":"The phrase 'the standard Gaussian random variable on H(U)' could be misread as implying the existence of a Gaussian measure on an infinite-dimensional Hilbert space; the intended meaning is the random distribution defined by the orthonormal series (2.5), and this could be stated more explicitly.","section":"Section 2.2"},{"comment":"Equation (2.7) uses the heat-kernel regularization h_epsilon, but the text does not specify the normalization convention that makes epsilon^{gamma^2/2} the correct factor. A sentence noting that the normalization is chosen so that E[h_epsilon(z)^2] ~ log(1/epsilon) would help readers verify the formula.","section":"Section 2.4"}],"recommendation":"minor_revision","confidential_remarks":"The paper is an expository survey with no new results, so the main risk is the usual one for surveys: the accuracy of its description of the state of the art depends on the cited theorems. I found no internal inconsistencies or misstatements of the cited results, and the open problems are properly qualified. The manuscript is appropriate for the target venue once the minor typographical and notational corrections above are made."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"I've read the Gwynne survey (arXiv:1908.05573). Quick take: it's an expository article, so don't expect new theorems. What it does well is give a genuinely clear and careful map of the LQG/random planar map convergence program as of 2019–2020. The three modes of convergence—Gromov–Hausdorff, embedding, mating-of-trees—are separated cleanly, and the text is disciplined about which mode is proven for which model. It is also honest about what is open: the value of d_gamma for gamma ≠ sqrt(8/3), and Gromov–Hausdorff convergence for non-uniform maps. For its stated audience (second-year graduate student) the level is right; it does not assume much background.\n\nThe soft spots are mostly intrinsic to a survey. There is no new content, so the novelty is zero by design. The author is a major contributor to the subject, and his own papers with Miller, Holden, Sun, and Sheffield dominate the bibliography. That is not a flaw here, because the cited results are real, published, and accurately described. I noticed a few compressed statements: the GFF is described as a 'standard Gaussian random variable' on an infinite-dimensional space, which could confuse a reader new to the theory, and the Hausdorff dimension of the LQG measure support is quoted without a precise citation. These are minor. The paper dates from 2019 (v3 2021), so the open problems section is slightly dated; some items, e.g., certain embedding convergences, have been addressed or partially addressed since.\n\nThe stress-test note matches my reading. I don't think there is a load-bearing flaw; the central claims are appropriately qualified in Section 3 and the open problems section. The abstract's language about 'canonical models' is clearly a motivational summary, not a theorem, and the text itself is careful about that.\n\nWho is this for? Anyone who wants a compact, reliable entry point into the LQG/random planar map literature, or needs to know exactly what has been proven in which convergence mode. I'd point a student to this before the primary sources. I would not typically cite it in my own technical papers—I would cite the originals—but I would recommend it without hesitation.\n\nRecommendation for an editor: yes, send this to peer review. A survey like this deserves a careful referee who can check the accuracy of the summary and the distinctions between convergence modes. The referee's job is to verify the citations, not to assess new mathematics.","headline":"A clear, honest survey of the LQG/random planar map convergence program; nothing new but a solid map of what is proven, worth a serious referee.","tokens_in":13689,"tokens_out":3871,"would_cite":false,"duration_ms":34652,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60D05","60G60","81T40"],"pacs":[],"model":"deepseek-v4-flash","headline":"Liouville quantum gravity surfaces are the canonical random two-dimensional geometries, the paper argues, and random planar maps converge to them in three distinct senses.","keywords":["Liouville quantum gravity","random planar maps","Gaussian free field","Brownian map","mating of trees","Gaussian multiplicative chaos","fractal surfaces","scaling limits"],"falsifier":"Run the renormalization construction for a fixed Gaussian free field