{"id":"981daa2a-1a70-4312-82ac-eed9558656c6","arxiv_id":"1908.07512","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper proves the conjectured critical fluctuations for the spherical SK model in an external field and establishes a three-parameter family of limits with a Curie-Weiss term.","lead":"A rigorous proof shows that the ground-state energy fluctuations of the spherical Sherrington-Kirkpatrick spin glass in a critically tuned external field converge to a new interpolating random distribution. The paper also adds a critical ferromagnetic term, producing a three-parameter family that contains previously known Tracy-Widom and Baik-Ben Arous-Peche-style limits.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.7 invokes Lemma 3.8 for an H^1 limit as stated; the gap is real but repairable by an H^1 version of Lemma 3.8.","rationale":"The reader's weakest assumption is the correct place to stress-test: Lemma 3.7 is the new discrete-to-continuous convergence result on which Proposition 3.3, and therefore the identification of the limiting point process, rests. As printed, Lemma 3.8 is misapplied because its C^1 hypothesis is not verified for f^{[1]} = f' - yf. However, the gap does not appear fatal: the conclusion of Lemma 3.8 only needs the limit in H^1_loc, and the proof can be adapted using Cauchy-Schwarz and the compact embedding H^1(I) into C^{1/3}(I). Thus the central argument likely survives after a routine repair, so the reader's CONDITIONAL verdict is appropriate and unchanged. Separately, the reader's other observation is also supported: Theorem 1.3 states N^{1/2}(1-\\mu_N) -> w, but the proof and the paper's own introduction require N^{1/3}(1-\\mu_N) -> w (i.e., w_N = N^{1/3}(1-\\mu_N)). With the printed N^{1/2} scaling, w_N -> 0 for every finite w, so the claimed w-dependence of the limit cannot hold. This is a localized statement-level typo, not a defect in the method, and is fixable by changing N^{1/2} to N^{1/3}. Both issues support CONDITIONAL, not ACCEPT or REJECT.","tokens_in":22891,"tokens_out":28888,"duration_ms":787074,"concrete_test":"Re-prove Lemma 3.7 with Lemma 3.8 weakened to H^1_loc limits: in the final estimate of the proof, replace y_2(x) m_N^{-1} sup_I |f'| by y_2(x) m_N^{-1/2} ||f'||_{L^2(I)} (plus the boundary term y_2(x)|f(x)| near 0). If the modified proof yields compact-uniform convergence of D_N f_N to f' under the stated hypotheses, the regularity objection is settled as a repairable gap; if the estimate fails, Proposition 3.3 lacks a complete proof as written.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.5: in the proof of Lemma 3.7, the sequence f_N^{[1]} = D_N f_N - (y_N^1)_\\times f_N - (y_N^2)_\\times 2^{-1}(T_N+T_N^*)f_N is shown to converge locally weakly in L^2 to f^{[1]} = f' - yf, and D_N f_N^{[1]} converges locally weakly to (f^{[1]})'. The text then applies Lemma 3.8 to conclude compact-uniform convergence. Lemma 3.8 is stated only when the limit function is C^1, and its proof uses sup_I |f'|. The limit f^{[1]} is known only to lie in H^1_loc (continuity is automatic in 1D, but C^1 is not). Thus the invocation is not justified as printed. This step is load-bearing: Proposition 3.3, and hence the identification of the limiting point process, depends on the compact-uniform convergence of the resolvent entries. The gap is repairable: Lemma 3.8 remains true for f in H^1_loc, because the estimate |g_N(x)-f_N(x)| ≤ m_N^{-1}|g_N'(x)| combined with weak L^2 convergence and the compact embedding H^1(I) ↪ C^{1/3}(I) gives uniform convergence; the final sup|f'| bound in Lemma 3.7 can be replaced by y_2(x) m_N^{-1/2} ||f'||_{L^2}. The manuscript should state and prove that weaker lemma.