{"id":"5f1ef586-b637-4216-8c9d-f02bcb8196b4","arxiv_id":"2607.27162","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"On bounded-degree non-negatively Ollivier-curved graphs, random-walk displacement is at most t exp(C√(log t log log t)) and log-volume is at most exp(C√(log r log log r)), pointwise in the root.","lead":"Bounded-degree graphs with non-negative Ollivier–Ricci curvature have pointwise near-diffusive random walks and subexponential volume growth. This upgrades earlier averaged bounds and moves closer to the conjectured polynomial growth.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified","rationale":"The Reader correctly isolates Lemma 5.3/Prop. 5.4 as the technically most delicate step and correctly judges that it is standard for Wang-type change-of-measure arguments and does not threaten the claim. The bootstrap loop (Thms 3.7–3.11) and the final optimization of n(t) are transparent and close. No hidden assumption, circularity, or regime failure appears. Consequently the ACCEPT verdict with low correctness risk stands; no adjustment is warranted.","tokens_in":19381,"tokens_out":482,"duration_ms":9814,"concrete_test":"Independently re-derive the entropy-cost identity (5.9) and the sign of (∂s+Aλs)Vs+Ψκ(λs) in (5.11) from the definitions of Vs, Hs and λs=κ(eHs−1) without consulting the manuscript; confirm that the coefficient B=32/Pmin is large enough to absorb the 16B/u term coming from A0 dist2≤16. If the inequality holds with room, the load-bearing coupling step is secure.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Theorem 2.1) rests on two pillars that both close under ordinary scrutiny: (i) non-negative Ollivier curvature produces LH(32d) via the geodesic-drift coupling of Lemma 5.3 / Prop. 5.4 / Thm 5.1, and (ii) LH(A) produces the stated subexponential bounds by the self-improving displacement–entropy–volume loop of §3 whose iteration count is optimized to √(log t log log t). The mass-adjustment construction that forces π(a,b−)=K(b,b−) while preserving the W1 bound is standard and the entropy-cost cancellation (5.9)–(5.11) is algebraic once the rate λs=κ(eHs−1) is chosen. The bootstrap constants remain finite for every finite stage n and the final n(t)∼√(log t/log log t) choice is elementary. Residual risk is only the ordinary possibility of an arithmetic slip in constant tracking, not a structural gap.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The paper proves that every (possibly infinite) connected graph of bounded degree and non-negative Ollivier–Ricci curvature satisfies pointwise near-diffusive displacement and pointwise subexponential volume growth: uniformly in the root x, E_x dist(x,X_t)^2 ≤ t exp[C_d √(log t log log t)] and log Vol(B(x,r)) ≤ exp[C_d √(log r log log r)] for t,r ≥ e^e. The argument is organized in three blocks: (i) non-negative Ollivier curvature plus deg ≤ d implies a log-Harnack inequality LH(32d) via a geodesic-drift change-of-measure coupling; (ii) LH(A) alone yields the stated bounds by a self-improving displacement–entropy–volume bootstrap whose iteration depth is optimized to √(log t / log log t); (iii) LH(A) plus a diffusive-moment hypothesis implies volume doubling. Discrete-time displacement is recovered from the continuous-time statement by a Poissonization comparison.","tokens_in":19527,"tokens_out":1112,"duration_ms":35774,"significance":"The result upgrades the averaged subexponential bounds of Hutchcroft–Münch to uniform pointwise control on every rooted ball and every starting point, including infinite graphs. This is a genuine local-to-global advance under Ollivier curvature and supplies the strongest available evidence toward their polynomial-growth/diffusive conjecture. The modular architecture (curvature ⇒ LH; LH ⇒ bootstrap; LH + diffusion ⇒ doubling) is clean, the constants are tracked explicitly through the loop, and the coupling argument is a carefully executed discrete analogue of Wang’s change-of-measure method. These features make the paper a substantial and reusable contribution to discrete curvature and geometric analysis on graphs.","major_comments":[],"minor_comments":[{"comment":"In §3 (Heat entropy), the same symbol H_t(x) is used both for the m-weighted heat entropy and for Shannon entropy; the subsequent identity H_t = H_t + E log m(X_t) is therefore ambiguous in plain text. Introduce a distinct notation (e.g., script or tilde) for Shannon entropy and keep it consistent through Lemmas 3.3–3.5.","section":"§3, Heat entropy"},{"comment":"Lemma 3.6: the factor 4/η in the conclusion is slightly loose relative to the integral lower bound η/(2 M^{-a}); a one-line remark that any constant >2/η works would help the reader track the later C_E(A).","section":"Lemma 3.6"},{"comment":"Theorem 3.11: the universal c_0 appearing in the Davies–Gaffney–Grigor’yan lower bound (3.10) is never given a numerical value. Since C^* := 