{"id":"77a54b7c-eb68-4a60-8365-8ca11f76b6e1","arxiv_id":"2607.16770","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Subextensive random boundary perturbations leave the quenched specific free energy of the Edwards-Anderson model unchanged, with almost-sure self-averaging.","lead":"This paper proves that random boundary perturbations in the Edwards-Anderson spin-glass model do not change the limiting specific free energy whenever their total energy is subextensive, and provides an almost-sure self-averaging statement. The result is a rigorous confirmation of an expected fact, useful as a clean general theorem covering surface fields, exterior spins, and periodic boundaries.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified.","rationale":"The paper's core claim—that subextensive boundary perturbations do not affect the limiting quenched specific free energy—is proven correctly. Lemma 7's pointwise bound and the subextensive envelope conditions are sufficient to transfer the free-boundary limit and self-averaging to the perturbed model. Proposition 5's superadditivity argument is standard: conditioning on internal variables, convexity of log Z_Λ(u), and E u = 0 yield P(Λ) ≥ ΣP(Λ_i); the tiling argument then gives convergence. The reliance on mean-zero couplings is explicit in Eq. (3) and is exactly the condition that makes the Jensen step work; this is a stated hypothesis, not a hidden assumption. Proposition 6's Efron–Stein estimate is quantitatively correct and the Borel–Cantelli step is valid for d ≥ 2. The numerical figures are not reproducible as presented, but they are illustrative and do not affect the theorem. The reader's weakest assumption is therefore not a load-bearing concern for the stated theorem; it only indicates a possible direction for generalization. The verdict should remain CONDITIONAL because of the non-fatal reproducibility issues, but no mathematical correction is needed.","tokens_in":5818,"tokens_out":20371,"duration_ms":183463,"concrete_test":"Attempt to re-prove the superadditivity inequality (20) with E J_e = m ≠ 0 by recentering the cross-piece couplings u at m: verify whether E_u log Z_Λ(u) ≥ log Z_Λ(m) and whether log Z_Λ(m) factorizes over the pieces. If the inequality fails, the zero-mean hypothesis in Eq. (3) is essential and should be stated in the abstract; if it still holds, the theorem can be strengthened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No significant objection identified. The central theorem is supported by a rigorous proof: Lemma 7 provides a deterministic comparison of the perturbed and free specific free energies; Proposition 5 establishes the free-boundary thermodynamic limit via superadditivity of E log Z, using the mean-zero hypothesis (3) in the Jensen step; Proposition 6 gives an Efron–Stein variance bound of order 1/V_L, yielding almost-sure self-averaging. The reader's flagged zero-mean assumption is an explicit hypothesis, not a hidden gap. The proof of Proposition 5 is terse but complete. The only caveat is the undocumented numerical figures (Figs. 2–4), which are not load-bearing for the mathematical claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a boundary-insensitivity theorem for the quenched specific free energy of the short-range Edwards–Anderson model on cubic boxes. Theorem 1 states that if a random boundary perturbation has an envelope A_L that is subextensive both in expectation and almost surely (A_L/V_L -> 0), then the perturbed quenched specific free energy converges to the same deterministic limit as the free-boundary model, and the quenched free energy self-averages almost surely. The proof is self-contained: Lemma 7 gives a deterministic comparison of partition functions; Proposition 5 establishes the free-boundary thermodynamic limit via superadditivity of the quenched pressure, using the mean-zero coupling assumption (3) in a Jensen step; Proposition 6 gives an Efron–Stein variance bound of order 1/V_L, yielding almost-sure self-averaging for d >= 2. Corollaries 2–4 verify the envelope hypothesis for scalar surface fields, exterior-spin boundary conditions, and periodic wrap-around bonds, with O(L^{-1}) comparisons of quenched means. The paper explicitly notes that it does not address convergence of finite-volume Gibbs measures, only the specific free energy.","tokens_in":5967,"tokens_out":5024,"duration_ms":49243,"significance":"If the result holds, it provides a clean and general statement of a physically expected fact: subextensive random boundary perturbations do not affect the limiting quenched specific free energy, and they also inherit a.s. self-averaging. The proof is elementary and fully self-contained, relying only on standard inequalities (Jensen, Efron–Stein, Borel–Cantelli), with no fitting parameters or hidden assumptions beyond those stated. The explicit hypotheses, especially the mean-zero condition (3), are used transparently. The paper also gives a useful unification of several common boundary conditions (surface fields, exterior spins, periodic bonds) under one theorem. Although the result is expected and incremental, it is a valuable reference statement for the mathematical spin-glass literature, and the Gaussian strengthening in the remark is a nice extra. The numerical figures are illustrative but not load-bearing.","major_comments":[],"minor_comments":[{"comment":"The sentence 'Whenever B_L is random, we assume that it is measurable and that the resulting sample specific free energy is integrable' appears twice verbatim. Please delete the duplicate.","section":"Section 2, after Eq. (5)"},{"comment":"The line 'observing that the corresponding limsup is bounded by the same supremum' is terse. Since the inequality P(Λ_L)/L^d ≤ sup_m P(Λ_m)/m^d holds trivially for every L, the limsup is bounded by the same supremum; combining this with (23) proves convergence. For the reader's convenience, spell