{"id":"4d2ad44c-8717-4a64-a77f-e6959a63d39e","arxiv_id":"2601.14909","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For random ideal hyperbolic polyhedra, the mean dihedral-angle sum around a face determines the conformal type, recurrence, and boundary behavior of the associated random walk.","lead":"This paper studies random infinite polyhedra whose faces sit at infinity of hyperbolic 3-space, together with their dual 'angle graphs'. It proves that a single per-face number — the sum of surrounding dihedral angles — decides whether the structure is parabolic or hyperbolic, which in turn decides whether a random walk on the graph returns or escapes to the boundary, and how fast.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 4.5's Birkhoff step requires a third-moment integrability condition not implied by E[deg^2(ρ)]<∞; Theorems 1.4/1.5 inherit this gap.","rationale":"The reader's weakest_assumption was the refined ring lemma, but the more precisely locatable gap is the integrability of S along the random walk. Lemma 4.5 and Theorem 5.1 both depend on the Birkhoff average of S(X_i) under the reversible measure; the stated E[deg^2(ρ)]<∞ only controls the unimodular expectation of S, not the degree-biased expectation E[deg(ρ)S(ρ)]/E[deg(ρ)]. This is a concrete, internal gap rather than a merely compressed proof. It does not refute the dichotomy theorem, which appears reducible to AHNR, so the verdict remains CONDITIONAL. The refined ring lemma is also under-verified, but the moment gap is the single most decisive obstruction to the boundary/speed conclusions as written.","tokens_in":34515,"tokens_out":19841,"duration_ms":195841,"concrete_test":"Take the degree-biased reversible measure of Lemma 5.2 and compute E_rev[S(ρ)] = E[deg(ρ)S(ρ)]/E[deg(ρ)] for a tame IAG with a degree distribution D satisfying E[D^2]<∞ but E[D^3]=∞ (or construct the nearest heavy-tailed example in the IAG class). If E_rev[S(ρ)]=∞, Lemma 4.5's cancellation argument fails and Theorems 1.4/1.5 require the additional assumption E[deg(ρ)S(ρ)]<∞. A minimal analytical check is to verify whether the sequence S(X_i) is integrable under the reversible measure from the stated hypotheses; if not, the proof must be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4.2's exponential-decay transfer depends on showing (1/n) Σ_{i=N_m(n)}^{n-1} S(X_i) → 0. The proof invokes Birkhoff on S(X_i) under the reversible/stationary measure, but the stated hypothesis is E[deg^2(ρ)]<∞ on the original unimodular measure. Passing to the reversible measure via degree-biasing (Lemma 5.2) makes the relevant mean E_rev[S(ρ)] = E[deg(ρ)S(ρ)] / E[deg(ρ)], and E[deg(ρ)S(ρ)] is not bounded by E[deg^2(ρ)]: it is a cross-moment involving deg(ρ) times the sum of neighbour degrees and can diverge while E[deg^2]<∞ (e.g., heavy-tailed unimodular planar graphs with E[D^3]=∞). Lemma 4.9 proves E[S(ρ)]=E[deg^2(ρ)] only under the unimodular measure, not the reversible one, so it does not supply the needed integrability. The same missing condition appears in Theorem 5.1 via E_rev[|log R_1|] ≤ A E_rev[S(ρ)]. Thus the proof as written has an internal gap; the claimed almost-sure exponential decay and positive speed require either an additional finite third-moment-type assumption (E[deg(ρ)S(ρ)]<∞) or a different argument that does not rely on the stationary mean of S being finite.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a boundary and dichotomy theory for random infinite ideal angled graphs (IAGs), i.e., weighted planar graphs (G,Θ) arising as dual 1-skeleta of ideal hyperbolic polyhedra, and equivalently for ideal circle packings (ICPs). The main claims are: (1) for an ergodic unimodular random tame IAG with finite expected degree, the expected angle character E[T(ρ)] is always at least 2π, with equality characterizing invariant amenability, VEL-parabolicity, ICP-parabolicity, and recurrence, and strict inequality characterizing the hyperbolic/transient side (Theorem 1.2 and Corollary 1.3); (2) in the ICP-hyperbolic case under E[deg²(ρ)]<∞ and an angle upper bound, the simple random walk converges a.s. to a point on ∂D, the exit measure is non-atomic with full support, ∂D realizes the Poisson boundary, and the hyperbolic speed is positive and equals the exponential decay rate of circle radii (Theorems 1.4 and 1.5). The proofs use a mass-transport identity for E[T]=E[θ], a