{"id":"7530eeb5-d72b-44f1-a822-36dfe2217117","arxiv_id":"2607.15213","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"For expander, almost-Ramanujan, and random regular graphs, the paper claims asymptotic Brill-Noether existence at half-canonical degree, up to a constant factor.","lead":"This math paper claims asymptotic Brill-Noether existence for divisors on graphs, focusing on the half-canonical degree in regular expander and random graphs. If correct, it would be a significant step on a wide-open conjecture, but a key inequality in the odd-valence part of the proof is false.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Odd-valence cases rely on Lemma 4.9, whose asserted upper bound has the wrong direction; the correct Rayleigh inequality gives a lower bound, so the subtraction in Lemma 4.10 does not establish the claimed rank lower bound.","rationale":"The reader's weakest assumption identifies exactly the load-bearing flaw: Lemma 4.9's energy bound is reversed. My independent check of the Rayleigh quotient for L^+ confirms that the asserted upper bound is actually a lower bound, so the triangle-inequality argument in Lemma 4.10 loses its force. Since the odd-valence cases are essential to Theorem 1.2's fixed-k≥5 statement, the submitted proof does not establish the central claim. The verdict should remain REJECT. I do not see a separate independent objection; the even-valence and Q-divisor portions may be salvageable, but the paper as written is not.","tokens_in":21270,"tokens_out":4258,"duration_ms":35842,"concrete_test":"Re-derive Lemma 4.9 analytically: since L^+ has maximum eigenvalue 1/λ_min and minimum eigenvalue 1/λ_max, the Rayleigh bound gives sqrt(E_G(D_S)) ≥ sqrt(n)/(2 sqrt(λ_max)), so the stated ≤ is the reverse inequality. Numerically, take a random 3-regular almost-Ramanujan graph on n = 100 vertices, choose any bisection S, compute E_G(D_S) by solving L x = D_S, and compare sqrt(E_G(D_S)) with sqrt(n)/(2 sqrt(λ_max)). For a generic bisection the value strictly exceeds the Lemma 4.9 bound, confirming the error.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central defect is Lemma 4.9. For D_S in Div_0(G,Q) with entries ±1/2, E_G(D_S) = D_S^T L^+ D_S, where L^+ is the pseudoinverse of the Laplacian and has eigenvalues 1/λ_i. The Rayleigh quotient of L^+ satisfies ||D_S||²/λ_max ≤ D_S^T L^+ D_S ≤ ||D_S||²/λ_min. Since ||D_S||² = n/4, the correct inequality is sqrt(E_G(D_S)) ≥ sqrt(n)/(2 sqrt(λ_max)), not the asserted ≤. Lemma 4.10 then uses the triangle inequality h(D_S) ≥ h(O) − sqrt(E_G(D_S)). A lower bound on the subtracted term cannot yield a lower bound on h(D_S); one would need an upper bound, and the only available Rayleigh inequality points in the opposite direction. Lemma 2.19(2) is calibrated to the false subtraction, so the asymptotic Ω(sqrt(g)) conclusion for odd k does not follow. This affects Items (2), (3), and the odd-valence part of Item (4) of Theorem 1.2, i.e. essentially all cases with fixed k ≥ 5 except even k. The even-valence argument and the Q-divisor lower bound are unaffected, but the headline claim 'for any fixed k ≥ 5' is not supported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a geometric method, based on the energy quadratic form E_G on Div_0(G,R), to prove asymptotic Brill-Noether existence (ABNE) at the half-canonical degree for several families of regular graphs. The main contributions are: a description of the holes of the Laplacian-energy pair (L_G,E_G) as Crit_△(L_G) for regular graphs, a Cheeger-type lower bound on the covering radius with respect to E_G, norm-conversion inequalities to pass to ℓ¹ and to the simplex distance, and a parity-conversion step for odd-regular graphs by adding a Q-divisor D_S. Theorem 1.2 claims ABNE for even-valence spectral expanders, almost-Ramanujan graphs of fixed valence k≥5, random k-regular graphs (Types U and I) for k≥5, and almost-Ramanujan graphs of unbounded valence up to a √(log k(n) n) factor. The proof of the odd-valence cases rests on Lemma 4.9, which asserts an upper bound on √E_G(D_S). I find that this lemma has the wrong inequality direction, and the subsequent subtraction in Lemma 4.10 does not establish the claimed rank lower bound. The even-valence and Q-divisor parts appear plausible, but the odd-valence cases, which are essential for the headline 'any fixed k≥5' statement, are not proved.","tokens_in":21584,"tokens_out":7593,"duration_ms":83271,"significance":"If correct, the paper would confirm