and a fixed $\\gamma \\in (0,2)$ with two different mollifiers and check whether the rescaled metrics converge to the same limit; the uniqueness theorem says they must, so two distinct limits would refute the metric existence claim. Short of that, a precise simulation of the Hausdorff dimension for $\\gamma=\\sqrt{2}$ that lands outside the proven interval $3.550408$ to $3.63299$ would show the theory is internally inconsistent.","tokens_in":12810,"feed_emoji":"🌀","tokens_out":6814,"duration_ms":60685,"temperature":0.7,"pith_summary":"Random planar maps—graphs drawn in the plane with no crossing edges, viewed as discrete surfaces—should have a continuum limit as their number of edges grows. This paper argues that the correct universal limit is a Liouville quantum gravity (LQG) surface, and that LQG surfaces play the role for random surfaces that Brownian motion plays for random paths. The paper states this as a precise mathematical program and reviews the three senses in which convergence has been proved: as metric spaces, through embeddings into the plane, and through 'mating-of-trees' encodings by planar Brownian motion. A sympathetic reader comes away with the picture of a single family of canonical fractal random geometries, parametrized by a roughness parameter $\\gamma \\in (0,2)$, that organizes many discrete random surface models.","feed_headline":"LQG surfaces are the canonical random two-dimensional geometries","feed_subtitle":"An expository argument that Liouville quantum gravity is to random surfaces what Brownian motion is to random paths.","key_machinery":"The load-bearing object is the $\\gamma$-LQG surface, defined as a random distribution $h$ (the Gaussian free field) on a domain $U$, equipped with the formal Riemannian metric tensor $e^{\\gamma h}(dx^2+dy^2)$. The area measure is obtained by regularization: $\\varepsilon^{\\gamma^2/2} e^{\\gamma h_\\varepsilon} d^2z$ converges to a Gaussian multiplicative chaos measure, and the metric is obtained by a much harder renormalized 'Liouville first passage percolation' procedure using the exponent $\\gamma/d_\\gamma$, whose limit was proved to exist and be unique for all $\\gamma \\in (0,2)$. The mating-of-trees bijections, which encode decorated planar maps by random walks converging to correlated Brownian motion, are the bridge that connects discrete maps to the continuum surface.","core_discovery":"The central claim is that for every $\\gamma \\in (0,2)$, there is a unique random surface—the $\\gamma$-LQG surface, built from the two-dimensional Gaussian free field $h$ via the formal metric tensor $e^{\\gamma h}(dx^2+dy^2)$—that is the universal scaling limit of the associated family of random planar maps, in the same sense that Brownian motion is the universal scaling limit of random walks. The paper surveys the three proven modes of convergence: Gromov-Hausdorff convergence of uniform planar maps to the Brownian map, which is the $\\sqrt{8/3}$-LQG quantum sphere; convergence of embedded maps toward the $\\gamma$-LQG measure; and mating-of-trees convergence, in which a decorated planar map is encoded by a two-dimensional Brownian motion with correlation $-\\cos(\\pi \\gamma^2/4)$ and the continuum surface is reconstructed from that motion by a continuum mating-of-trees bijection. It also records the main open problem, the value of the Hausdorff dimension $d_\\gamma$ for $\\gamma \\neq \\sqrt{8/3}$.","pith_inferences":["If the universality claim is right, the three convergence modes should ultimately be compatible: the same $\\gamma$-LQG surface should arise as the Gromov-Hausdorff limit, the embedding limit, and the peanosphere limit for the same model; Problem 2 asks for exactly this for weighted maps, but the paper does not conjecture that they coincide as metric measure spaces.","A closed-form expression for $d_\\gamma$ would likely propagate to many observables, including the spectral dimension of random walk on random planar maps and the scaling of graph distances, since the paper lists these as expressible in terms of $d_\\gamma$.","The metric construction is two-dimensional in an essential way: the paper notes the associated metric has only been constructed in dimension 2, so a higher-dimensional canonical random geometry would need a new idea, possibly along the lines of the proposed higher-dimensional Brownian map analog."],"forward_implications":["Uniform triangulations and quadrangulations, with graph distance