this out or explicitly cite Fekete's lemma for the superadditive scalar sequence P(Λ_L). The argument is correct as written, but the presentation should be clearer.","section":"Proposition 5"},{"comment":"The numerical figures lack any description of the simulation setup: model parameters (β, disorder distribution, dimension), number of samples, error bars, and how f∞ was estimated. If they are meant to illustrate the theorems, a short caption or text description is needed. If they are not part of the mathematical content, they should be removed or labeled as schematic. In particular, Fig. 4's caption says the curves 'confirm' an exact O(L^{-1}) suppression rate, while the paper proves upper bounds; please soften this claim.","section":"Figures 2–4"},{"comment":"The phrase 'The right-hand side is summable when d >= 2' refers to the Chebyshev bound in (28), not to the variance bound in (27). Rewording will avoid confusion: after (28), say 'the right-hand side of (28) is summable for d >= 2.'","section":"Proof of Proposition 6, after Eq. (27)"}],"recommendation":"minor_revision","confidential_remarks":"This is a sound, clearly written short note. The main theorem is correct and the proof is complete; the only flagged issue is a terse line in Proposition 5 that should be elaborated for readability. The numerical figures are not needed for the mathematical claims and lack experimental details; I suggest the authors either supplement or remove them. The paper fits the scope of math-ph and, after the minor revisions listed, should be acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take on arXiv:2607.16770. It does what it says: proves a general envelope theorem for random boundary perturbations in the EA model, with full-sequence a.s. self-averaging. The key ingredients are Lemma 7's deterministic comparison and an Efron–Stein variance bound, and the paper packages them into a theorem that covers scalar surface fields, exterior spins, and periodic wrap-around bonds in one place. I don't know a prior statement that isolates the subextensive envelope condition this cleanly. The proofs are correct, and the O(1/L) bounds for the specific examples come from simple surface counting and are plausibly sharp.\n\nThe main theorem's proof is fine. The only genuinely soft spot is Proposition 5: the convergence of the free-boundary pressure is asserted via a one-line limsup argument that implicitly uses superadditivity and Fekete's lemma. It's true and standard, but it should be spelled out, because the superadditivity itself (Eq. (20)) is load-bearing. The Jensen step uses mean-zero couplings, which is an explicit hypothesis, not a hidden gap. The paper doesn't discuss E J_e ≠ 0, but that's a different model, not a flaw.\n\nThe numerical figures are the actual problem. Figures 2–4 present simulation results with no parameters, no code, no data, and no simulation protocol. They are decorative and non-reproducible. The math doesn't need them, so the cleanest fix is to delete them or move them to an appendix with full documentation. As they stand, they undercut the otherwise careful tone.\n\nThe references look appropriate: [2–4] cover thermodynamic limits, [9] is the relevant random-boundary result, [10, 11] are standard tools. No self-citation, no fitting. The scope remarks are honest about not addressing Gibbs-state convergence.\n\nWho should read it: people needing a precise boundary-stability statement for quenched free energies in disordered Ising models, especially in metastate discussions. It doesn't resolve an open problem, but it's solid mathematical physics. I'd send it to a serious referee; the referee should ask for Proposition 5 to be written out and insist that the numerical figures be either documented or removed.","headline":"Clean, correct envelope theorem for random boundary perturbations; the real issue is unreproducible numerical figures and a terse but fixable superadditivity proof.","tokens_in":6408,"tokens_out":2003,"would_cite":true,"duration_ms":19233,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["82B44","82B20","60K35"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that any random boundary perturbation whose total energy grows slower than the volume leaves the limiting quenched free energy of the Edwards–Anderson spin glass unchanged, and that this free energy self-averages almost sur","keywords":["Edwards–Anderson model","quenched free energy","spin glass","boundary perturbation","self-averaging","thermodynamic limit","surface disorder","superadditivity"],"falsifier":"Compute the quenched specific free energy difference between periodic and free boundary conditions in d = 2 with zero-mean couplings and check whether it approaches zero as L → ∞; if the difference does not approach zero, the envelope comparison fails. Alternatively, run the same model with couplings of non-zero mean and see whether the free-boundary specific free energy still converges to a boundary-independent limit; a single instance where it does not would expose the zero-mean premise.","tokens_in":5735,"feed_emoji":"🎲","tokens_out":10493,"duration_ms":92071,"temperature":0.7,"pith_summary":"This paper establishes a boundary-stability theorem for the quenched specific free energy of the short-range Edwards–Anderson spin glass. It shows that if a random perturbation lives on the boundary of a cubic box and its energy envelope A_L satisfies A_L/V_L → 0 both in expectation and almost surely, then the limiting free energy is exactly the same as with free boundary conditions, and the sample free energy converges almost surely to the same deterministic constant. The proof bounds the perturbed partition function above and below by free-boundary partition functions shifted by e^{±β A_L}, then uses a superadditivity argument for the quenched free-boundary pressure and a variance bound for self-averaging. The hypotheses are checked for