geometric dichotomy based on the AHNR framework, a refined ring lemma for ICPs, and an induced-walk argument with exponential decay of radii.","tokens_in":34756,"tokens_out":10859,"duration_ms":112318,"significance":"If the main theorems are correct, the paper would substantially extend the Angel–Hutchcroft–Nachmias–Ray theory from triangulations and general planar maps to angle-weighted ideal circle packings and ideal hyperbolic polyhedra, providing a single geometric invariant — the expected angle character — that detects amenability, conformal type, recurrence/transience, and boundary identification. The conceptual framework is attractive, and the mass-transport proof of Theorem 1.1 is explicit and honest: E[T(ρ)]=E[θ(ρ)] is derived cleanly from a concrete unimodular transport. The paper also identifies the refined ring lemma as the key bridge from local geometry to global random-walk behavior, which is the right structural insight. However, as it stands the central proof contains several load-bearing gaps: the dichotomy proof is partly self-referential and leaves the angle identities unchecked; the counting lemma used for exponential decay is false as stated for arbitrary disc families; and the Birkhoff step in the exponential-decay and speed theorems requires a moment control that is not provided by E[deg²(ρ)]<∞. These issues touch Theorems 1.2, 1.4, and 1.5 directly, so the current manusc","major_comments":[{"comment":"The proof of Theorem 1.2 opens with ‘by The Dichotomy Theorem 1.2’, which is a self-reference unless an external theorem from [4,5] is intended; this must be stated explicitly and its hypotheses checked. More importantly, Eqs. (10)–(13) assert that the mass transport F has outgoing total 4π and incoming total 2T(ρ) in the parabolic case and <2T(ρ) in the hyperbolic case. These identities are not derived. In particular, the notation f_i=(v_i,ρ,v_{i+1}) in (11) appears to assume that the faces incident to ρ are triangles, whereas IAGs are defined for arbitrary cellular decompositions. This is the central step establishing E[T(ρ)]=2π iff ICP-parabolic; it needs a complete angle bookkeeping for faces of arbitrary length, or a precise reduction to [5].","section":"§3.2, Eqs. (10)–(13)"},{"comment":"Lemma 4.8 is false as stated for arbitrary families of discs in D: taking countably many discs with centers in a small neighbourhood and radii 1/2 gives N(τ)=∞ while Cτ^{-2} is finite. The proof claims a decomposition into at most Q pairwise-disjoint subfamilies using Besicovitch covering; Besicovitch provides a bounded number of covering subfamilies, not a partition of the index set. The subsequent bound #{v: r(v)≥e^{-cm/4}} ≤ C e^{cm/2} and hence the Borel–Cantelli step in (21) depend on this lemma. A correct counting argument must use the specific geometric constraints of an embedded ICP, e.g. disjointness of the associated quadrilaterals or a separation property following from planarity, rather than an arbitrary disc-family estimate.","section":"§4.2, Lemma 4.8"},{"comment":"The proof applies Birkhoff’s ergodic theorem to S(X_i) under the reversible/stationary law. It cites Lemma 4.9 for E[S(ρ)]<∞, but Lemma 4.9 computes the expectation under the original unimodular measure. Under the degree-biased reversible measure, E_rev[S(ρ)] = E[deg(ρ)S(ρ)] / E[deg(ρ)], and E[deg(ρ)S(ρ)] is not bounded by E[deg²(ρ)]; it is a cross-moment involving deg(ρ) times the sum of neighbour degrees. Thus the claimed a.s. vanishing in (29) is not justified. The same missing integrability appears in Theorem 5.1, where E_rev[|log R_1|] ≤ A E_rev[S(ρ)] is asserted. The exponential-decay and positive-speed theorems therefore require an additional hypothesis such as E[deg(ρ)S(ρ)]<∞ (or E[deg³(ρ)]<∞), or a separate argument that avoids the stationary mean of S being finite.","section":"§4.2, after Eq. (27), and §5, Eq. (37)"},{"comment":"In the argument that supp(μ)=∂D, the proof chooses an arc (θ,ψ) disjoint from supp μ, defines A as the vertices whose hyperbolic centers lie in the sector S(θ,ψ), and then asserts ‘Since circles accumulate at the boundary inside this sector, there are infinitely many such vertices u’. This is precisely the kind of statement that needs proof: supp μ being a proper subset of ∂D does not by itself imply that the circle centers accumulate at every point of that arc. If the packing has a boundary gap in that