2^{5/3} c_0^{-1} log 2 enters the stage constants, either cite a concrete c_0 from the literature or note that any positive c_0 is absorbed into C_D(A).","section":"Theorem 3.11"},{"comment":"Proposition 5.4 / Lemma 5.3: the construction that forces π_{a,b}(a,b_-)=K(b,b_-) while preserving the W_1 bound is correct, but the verification that the modified plan remains non-negative relies on π(a,b)≥1/2. A short parenthetical recalling K(a,a)≥3/4 and K(b,V\\{b})≤1/4 would make the argument self-contained for readers unfamiliar with the lazy-kernel arithmetic.","section":"§5, Lemma 5.3"},{"comment":"Several minor typographical inconsistencies appear: “locally-finite” vs “locally finite”, “Ollivier-Ricci” vs “Ollivier–Ricci”, and the future date “July 2026” on the title page. Normalize hyphenation and fix the date before publication.","section":"Throughout"},{"comment":"In the display after (3.19), the recursive inequality for log(1+M_{n+1}) is written with an implicit absorption of lower-order terms into C(A); stating the precise inductive hypothesis used for the bound log(1+M_n)≤C(n+2)log(e+n) would remove any doubt about the constant bookkeeping.","section":"Theorem 3.12"}],"recommendation":"accept","confidential_remarks":"The manuscript is ready for acceptance; the empty major-comments list reflects that the two load-bearing pillars (Ollivier⇒LH via the geodesic-drift coupling, and the LH bootstrap) close under ordinary scrutiny. Residual risk is only ordinary arithmetic slip in constant tracking, which does not affect the form of the final bounds. Fit for a geometry/analysis journal is excellent."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The real news is that Zhou gets the pointwise, root-uniform versions of the subexponential volume and near-diffusive displacement bounds that Hutchcroft–Münch only had in averaged form, and does it on possibly infinite bounded-degree graphs. That closes the natural next question in the program.\n\nWhat works well is the architecture. Everything runs through an abstract log-Harnack condition LH(A). Ollivier + deg ≤ d gives LH(32d) by a geodesic-drift change-of-measure (the discrete Wang coupling in §5). Then a self-improving displacement–entropy–volume loop under LH(A) produces the √(log t log log t) factor after optimizing the number of iterations. The three blocks are cleanly separated, constants are tracked explicitly, and the usual tools (data-processing for KL, Davies–Gaffney–Grigor’yan, Girsanov for pure-jump processes) are applied correctly. The doubling implication under an extra diffusive hypothesis is a nice bonus and shows the method is flexible. The birth–death counter-example correctly shows that pure Ollivier does not force diffusivity once weights are allowed, so the extra assumption is necessary.\n\nSoft spots are ordinary, not structural. The mass-adjustment that forces the coupling to put exactly the right mass on the geodesic edge (Lemma 5.3) is the load-bearing step for the entropy cost; it looks standard and the cancellation (5.9)–(5.11) is algebraic once the rate is chosen, but it is the place an arithmetic slip could hide. The γ ≤ 2/3 threshold in the volume-to-displacement step is technical bookkeeping. Neither threatens the logic. The bounds still sit short of the polynomial/diffusive conjecture, which the paper does not claim to settle.\n\nThis is for people working on discrete Ricci curvature, Markov-chain geometry, or geometric group theory who already care about the Hutchcroft–Münch results. The math is careful, the citations are honest, and there is no circularity. I would send it to referees without hesitation and would cite the pointwise statements.","headline":"Pointwise upgrade of Hutchcroft–Münch under Ollivier curvature: clean LH bootstrap plus a working discrete Wang coupling; solid and worth engaging.","tokens_in":20227,"tokens_out":520,"would_cite":true,"duration_ms":12063,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C21","60J27","05C81","58J65"],"pacs":[],"model":"grok-4.5","headline":"Bounded-degree graphs with non-negative Ollivier–Ricci curvature have pointwise near-diffusive walks and subexponential volume growth at every root.","keywords":["Ollivier-Ricci curvature","random walks on graphs","volume growth","log-Harnack inequality","displacement estimates","bounded-degree graphs","heat entropy"],"falsifier":"Exhibit a single infinite connected graph of bounded degree and non-negative Ollivier–Ricci curvature whose ball volumes satisfy log Vol(B(x,r)) / exp(c √(log r log log r)) → ∞ for every c, or whose continuous-time walk satisfies E_x dist(x,X_t)^2 / (t exp(c √(log t log log t))) → ∞ for every c.","tokens_in":20178,"feed_emoji":"📐","tokens_out":1038,"duration_ms":26065,"temperature":0.7,"pith_summary":"This paper shows that a purely local curvature condition on a graph already controls global geometry and random-walk speed, uniformly