random scalar surface fields, fixed random exterior spins, and periodic wrap-around bonds, each with expected boundary corrections of order 1/L. The result concerns only the free energy; it says nothing about convergence of finite-volume Gibbs measures.","feed_headline":"Boundary noise can't move the spin-glass free energy","feed_subtitle":"Random surface fields, exterior spins, and wrap-around bonds leave the limiting quenched free energy unchanged.","key_machinery":"The load-bearing object is the energy envelope A_L of the boundary perturbation and the deterministic comparison inequality it produces: if sup_σ |B_L(σ)| ≤ A_L, then e^{−β A_L} Z_L^{free} ≤ Z_L^B ≤ e^{β A_L} Z_L^{free}, so the sample free energies differ by at most A_L/V_L. Two background inputs complete the argument: first, the quenched free-boundary pressure is superadditive—convexity of log Z as a function of the cross-piece couplings yields P(Λ) ≥ Σ P(Λ_i), which gives the free-boundary thermodynamic limit; second, a variance estimate obtained by swapping one coupling or field for an independent copy gives Var(hat f_L^{free}) ≤ C/V_L, and a standard almost-sure argument using the summab","core_discovery":"The central discovery is that the thermodynamic limit of the quenched specific free energy is insensitive to any boundary perturbation whose total energy is subextensive. Concretely, for the Edwards–Anderson model on cubic boxes with zero-mean, finite-variance couplings, if sup_σ |B_L(σ)| ≤ A_L with E A_L/V_L → 0 and A_L/V_L → 0 almost surely, then there is a deterministic constant f∞ such that both the disorder average and the sample free energy converge to f∞, and the difference between the sample value and the mean tends to zero almost surely. The proof isolates a deterministic comparison lemma and verifies the envelope hypothesis for random scalar surface fields, fixed random exterior sp","pith_inferences":["The proof structure suggests the theorem extends beyond nearest-neighbor cubic boxes to any finite-range disordered system with zero-mean couplings and a boundary term whose total energy is subextensive; only the comparison lemma and self-averaging are essential.","If the zero-mean coupling assumption E J_e = 0 fails, the superadditivity step breaks down; whether the free-boundary limit still exists under non-zero-mean disorder is left open by this method.","A natural next step, left implicit, is whether the same subextensive perturbations that leave the free energy unchanged nevertheless select a particular Gibbs state; the paper's scope remark makes this the immediate follow-up.","A concrete numerical check follows from the paper's bounds: measure the difference between periodic and free quenched free energies in d = 2 and verify that it approaches zero as L grows; persistent deviations would indicate a missing term in the envelope argument."],"forward_implications":["If the theorem is correct, the limiting quenched specific free energy is a single deterministic constant shared by free, random-surface-field, random-exterior-spin, and periodic boundary conditions.","Sample-to-sample fluctuations of the specific free energy vanish almost surely along the full sequence of boxes in dimension at least two, not just in probability.","Expected boundary corrections decay as O(1/L), meaning finite-size free energies at side length L approach the thermodynamic limit with surface-to-volume errors.","For Gaussian disorder, the paper's Gaussian-concentration remark gives exponential concentration and full-sequence almost-sure self-averaging in every dimension d ≥ 1.","The result deliberately does not assert convergence of finite-volume Gibbs measures, so chaotic size dependence of the states remains compatible with the theorem."],"fun_headline_variants":["Subextensive boundary effects vanish in spin glass free energy","Spin glass free energy immune to weak surface perturbations","Edwards-Anderson free energy ignores subextensive boundary energy","Boundary perturbations with negligible energy don't change spin glass limit","Weak surface randomness can't move the Edwards-Anderson free energy"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the random bulk couplings have mean zero; the proof of the free-boundary limit uses convexity of the log-partition function as a function of the cross-piece couplings, and that step requires the conditional mean of those couplings to vanish, so if E J_e ≠ 0 the superadditivity argument collapses and the paper gives no alternative proof of convergence.","fun_headline_variants_meta":{"raw":{"variants":["Subextensive boundary effects vanish in spin glass free energy","Spin glass free energy immune to weak surface perturbations","Edwards-Anderson free energy ignores subextensive boundary energy","Boundary perturbations with negligible energy don't change spin glass limit","Weak surface randomness can't move the Edwards-Anderson free energy"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000552,"raw_usage":{"total_tokens":2426,"prompt_tokens":657,"completion_tokens":1769,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":401,"completion_tokens_details":{"reasoning_tokens":1701}},"tokens_in":401,"tokens_out":1769,"duration_ms":12037,"temperature":1.0,"reasoning_tokens":1701,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T20:00:47.404739+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the quenched specific free energy difference between periodic and free boundary conditions in d = 2 with zero-mean couplings and check whether it approaches zero as L → ∞; if the difference does not approach zero, the envelope comparison fails. Alternatively, run the same model with couplings of non-zero mean and see whether the free-boundary specific free energy still converges to a boundary-independent limit; a single instance where it does not would expose the zero-mean premise.","supporting_citations":[],"review_version":1}