rescaled by $n^{-1/4}$ and vertex masses by $1/n$, converge in the Gromov-Hausdorff-Prokhorov sense to the Brownian map, identified with the $\\sqrt{8/3}$-LQG quantum sphere.","For every $\\gamma \\in (0,2)$, mated-CRT maps embedded by the Tutte embedding have vertex counting measure converging to the $\\gamma$-LQG measure; uniform triangulations embedded by the Cardy embedding converge to $\\sqrt{8/3}$-LQG in all three senses simultaneously.","The existence and uniqueness of the $\\gamma$-LQG metric implies that $d_\\gamma$—the Hausdorff dimension of the metric space—is a well-defined function of $\\gamma$, equal to $4$ at $\\gamma=\\sqrt{8/3}$, and strictly increasing with known bounds such as $3.550408 \\leq d_{\\sqrt{2}} \\leq 3.63299$.","Mating-of-trees convergence holds for many weighted models (spanning trees, percolation, bipolar orientations, Fortuin-Kasteleyn), giving scaling limits of functionals and exponents, even where metric-space convergence is not yet available."],"supporting_citations":[{"why":"Supplies the existence and uniqueness of the $\\gamma$-LQG metric for every $\\gamma \\in (0,2)$, the central analytic fact on which the metric space picture rests.","marker":"[GM21]"},{"why":"Identifies the Brownian map with a special $\\sqrt{8/3}$-LQG surface, the quantum sphere, thereby grounding Gromov-Hausdorff convergence in the LQG framework.","marker":"[MS20]"},{"why":"Provides the continuum mating-of-trees construction that builds a $\\gamma$-LQG surface from correlated planar Brownian motion.","marker":"[DMS14]"},{"why":"Establishes the $\\gamma$-LQG area measure via Gaussian multiplicative chaos and the KPZ formula, giving the measure side of the surface.","marker":"[DS11]"},{"why":"Proves tightness of the renormalized Liouville first passage percolation metrics, a key step toward the metric construction.","marker":"[DDDF19]"},{"why":"Shows uniqueness and universality of the Brownian map as the scaling limit of uniform planar maps.","marker":"[Le 13]"},{"why":"Proves that uniform random quadrangulations converge to the Brownian map in the Gromov-Hausdorff sense.","marker":"[Mie13]"},{"why":"Establishes the Gaussian free field as a random distribution, the basic random input for LQG surfaces.","marker":"[She07]"},{"why":"Proves embedding convergence for mated-CRT maps under the Tutte embedding for every $\\gamma \\in (0,2)$.","marker":"[GMS17]"},{"why":"Proves Cardy embedding convergence for uniform triangulations simultaneously in the metric and peanosphere senses.","marker":"[HS19]"}],"fun_headline_variants":["LQG: the Brownian motion of random surfaces","The canonical random surface: LQG","From planar maps to Liouville quantum gravity","LQG: the universal geometry of random surfaces","A graduate-friendly tour of LQG surfaces"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The survey stands or falls with the correctness of the external theorems it cites, above all the existence and uniqueness of the random metric for every roughness parameter in (0,2), the identification of the uniform-map limit with the roughness $\\sqrt{8/3}$ case, and the encoding of the continuum surfaces by correlated Brownian motion.","fun_headline_variants_meta":{"raw":{"variants":["LQG: the Brownian motion of random surfaces","The canonical random surface: LQG","From planar maps to Liouville quantum gravity","LQG: the universal geometry of random surfaces","A graduate-friendly tour of LQG surfaces"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000643,"raw_usage":{"total_tokens":2918,"prompt_tokens":870,"completion_tokens":2048,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":486,"completion_tokens_details":{"reasoning_tokens":1981}},"tokens_in":486,"tokens_out":2048,"duration_ms":14485,"temperature":1.0,"reasoning_tokens":1981,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:08:06.379033+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the renormalization construction for a fixed Gaussian free field and a fixed $\\gamma \\in (0,2)$ with two different mollifiers and check whether the rescaled metrics converge to the same limit; the uniqueness theorem says they must, so two distinct limits would refute the metric existence claim. Short of that, a precise simulation of the Hausdorff dimension for $\\gamma=\\sqrt{2}$ that lands outside the proven interval $3.550408$ to $3.63299$ would show the theory is internally inconsistent.","supporting_citations":[],"review_version":1}