an asymptotic form of Baker's Brill-Noether existence conjecture on expander and random regular graphs, a significant advance for a problem that lacks general techniques. The reformulation via energy pairings, the hole description, and the Cheeger inequality are original and interesting. The even-valence parts are well argued and credible. However, the odd-valence parity-conversion argument is load-bearing and fails: Lemma 4.9 asserts an upper bound where the Rayleigh inequality gives a lower bound. Since fixed valences k≥5 include both even and odd k, and the odd cases form the majority, the central advertised result is not established. The paper is therefore not suitable for publication in its current form, despite the value of some of its components.","major_comments":[{"comment":"The inequality asserted in Lemma 4.9, √E_G(D_S) ≤ √n/(2√λ_max(G)), has the wrong direction. Since E_G(D_S)=D_S^T L_res^{-1} D_S and the eigenvalues of L_res^{-1} are 1/λ_i, the Rayleigh principle gives ||D_S||²/λ_max ≤ E_G(D_S) ≤ ||D_S||²/λ_min. With ||D_S||²=n/4, the correct lower bound is √E_G(D_S) ≥ √n/(2√λ_max). The proof says the claim follows 'by the Rayleigh inequality', but that inequality yields exactly the opposite inequality. The displayed bound in Lemma 4.9 is therefore false as stated.","section":"4.3, Lemma 4.9"},{"comment":"Lemma 4.10 derives h_{E_G,Crit}(D_S) ≥ h_{E_G,Crit}(O) − √E_G(D_S) and then subtracts the quantity from Lemma 4.9. A lower bound on √E_G(D_S) cannot be subtracted to obtain a lower bound on h(D_S); one would need an upper bound. With the correct Rayleigh inequality, the subtracted term is ≥ √n/(2√λ_max), which for expander graphs is of order √n, i.e. the same order as the target √g. Therefore the chain in Lemma 4.10 does not prove h(D_S)∈Ω(√g). Consequently the odd-valence cases in Theorem 1.2 Items (2) and (3) for odd k, the odd-valence part of Item (4), and the corresponding applications in Corollary 1.3 and Theorems 5.2–5.4 are not established. The asymptotic expression in Lemma 2.19(2) is calibrated to the false subtraction and does not repair the problem.","section":"4.3, Lemma 4.10 and odd-valence proof"}],"minor_comments":[{"comment":"The sentence about the distance function says it satisfies all metric properties 'except possibly symmetry, i.e. d_C(p1,p2)=d_C(p2,p1)'. This is incoherent: the identity is symmetry. It should read 'except possibly symmetry, i.e., d_C(p1,p2) need not equal d_C(p2,p1)'.","section":"Section 2.2"},{"comment":"In the paragraph after Theorem 1.2, 'the chief difficulty in tacking Brill-Noether existence' should be 'tackling'.","section":"Section 1"},{"comment":"The notation S^{(n)} in Lemma 4.10 and the statement 'Throughout the following proof, S^{(n)} is an arbitrary vertex subset' is potentially confusing: the proof of Theorem 1.2 does not specify how S^{(n)} is chosen, and if the argument worked, any choice would do. Please state explicitly that the construction is choice-free if that is intended.","section":"Section 4.3"}],"recommendation":"reject","confidential_remarks":"The reader's and skeptic's diagnosis is correct: Lemma 4.9 has the wrong inequality direction, and Lemma 4.10 depends on it in an essential way. The even-valence parts are plausible, but the odd-valence cases are needed for the paper's main claim 'for any fixed k≥5'. I do not see a local fix: the correct Rayleigh bound gives a lower bound, not an upper bound, and for expanders the subtracted term is of the same order as the target. I recommend rejection. If a valid upper bound for E_G(D_S) or an alternative parity construction is found, a revised version could be reconsidered."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here is the short version: the paper introduces a genuinely new route to asymptotic Brill–Noether existence on regular graphs — using the energy quadratic form on the Laplacian lattice and a Cheeger-style lower bound on the covering radius — and the even-valence cases look plausible. But the odd-valence cases, which cover most fixed k ≥ 5, rest on Lemma 4.9, and that lemma asserts the wrong direction of the Rayleigh inequality. The correct bound is sqrt(E_G(D_S)) ≥ sqrt(n)/(2 sqrt(λ_max)), not ≤. Lemma 4.10 then subtracts this quantity from a lower bound; to get a lower bound on h(D_S) you would need an upper bound on the subtracted term. So the subtraction cannot establish the claimed Ω(√g) rank bound. The flaw is load-bearing: it affects items (2), (3), and the odd-valence part of (4) of Theorem 1.2, i.e. essentially all cases with fixed odd k ≥ 5.