sector, the set A may be finite and the mass-transport contradiction does not follow. The step needs either a proof that the accumulation set of the ICP has no open gaps, or a different argument for full support.","section":"§4.3, full-support proof"},{"comment":"The proof of the refined ring lemma is too compressed. Lemma 4.2 is stated with a one-line proof, and Lemma 4.4’s iterative contradiction is under-bookkept, particularly the passage ‘This process continues until … which contradicts Lemma 2’. Since Theorem 4.3 is the bridge transferring exponential decay from the induced walk to the original walk in Eq. (26), and is also used in Theorem 5.1 via Eq. (36), this estimate is load-bearing for the boundary and speed theorems. The exponential dependence on S(u) — the main place where unbounded degrees are controlled — needs a complete and self-contained proof.","section":"§4.1, Theorem 4.3 (Refined Ring Lemma)"}],"minor_comments":[{"comment":"Typo ‘ince’ in ‘ince these angles are independent’. More substantially, the notation IAG+ in Definition 2.4 is introduced without a separate displayed definition of the space IAG+.","section":"§3.2"},{"comment":"The angle condition is written ‘Θ2 (0,π )E’ and ‘Θ2 (0,π−ε]E’ with missing superscripts; these should be Θ ∈ (0,π)^E and Θ ∈ (0,π−ε]^E.","section":"§2.1, Definition 2.1 and 2.4"},{"comment":"The notation E♮ in Eq. (37) is used before the reversible measure has been consistently named; elsewhere it is E_rev. Please unify the notation.","section":"§5"},{"comment":"The Gromov/Martin boundary identification relies on the unpublished/preparation reference [26]. If these statements are part of the main contribution, they should either be stated as assumptions in the theorems or proved here; otherwise the dependence should be clearly flagged as external.","section":"§1.3, Remark after Theorem 1.5"},{"comment":"The abstract states the character dichotomy without the hypotheses of Theorem 1.2 (tameness, ergodicity, E[deg(ρ)]<∞). It would help readers to include these qualifiers in the abstract.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The manuscript has a strong conceptual framework and the main results are plausible, but several central arguments are incomplete. The most concerning issue is Lemma 4.8, whose statement is false for arbitrary disc families; even if the intended ICP context saves the counting estimate, the proof as written does not. The missing third-moment/cross-moment control in the Birkhoff step is a concrete gap in Theorems 1.4 and 1.5, and the dichotomy proof in §3.2 is not self-contained. I would recommend major revision rather than rejection, because the gaps appear fixable by adding hypotheses, proving the counting lemma under ICP-specific separation conditions, and expanding the angle bookkeeping. The editor may also wish to verify the availability of the cited in-preparation works [26] and [40], since the paper’s boundary-theory claims depend on them."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my read. The paper's real contributions are the angle character T, the identity E[T]=E[θ] via mass transport, and the extension of boundary convergence results from triangulations to angle-weighted cellular decompositions with unbounded degree. The refined ring lemma (Thm 2.9) is the right kind of geometric estimate, and if it holds the rest of the architecture should work. I believe the dichotomy statement itself is likely correct, because modulo the angle identities it reduces to the published AHNR dichotomy [5].\n\nThe problems are concentrated in two places. First, the proof of Thm 1.2 literally invokes 'The Dichotomy Theorem 1.2' where a citation to AHNR [5] is needed, and the angle identities (10)–(13) are asserted without derivation. That is fixable, but as written it is circular in form. Second, Lemma 4.5 and Theorem 5.1 have a genuine integrability gap. The proof applies Birkhoff's theorem to S(X_i) under the reversible measure, but the stated hypothesis E[deg^2(ρ)] < ∞ only gives E[S(ρ)] < ∞ under the unimodular measure. Passing to the reversible measure via degree biasing requires E[deg(ρ)S(ρ)] < ∞, which is a third-moment-type condition and is not implied by E[deg^2] < ∞. So the exponential decay of radii along the full walk and the positive-speed theorem are not established as written.