{"id":"d6760fb3-b232-4b5a-b356-3ccd578ec270","arxiv_id":"2607.15522","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Connected bounded-degree graphs satisfying CD(0,∞) for the unnormalised Laplacian are volume doubling and satisfy scale-invariant L² Poincaré inequalities with dilation two.","lead":"This paper proves that bounded-degree graphs with nonnegative Bakry–Émery curvature grow polynomially and satisfy scale-invariant Poincaré inequalities, settling a conjecture in graph curvature. It gives the first dimension-free derivation of these global geometric properties from local curvature.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Displayed semigroup estimate (27) has wrong Gamma-scaling; Theorems 1.1 and 1.3 rely on the corrected version, so the written proof is incomplete.","rationale":"I read the paper in good faith and checked the main architecture: point-mass identities, resolvent smoothing, modified Li–Yau, Harnack, volume doubling, censored-kernel localisation, and the Whitney-type reduction to dilation two. The proof is coherent and I found no fatal gap in the geometric argument. The single most load-bearing issue I found is the incorrect statement and proof of the semigroup estimate (27). It is not merely cosmetic: (67) in Theorem 1.1 and the displacement bound in Lemma 6.2 are explicitly derived from (27), and the stated version cannot yield those bounds. However, the correct inequality follows immediately from (26) with a √||Γ(f)||∞ factor, and all subsequent applications use exactly that corrected form. So the central claim appears true and the proof is repairable. The reader's conditional verdict was based on the unverifiable private-preprint footnote in §1.4; that is a legitimate but non-central issue. My concern is different, hence disagreement with the reader's identified weakest assumption, but I do not move the verdict because the paper should already be conditional pending these fixes.","tokens_in":841,"tokens_out":770,"duration_ms":325337,"concrete_test":"Recompute (27) from (26): write ||ΔP_s f||∞ ≤ sqrt(N/(2s))·sqrt(||Γ(f)||∞), integrate from 0 to t, and compare with the displayed text. If the correct value is sqrt(2N t ||Γ(f)||∞), the displayed (27) is false. Then substitute u0 from (65) into the corrected inequality and verify that (67) is exactly C√(N d* t)/r; the concern is settled if (67) does not follow from the displayed (27).","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3's Laplacian-smoothing estimate (27) is stated and proved with the wrong scaling. From (26), ||ΔP_s f||∞ ≤ sqrt(N/(2s) · ||Γ(f)||∞). Integrating gives ||P_t f − f||∞ ≤ sqrt(2N t ||Γ(f)||∞). The paper instead writes sqrt(2N t) ||Γ(f)||∞, and its proof line integrates sqrt(N/(2s)) ||Γ(f)||∞, treating ||Γ(f)||∞ as though it had been pulled outside the square root. The displayed inequality is dimensionally inconsistent: the left side has units of f, while the RHS as written has units of f² (times a dimensional time factor). This is load-bearing because Theorem 1.1's lower bound (67) and Lemma 6.2's displacement bound are obtained by applying (27); with the displayed inequality those bounds do not follow, and the constant choices in (69)–(70) collapse. The argument is repairable: replacing (27) by the sqrt-Γ version restores (67) and Lemma 6.2 verbatim. Thus the central theorems are likely correct, but the written proof contains a concrete gap at a key estimate. A separate minor typo also occurs in the proof of Corollary 1.2.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies connected simple graphs with bounded degree whose unnormalised graph Laplacian satisfies the classical dimension-free Bakry--Émery condition CD(0,∞), i.e. Γ2≥0. It claims two main results: Theorem 1.1 establishes volume doubling, #B(x,2r) ≤ K(d*) #B(x,r), and Theorem 1.3 establishes a scale-invariant L2 Poincaré inequality with dilation two, for all integer radii and with constants depending only on the maximum degree. Corollary 1.2 derives polynomial growth, resolving the Cushing--Liu--Peyerimhoff conjecture in a stronger form. The proof combines the modified nonlinear heat flow of Münch and Pajot--Russ with new point-mass consequences of Γ2≥0, a positive-resolvent smoothing estimate, a modified