\n\nWhat is good: the hole description in Theorem 3.1, identifying Crit of the Laplacian-energy pair with acyclic orientations, is clean and carefully argued; the Cheeger inequality for the covering radius (Lemma 3.6) is a real new tool; and the conversion from energy to ℓ1 via Proposition 4.1 is straightforward and useful. The even-valence proof works: for constant even k, spectral gap is bounded below and λ_max ≤ 2k, so Proposition 4.3 gives Ω(√g). The unbounded even-valence argument through the Poisson equation and diameter is also plausible, although it inherits the same issue when k(n) is odd.\n\nThe soft spot is not a minor gap; it is the central mechanism for odd valences. The paper's reliance on the author's earlier work [2,36] is not circular — those results are published and independent. The self-citation pattern is not a problem here. But Lemma 4.9 as written is false, and the proof of Lemma 4.10 uses it in the wrong direction. A reviewer should ask whether a different choice of D_S or an upper bound on E(D_S) can be obtained by another method; as submitted, the odd-valence theorem is unsupported.\n\nWho this is for: people working on graph Brill–Noether or on lattice/spectral methods for graphs will find the framework worth studying, but they should not rely on the stated theorem for odd k until the proof is repaired. It deserves a serious referee — the even-valence case and the geometric framework are substantial, and the error is identifiable and possibly fixable. I would not accept the current version, but I would send it to review rather than desk-reject.","headline":"Promising framework, but the odd-valence proof uses the Rayleigh inequality in the wrong direction, so the claimed Theorem 1.2 for fixed k ≥ 5 is not established as written.","tokens_in":22075,"tokens_out":3916,"would_cite":false,"duration_ms":105897,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05C35","05C50","05C80"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that for any fixed valence k at least five, almost all simple connected k-regular graphs contain divisors of degree g-1 with rank Omega(sqrt(g)), giving an asymptotic confirmation of the Brill-Noether existence conjecture a","keywords":["Brill-Noether theory","graph divisors","rank of divisors","Laplacian lattice","energy quadratic form","covering radius","expander graphs","random regular graphs"],"falsifier":"Evaluate E_G(D_S) for a bisection S on a concrete odd-regular graph (e.g., the 3-regular Petersen graph) and compare with n/(4 lambda_max). If E_G(D_S) exceeds that value, the odd-valence construction collapses. A simpler check is the sign in the variational argument: for a positive-definite operator, the min-max theorem bounds the quadratic form from below, not from above.","tokens_in":21137,"feed_emoji":"🌐","tokens_out":8619,"duration_ms":75744,"temperature":0.7,"pith_summary":"The paper aims to prove an asymptotic version of the Brill-Noether existence conjecture for graphs: on well-connected regular graph families, there should be divisors of degree g-1 (the half-canonical degree) whose rank grows at least like the square root of the genus. It establishes this for spectral expanders of even valence, for near-optimal spectral expanders of any fixed valence at least five, and for almost all random regular graphs in two standard random models; for unbounded valence it achieves the bound up to a logarithmic factor. The value is that the proof avoids the usual dependence on the Picard group and reduced divisors, instead using a geometric covering-radius argument for the Laplacian lattice equipped with the energy quadratic form. If correct, this confirms the asymptotic half-canonical case of the conjecture on a large and natural class of graphs.","feed_headline":"sqrt(g) rank on half-canonical divisors for almost all regular graphs","feed_subtitle":"The paper's energy-pairing proof establishes the half-canonical rank bound for expander and random regular graph families.","key_machinery":"The energy quadratic form E_G on degree-zero divisors, defined as the pairing of a divisor with the inverse Laplacian acting on it, together with the Laplacian lattice L_G. The paper proves that the holes of the pair (L_G, E_G) coincide with the critical set Crit(L_G) appearing in the covering-radius formulation of Brill-Noether existence, and that the covering radius of this set is controlled from below via an inequality