\n\nThere are smaller issues: the abstract claims Martin boundary identification, but the body only proves Poisson boundary; the Martin result depends on an in-preparation paper [26]. The paper also leans heavily on same-group preprints, which makes verification harder.\n\nIf the moment condition is added and the ring lemma proof is written out, I think the main theorems are likely true and the paper would be a substantial contribution. As it stands, it's not ready. I'd send it to a serious referee, and I'd tell the referee to focus on the ring lemma and the Birkhoff step.","headline":"Serious extension of AHNR to ideal circle packings with a nice new invariant, but two proofs have load-bearing gaps that need fixing before this is citable.","tokens_in":35396,"tokens_out":3847,"would_cite":false,"duration_ms":39922,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60K37","52C26","51M10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The expected sum of dihedral angles around a vertex decides whether a random infinite ideal hyperbolic polyhedron is parabolic or hyperbolic — and in the hyperbolic case the random walk escapes to a well-defined boundary at positive speed.","keywords":["ideal angled graphs","ideal hyperbolic polyhedra","ideal circle packings","unimodular random graphs","dichotomy theorem","Poisson boundary","random walk","ring lemma"],"falsifier":"Compute, for a tame IAG with embedded ICP and angles bounded away from π, the ratio r(v)/r(u) along an edge where the flower degree S(u) is large; the Refined Ring Lemma predicts log(r(v)/r(u)) ≥ −C·S(u) with C depending only on the angle bounds. A concrete example with log(r(v)/r(u)) < −C·S(u) for a fixed C would falsify the lemma and, with it, the positive-speed and boundary-convergence theorems. Alternatively, an ergodic unimodular IAG with E[T(ρ)]=2π that is transient, or E[T(ρ)]>2π that is recurrent, would falsify the dichotomy.","tokens_in":34157,"feed_emoji":"⭕","tokens_out":7231,"duration_ms":60803,"temperature":0.7,"pith_summary":"This paper establishes a dichotomy for random infinite ideal polyhedra in hyperbolic 3-space, equivalently for random ideal circle packings via their ideal angled graphs. For an ergodic unimodular random IAG with finite expected degree, the expected angle character E[T(ρ)] = E[Σ_{e∋ρ} Θ_e] is always at least 2π; it equals 2π exactly when the structure is almost surely parabolic — invariantly amenable, recurrent, admitting a circle packing in the plane — and exceeds 2π exactly when it is almost surely hyperbolic — invariantly non-amenable, transient, admitting a packing in the unit disk. In the hyperbolic case the simple random walk converges almost surely to a boundary point of the unit disk, the exit measure is non-atomic with full support, the disk boundary realizes the Poisson boundary, and the walk escapes with positive linear hyperbolic speed equal to the exponential decay rate of circle radii. The result extends the known dichotomy for planar triangulations to the much larger class of angle-weighted planar graphs by proving a refined ring lemma for ideal circle packings.","feed_headline":"One angle sum decides random ideal polyhedra's fate","feed_subtitle":"Expected angle character E[T]=2π marks parabolic/recurrent ideals; E[T]>2π marks hyperbolic/transient ones.","key_machinery":"The central object is the angle character T(ρ) := Σ_{e∋ρ} Θ_e, the sum of dihedral angles of the ideal polyhedron at the face corresponding to the root vertex, which plays the role that combinatorial curvature plays for triangulations. The paper's main new estimate is a Refined Ring Lemma: for a tame IAG with angles in [0, π−ε1] and the strengthened Rivin condition (C′2), the radius ratio r(v)/r(u) across any edge u~v is at least e^{−C·S(u)}, where S(u)=Σ_{v′∼u} deg(v′) is the flower degree and C depends only on the angle bounds. This exponential local control of radii is what converts exponential decay of radii along the induced walk on an invariant percolation cluster into exponential deca","core_discovery":"The central claim is that for unimodular random ideal angled graphs the expected angle character E[T(ρ)] is a complete invariant determining the ideal-circle-packing type: E[T(ρ)] = 2π if and only if the graph is almost surely ICP-parabolic (equivalently invariantly amenable, VEL-parabolic, recurrent), and E[T(ρ)] > 2π if and only if it is almost surely ICP-hyperbolic (equivalently invariantly