Li--Yau/Harnack bound, and a finite-volume heat-kernel localisation argument.","tokens_in":25677,"tokens_out":20416,"duration_ms":197088,"significance":"If correct, these are substantial results: they remove the finite-dimensional CD(0,n), CDE, and edge-regularity assumptions appearing in earlier volume-growth and doubling theorems, and they add a metric-ball Poincaré inequality under the classical dimension-free condition. The proof strategy is original and the constants are explicit. The bounded-degree hypothesis is convincingly shown to be necessary via the normalised-antitree counterexamples, so Assumption A is not merely a technical convenience. The manuscript also gives a useful comparison table separating CD, CDE, CDE′, and related conditions.","major_comments":[{"comment":"Equation (27) and its proof line are dimensionally inconsistent. From (26), ∥ΔP_s f∥∞ ≤ sqrt(N/(2s)) sqrt(∥Γ(f)∥∞); integrating gives ∥P_t f−f∥∞ ≤ sqrt(2N t ∥Γ(f)∥∞). The manuscript displays sqrt(2Nt)∥Γ(f)∥∞ and in the proof integrates sqrt(N/(2s))∥Γ(f)∥∞, treating Γ(f) as if it were already square-rooted. This is load-bearing: with the printed (27), the step at (67) would give (C²d*/(2r²))√(2Nt), not the stated C√(Nd*t)/r, and Lemma 6.2's displacement bound does not follow. The repair is purely local — replace the display and the integration line by the sqrt-Γ version — and then (67) and (82) are restored exactly. Because (27) is used in the proofs of Theorems 1.1 and 1.3, the manuscript needs this correction before acceptance.","section":"§3.2, Theorem 3.6, Eq. (27)"}],"minor_comments":[{"comment":"The footnote referring to a private preprint 'available upon request' is not independently verifiable and is not needed for the main theorems; it should either be removed or replaced by a public reference.","section":"§1.4, footnote"},{"comment":"The inequality chain #B(x,r) ≤ K^k ≤ K r^{log2 K} ≤ r^{2 log2 K} is correct but terse; please spell out the separate treatment of r=1 and the use of r≥2 and K≥2 so the exponent ceiling is unambiguous.","section":"§5, proof of Corollary 1.2"},{"comment":"The finite-volume localisation uses balls B(o,Λr) with a real dilation Λ in (85), while the main theorems are stated for integer radii. Since the paper already defines open balls for real radii, please state explicitly that balls of real radius are used throughout Sections 6--7 to avoid a perceived mismatch.","section":"§6.2--§7"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a major result and, with fixes, I think it is correct. It proves volume doubling and a ball Poincaré inequality under classical CD(0,∞) with bounded degree, no finite dimension and no edge-regularity, which settles the Cushing–Liu–Peyerimhoff conjecture in stronger form. The proof architecture is honest and detailed: Assumptions A and B are explicit, the point-mass identities and resolvent smoothing are worked out, there are no fitted parameters and no circularity. The comparison table is useful and, as far as I can check, accurate. Credit where due: the modified-flow Li–Yau argument and the finite-volume localisation are substantial pieces of work, and the point-mass coercivity estimates are elegant.\n\nThe soft spots are real but local. The stress-test note is right about (27). From (26) you get sqrt(2Nt) times sqrt(||Gamma(f)||_infty), not sqrt(2Nt) ||Gamma(f)||_infty. The proof line integrates sqrt(N/(2s)) ||Gamma(f)||_infty, which has the same scaling error. This is load-bearing because Theorem 1.1's estimate (67) and Lemma 6.2 cite (27). That said, the fix is exactly what the stress-test says: replace (27) by the sqrt-Gamma version, and (67), Lemma 6.2, and the subsequent argument go through verbatim. So this is a repairable gap in the written proof, not a broken strategy.\n\nThere is also a separate exponent slip in the proof of Corollary 1.2: it writes K^{r log_2 K} where it should be K^{log_2 r + 1}. The stated doubling exponent still works with that correction. And the footnote in §1.4 about a private preprint disproving Conjecture 2 of [5] is unverifiable and not needed for the main theorems. The authors should either provide that preprint or remove the claim; it should not be left as 'available upon request.' The bounded-degree hypothesis itself is not a flaw, and the authors are right that the normalised antitrees show it cannot simply be dropped.