from every starting vertex. If degrees are bounded and Ollivier–Ricci curvature is non-negative, continuous-time random-walk displacement is at most t times a slowly growing subexponential factor, and the logarithm of ball volume grows at most like exp of the square root of log r times log log r. Earlier results gave only averaged or first-moment statements; the advance is a pointwise bound that holds for every root on possibly infinite graphs. The argument works by turning curvature into a log-Harnack inequality and then running a self-improving loop among displacement, heat entropy, and volume. A sympathetic reader cares because this is a discrete Bishop–Gromov-type statement: local non-negative curvature still forces almost-diffusive large-scale behavior.","feed_headline":"Curved graphs grow subexponentially from every root","feed_subtitle":"Non-negative Ollivier curvature forces near-diffusive walks and slow volume growth, pointwise","key_machinery":"The log-Harnack inequality LH(A): relative entropy between heat kernels started at neighboring points is at most A times (dist²/t + dist/√t). Non-negative Ollivier curvature supplies LH(32d) by a coupling-by-change-of-measure that adds a geodesic-drift jump. LH(A) then feeds a bootstrap loop—displacement controls entropy, entropy controls volume, volume improves displacement—whose iteration yields the subexponential bounds.","core_discovery":"On any connected graph of maximum degree d with non-negative Ollivier–Ricci curvature there is a constant C_d such that, for every vertex x and all t,r at least e^e, the continuous-time walk satisfies E_x dist(x,X_t)^2 ≤ t exp[C_d √(log t log log t)] and log Vol(B(x,r)) ≤ exp[C_d √(log r log log r)]. The same near-diffusive bound holds for the discrete-time lazy walk. The estimates are pointwise in the root, not merely averaged.","pith_inferences":["The remaining gap to the conjectured polynomial volume bound is now purely quantitative: one must remove the iterated-log factors produced by the bootstrap, not invent a new averaging device.","The same LH-plus-bootstrap loop should apply verbatim to other discrete curvatures (Forman, Bakry–Émery) once a comparable log-Harnack inequality is available.","On transitive or unimodular examples the pointwise bounds immediately recover and slightly strengthen the earlier averaged theorems without ergodic decomposition."],"forward_implications":["Pointwise subexponential volume growth holds on every (possibly infinite) bounded-degree non-negatively curved graph, not only on finite or unimodular ones.","The same graphs have pointwise near-diffusive continuous- and discrete-time random walks from every root.","Under the extra hypothesis of truly diffusive moments, the log-Harnack inequality upgrades to genuine volume doubling.","Any future proof of the Hutchcroft–Münch polynomial-growth conjecture can start from these pointwise subexponential bounds rather than from averaged estimates."],"fun_headline_variants":["Non-negative Ollivier curvature forces pointwise near-diffusive walks","Bounded-degree graphs with non-negative curvature grow subexponentially","Pointwise subexp volume growth under non-negative Ollivier-Ricci","Near-diffusive displacement on every root for non-negatively curved graphs","Non-negative Ollivier-Ricci implies slow pointwise volume growth"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"That non-negative curvature always lets you couple two lazy walks so the second particle can drift one step closer along a geodesic at a uniformly positive rate, without changing the law of the first particle.","fun_headline_variants_meta":{"raw":{"variants":["Non-negative Ollivier curvature forces pointwise near-diffusive walks","Bounded-degree graphs with non-negative curvature grow subexponentially","Pointwise subexp volume growth under non-negative Ollivier-Ricci","Near-diffusive displacement on every root for non-negatively curved graphs","Non-negative Ollivier-Ricci implies slow pointwise volume growth"]},"model":"grok-4.5","effort":"low","cost_usd":0.005383,"raw_usage":{"total_tokens":1475,"prompt_tokens":758,"num_sources_used":0,"completion_tokens":90,"cost_in_usd_ticks":53828000,"prompt_tokens_details":{"text_tokens":758,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":627,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":758,"tokens_out":90,"duration_ms":10454,"temperature":1.0,"reasoning_tokens":627,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-30T10:34:56.044743+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Exhibit a single infinite connected graph of bounded degree and non-negative Ollivier–Ricci curvature whose ball volumes satisfy log Vol(B(x,r)) / exp(c √(log r log log r)) → ∞ for every c, or whose continuous-time walk satisfies E_x dist(x,X_t)^2 / (t exp(c √(log t log log t))) → ∞ for every c.","supporting_citations":[],"review_version":1}