non-amenable, VEL-hyperbolic, transient). In the hyperbolic case, assuming a finite second moment of degree and angles uniformly bounded away from π, the Euclidean and hyperbolic circle centers z(X_n) and z_h(X_n) converge almost surely to a common boundary point δ on ∂D; the exit measure is non-atomic","pith_inferences":["If the dichotomy is correct, the expected angle character E[T(ρ)] becomes a computable invariant for classifying random planar angle-weighted graphs as hyperbolic or parabolic, and could serve as a geometric order parameter for phase transitions in random polyhedral models.","The refined ring lemma suggests a general phenomenon for circle packings with prescribed intersection angles: local radius ratios are exponentially controlled by local degree structure; this may extend to non-ideal circle patterns or circle patterns on surfaces with cone singularities.","The authors conjecture that E[deg(ρ)] < ∞ (or a higher moment) suffices for Martin-boundary identification; if true, the bounded-degree hypothesis in the boundary theory could be relaxed substantially.","The equality between hyperbolic speed and logarithmic radius decay gives a new way to measure the asymptotic geometry of random ideal polyhedra from the circle packing alone, and might be testable numerically on concrete packings."],"forward_implications":["For any ergodic unimodular random IAG with finite expected degree, the single value E[T(ρ)] decides whether the associated random ideal polyhedron is parabolic (planar, recurrent, amenable) or hyperbolic (disk, transient, non-amenable).","In the hyperbolic case, the random walk converges almost surely to a point on the unit circle, so the circle boundary gives a concrete tail description of the walk.","The exit measure on the boundary has full support and no atoms, so every open arc is hit with positive probability and no single boundary point carries positive mass.","The walk's hyperbolic displacement grows linearly at a positive rate, and this rate is exactly the exponential decay rate of the radii of the visited circles.","The dichotomy and boundary coincidences extend to all tame ideal angled graphs, not just triangulations, so the same statements hold for the corresponding ideal hyperbolic polyhedra."],"fun_headline_variants":["Angle sum expectation decides ideal polyhedra fate","One number separates parabolic and hyperbolic ideals","E[T]=2π marks parabolic random ideal graphs","Ideal angled graphs: expected angle T tells all","Random IHP type fixed by expected angle character"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is the Refined Ring Lemma, whose claim is that in a tame ideal circle packing with angles uniformly bounded away from π, the radius of a neighboring circle can shrink by at most an exponential factor in the vertex's flower degree; the paper's proof of this lemma is the most compressed step, and if this radius-ratio bound fails, the exponential decay of radii, the boundary convergence, and the positive-speed theorem lose their engine.","fun_headline_variants_meta":{"raw":{"variants":["Angle sum expectation decides ideal polyhedra fate","One number separates parabolic and hyperbolic ideals","E[T]=2π marks parabolic random ideal graphs","Ideal angled graphs: expected angle T tells all","Random IHP type fixed by expected angle character"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000699,"raw_usage":{"total_tokens":3073,"prompt_tokens":904,"completion_tokens":2169,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":648,"completion_tokens_details":{"reasoning_tokens":2098}},"tokens_in":648,"tokens_out":2169,"duration_ms":22613,"temperature":1.0,"reasoning_tokens":2098,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T09:08:24.727513+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, for a tame IAG with embedded ICP and angles bounded away from π, the ratio r(v)/r(u) along an edge where the flower degree S(u) is large; the Refined Ring Lemma predicts log(r(v)/r(u)) ≥ −C·S(u) with C depending only on the angle bounds. A concrete example with log(r(v)/r(u)) < −C·S(u) for a fixed C would falsify the lemma and, with it, the positive-speed and boundary-convergence theorems. Alternatively, an ergodic unimodular IAG with E[T(ρ)]=2π that is transient, or E[T(ρ)]>2π that is recurrent, would falsify the dichotomy.","supporting_citations":[],"review_version":1}