\n\nOverall: the central theorems are very likely correct, and this deserves a serious referee. I would send it to review with a request to fix (27), the corollary proof, and the footnote. The contribution is significant: it removes the finite-dimensional and edge-regularity hypotheses in one stroke.","headline":"The main theorems are likely right and settle the conjecture, but the written proof has a wrong Gamma-scaling in (27), an exponent slip in Corollary 1.2, and an unverifiable footnote; all are repairable.","tokens_in":26208,"tokens_out":4272,"would_cite":true,"duration_ms":43330,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05C50","60J27","58J35"],"pacs":[],"model":"deepseek-v4-flash","headline":"On connected bounded-degree graphs, the dimension-free Bakry–Emery condition CD(0,infinity) implies volume doubling and a scale-invariant Poincare inequality with dilation two, settling a polynomial-growth conjecture in stronger form.","keywords":["Bakry-Emery curvature","CD(0,infinity)","volume doubling","Poincare inequality","modified heat equation","graph Laplacian","polynomial growth","heat-kernel localisation"],"falsifier":"Find a connected simple graph with uniformly bounded degrees that satisfies CD(0,infinity) for the unnormalised Laplacian but whose balls violate #B(x,2r) <= K #B(x,r) for any fixed K, for example by exhibiting exponential volume growth along a sequence of radii. The known exponentially growing antitrees do not qualify because their degrees are unbounded; a bounded-degree counterexample would disprove the theorem, and a computational search on bounded-degree families with Gamma2 >= 0 could look for ratio blow-up.","tokens_in":1569,"feed_emoji":"📈","tokens_out":2715,"duration_ms":91308,"temperature":0.7,"pith_summary":"The paper proves that a single local curvature inequality, Gamma2(f)(x) >= 0 for every function and vertex, forces strong large-scale geometry on every connected simple graph whose vertex degrees are bounded. Under that hypothesis alone, balls double in volume up to a constant depending only on the maximum degree, and every ball supports an L2 Poincare inequality with a fixed dilation, again with a degree-only constant. This settles the polynomial-growth conjecture in a stronger doubling form. The proof works without any finite-dimensional curvature bound by extracting point-mass consequences of the positive semidefinite form Gamma2(.,.)(x) and by localising the heat kernel through a censored-chain argument.","feed_headline":"Nonnegative curvature forces volume doubling on bounded-degree graphs","feed_subtitle":"The same condition also yields a scale-invariant Poincare inequality, settling the polynomial-growth conjecture.","key_machinery":"The load-bearing object is the pointwise quadratic form Q_x(f,g) = Gamma2(f,g)(x), which CD(0,infinity) makes positive semidefinite. Testing Q_x against the point mass e_x gives the identity 4Gamma2(f,e_x)(x) = Delta^2 f(x) - 2Delta f(x) and the coercivity (Delta^2 f - 2Delta f)^2 <= 4 d_x(d_x+3) Gamma2(f)(x); Cauchy-Schwarz for Q_x then yields an effective dimension bound at local minima of Delta f. Inverting the factor A(A+2I) through the positive resolvent R = (A+2I)^{-1} turns that local control into the semigroup estimate |Delta P_t f|^2 <= (N/t) R(P_t Gamma(f) - Gamma(P_t f)), which supplies displacement and exit-time control. Around these sit the modified nonlinear heat flow partial_t","core_discovery":"The central claim is that the classical dimension-free curvature condition CD(0,infinity), i.e. Gamma2(f)(x) >= 0 for the unnormalised Laplacian, is itself sufficient to force the two structural properties that in Riemannian geometry are equivalent to Gaussian heat-kernel estimates: uniform volume doubling and a scale-invariant L2 Poincare inequality on metric balls. On a connected simple graph with maximum degree d*, the paper proves #B(x,2r) <= K(d*) #B(x,r) and sum_{z in B(x,r)} |f(z)-f_{B(x,r)}|^2 <= C_P(d*) r^2 E_{B(x,2r)}(f) for all integer radii r >= 1. The constants depend only on d*. The proof replaces the missing finite-dimensional reduction by two projections of the same positive-","pith_inferences":["Inference: the point-mass coercivity identity likely extends to other positive-semidefinite bilinear forms associated with graph generators, so analogous semigroup smoothing estimates may hold for weighted or directed Laplacians with bounded edge weight.","Inference: because the proof treats the Laplacian as a bounded operator on l^infinity, it should carry over to weighted graphs with uniformly bounded total edge weight per vertex; whether unbounded-degree graphs with sufficiently slow degree growth admit the same conclusion is an open question the paper does not address.","Inference: the explicit constants are built from comfortable parameter choices, so a natural next step is to seek optimal or near-optimal K(d*) and C_P(d*), perhaps via higher-order point-mass identities beyond the two-ball.","Inference: a direct parabolic Harnack inequality for the standard heat kernel, without passing through the lazy-kernel equivalence, may be within reach of the same localisation machinery."],"forward_implications":["Iterating the doubling bound gives polynomial growth #B(x,r) <= r^{D(d*)} with D(d*) = ceil(2 log_2 K(d*)), the conjectured bound in explicit form.","Because the lazy walk with kernel I + Delta/(2d*) has Dirichlet form proportional to edge energy, the standard equivalence upgrades the two theorems to two-sided Gaussian heat-kernel estimates and a parabolic Harnack inequality on bounded-degree CD(0,infinity) graphs.","The conclusion-level package previously known under the stronger finite-dimensional exponential curvature-dimension condition is recovered from the weaker dimension-free classical curvature hypothesis.","The Poincare inequality holds with the conventional dilation two and with constants depending only on d*; the volume-doubling theorem is used only to reduce a general degree-dependent dilation to two.","Finite-dimensional self-improvement (CD(0,n) for some finite n) is not needed; the proof works even on graphs that admit no finite CD(0,n), as the authors note."],"fun_headline_variants":["Nonnegative curvature yields volume doubling and Poincaré on graphs","Bounded-degree graphs: nonnegative curvature forces doubling and Poincaré","Dimension-free curvature on bounded-degree graphs: doubling and Poincaré","Curvature condition implies volume doubling on bounded-degree graphs","On bounded-degree graphs nonnegative curvature gives doubling and Poincaré"],"cache_read_input_tokens":27520,"weakest_assumption_plain":"The load-bearing premise is the uniform degree bound sup_x deg(x) < infinity: it makes the Laplacian a bounded operator on l^infinity, so the heat semigroup, the nonlinear-flow ODE, and the finite-volume censored chains are all well defined; without it the conclusion is false, since normalised antitrees have nonnegative curvature yet fail polynomial growth.","fun_headline_variants_meta":{"raw":{"variants":["Nonnegative curvature yields volume doubling and Poincaré on graphs","Bounded-degree graphs: nonnegative curvature forces doubling and Poincaré","Dimension-free curvature on bounded-degree graphs: doubling and Poincaré","Curvature condition implies volume doubling on bounded-degree graphs","On bounded-degree graphs nonnegative curvature gives doubling and Poincaré"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001828,"raw_usage":{"total_tokens":7021,"prompt_tokens":736,"completion_tokens":6285,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":480,"completion_tokens_details":{"reasoning_tokens":6196}},"tokens_in":480,"tokens_out":6285,"duration_ms":40310,"temperature":1.0,"reasoning_tokens":6196,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T23:06:01.641327+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a connected simple graph with uniformly bounded degrees that satisfies CD(0,infinity) for the unnormalised Laplacian but whose balls violate #B(x,2r) <= K #B(x,r) for any fixed K, for example by exhibiting exponential volume growth along a sequence of radii. The known exponentially growing antitrees do not qualify because their degrees are unbounded; a bounded-degree counterexample would disprove the theorem, and a computational search on bounded-degree families with Gamma2 >= 0 could look for ratio blow-up.","supporting_citations":[],"review_version":1}