relating the energy norm to the spectral gap. The rank formula then converts this energy-geometric bound into a lower bound on the rank of the half-canonical divisor.","core_discovery":"The central discovery is that the covering radius of the critical set Crit(L_G) with respect to the energy quadratic form E_G is at least a spectral quantity of order sqrt(n * lambda_min), and that this lower bound is attained at the zero divisor. Combined with the rank formula that expresses the rank of the half-canonical divisor as half the ell-1 distance from the origin to Crit(L_G), and with norm-conversion inequalities relating the energy norm and the ell-1 norm, the paper derives the rank lower bounds. In the odd-valence case it adds a half-integer bisection divisor D_S to the half-canonical divisor, producing an integer divisor of half-canonical degree; the claimed energy bound for D_","pith_inferences":["If the claimed energy bound for the bisection divisor D_S fails (as a sign error in the variational argument suggests), the odd-valence cases of the theorem would need a different argument; the even-valence and unbounded-even-valence cases would remain intact.","The method of replacing the ell-1 norm by the energy quadratic form and then converting back is likely to extend to other weighted energy forms, as the paper suggests, potentially yielding rank bounds at degrees far from half-canonical.","A direct check of the energy bound on small odd-regular graphs (e.g., the Petersen graph) would either confirm the proof or expose the flaw; if the bound fails, the gap is concrete and testable."],"forward_implications":["The half-canonical degree case of the Brill-Noether existence conjecture holds asymptotically for all fixed-valence spectral expander graphs, including even-valence expanders and near-optimal spectral expanders of valence at least five.","For any fixed k >= 5, almost all simple connected k-regular random graphs (in the uniform and perfect-matching models) have divisors of degree g-1 with rank Omega(sqrt(g)).","Degrees g-1 - o(sqrt(g)) also support divisors of rank Omega(sqrt(g)) in the fixed-valence cases, by subtracting effective divisors.","For unbounded-valence expanders, divisors of degree g-1 with rank Omega(sqrt(g)/sqrt(log n)) exist, leaving a logarithmic gap to the full conjecture.","The path-reversal graphs associated to cycle-cocycle and cocycle reversal systems have diameter at least of order sqrt(n) (or sqrt(n/log n) for unbounded valence), a dynamical consequence of the rank bounds."],"fun_headline_variants":["Half-canonical rank for almost all regular graphs","Expander graphs hit Brill-Noether at half-canonical degree","Energy-pairing proof of graph rank bound","Almost-Ramanujan graphs satisfy half-canonical conjecture","Cheeger inequality unlocks graph divisor ranks"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The odd-valence parts of the theorem depend on the assertion that the bisection divisor D_S has energy at most sqrt(n)/(2 sqrt(lambda_max)), claimed in the text to follow from the variational characterization of eigenvalues; that characterization yields the opposite bound, so the upper bound on the energy of D_S is the load-bearing claim for odd k, and if it falls the rank lower bound for odd valences is not established.","fun_headline_variants_meta":{"raw":{"variants":["Half-canonical rank for almost all regular graphs","Expander graphs hit Brill-Noether at half-canonical degree","Energy-pairing proof of graph rank bound","Almost-Ramanujan graphs satisfy half-canonical conjecture","Cheeger inequality unlocks graph divisor ranks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000302,"raw_usage":{"total_tokens":1568,"prompt_tokens":727,"completion_tokens":841,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":471,"completion_tokens_details":{"reasoning_tokens":764}},"tokens_in":471,"tokens_out":841,"duration_ms":8652,"temperature":1.0,"reasoning_tokens":764,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T23:59:43.695070+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate E_G(D_S) for a bisection S on a concrete odd-regular graph (e.g., the 3-regular Petersen graph) and compare with n/(4 lambda_max). If E_G(D_S) exceeds that value, the odd-valence construction collapses. A simpler check is the sign in the variational argument: for a positive-definite operator, the min-max theorem bounds the quadratic form from below, not from above.","supporting_citations":[],"review_version":1}