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies the spherical Sherrington-Kirkpatrick model in an external field, and a joint external-field/Curie-Weiss critical regime, for general inverse temperature β. The main results identify the fluctuations of the low-lying critical values of the Hamiltonian. Theorem 1.1 states that when N^{1/6}h_N→h, the top N-k critical values, shifted by N^{2/3}(1+h_N^2/2), converge to a point process Λ^{h,k}_β built from the β-stochastic Airy operator; in particular the ground-state fluctuation N^{2/3}(1+h_N^2/2-E^β_{N,h_N}) converges to -TW^h_β. Theorem 1.3 extends this to a joint critical regime with a ferromagnetic Curie-Weiss term, producing a three-parameter family Λ^k_{β,w,h}. The proof passes through a Lagrange-dual representation, a general convergence theorem for spiked tridiagonal ensembles (Proposition 3.1), a discrete-to-continuum resolvent-entry result (Proposition 3.3), and new stochastic-operator facts in Section 4.","tokens_in":23212,"tokens_out":22817,"duration_ms":215930,"significance":"If the technical gaps are repaired, the paper confirms predictions of Baik-Lee and Fyodorov-Le Doussal and extends the Bloemendal-Virág spiked-random-matrix transition to the spherical SK setting. The limiting distributions are constructed from eigenfunctions of the stochastic Airy operator, not fitted to finite-N data, and the known Tracy-Widom and Bloemendal-Virág limits arise as special cases. The point-process-level statement is stronger than ground-state convergence alone and covers all β>0, with complex and quaternionic versions via the Householder reduction. The proof strategy is coherent and builds seriously on [4,14,18]. However, two load-bearing issues in the printed version must be addressed: the stated μ-scaling in Theorem 1.3 is inconsistent with the proof, and the proof of Lemma 3.7 invokes Lemma 3.8 outside the hypotheses of that lemma.","major_comments":[{"comment":"The stated scaling of the ferromagnetic coupling in Theorem 1.3 is inconsistent with the proof. The theorem requires N^{1/2}(1-μ_N)→w, but in the reduction in Section 3.3 the spike parameter is w_N=N^{1/3}(1−μ_N). Under the theorem's stated scaling, w_N→0 for every finite w, so the limiting process would be the w=0 member of the family Λ^k_{β,w,h}, not the claimed Λ^k_{β,w,h} with arbitrary w. The statement should read N^{1/3}(1−μ_N)→w, the scaling used in [4] and in the rest of the paper; with the current text, Eq. (22) and the subsequent application of Proposition 3.1 do not match.","section":"Theorem 1.3, Eq. (22); Section 3.3"},{"comment":"In the proof of Lemma 3.7, Lemma 3.8 is applied to f_N^{[1]} with limit f^{[1]}=f'−yf. At that point only f∈D_max is known, which gives f^{[1]}∈H^1_loc; the limit need not be C^1. Lemma 3.8 as stated assumes f∈C^1, and its proof uses sup_I|f'|, so the invocation is not justified as printed. This is load-bearing: the compact-uniform convergence of f_N^{[1]} is what upgrades the weak L^2 convergence of the discrete quasi-derivatives to the resolvent-entry convergence in Proposition 3.3, and hence to the limiting point process. The gap appears repairable by proving an H^1 version of Lemma 3.8 (for instance, combining |g_N(x)-f_N(x)|≤m_N^{-1}|g_N'(x)| with weak L^2 convergence and Morrey compactness, and replacing sup|f'| by y_2(x)m_N^{-1/2}||f'||_{L^2}); the manuscript should state and prove that version.","section":"Section 3.5, Lemma 3.7; Lemma 3.8; Proposition 3.3"}],"minor_comments":[{"comment":"The displayed relation in Eq. (65) appears to be N^{2/3}(2−H/N)=L, but the preceding calculation gives N^{2/3}(1−H/N)=L. As printed the constant would affect the h_N^2/2 shift in Theorem 1.1; this should be corrected to match the derivation.","section":"Section 3.3, Eq. (65)"},{"comment":"The text states that for fixed h, √N(E_N,h−√(1−h²)) converges to a Gaussian limit. Given the immediately preceding statement that E_N,h→√(1+h²), the formula should be √(1+h²) rather than √(1−h²).","section":"Section 1, paragraph after Eq. (7)"},{"comment":"The word 'quarternionic' should be 'quaternionic'.","section":"Abstract and Introduction"},{"comment":"The notation R^* = R ∪ {∞} is potentially confusing because R^* is commonly used for the nonzero reals; a different symbol such as R_* would avoid ambiguity.","section":"Section 1.1, notation"},{"comment":"The phrase 'a w-spiked (∞-spike) tridiagonal ensemble' is awkward; it should say that the pure SK case corresponds to w=∞, while the Curie-Weiss case corresponds to finite w when the scaling is corrected.","section":"Section 3.3, paragraph after Eq. (58)"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a real proof of a conjectured critical fluctuation theory, not a numerical or heuristic paper. It confirms the N^{1/6} h_N critical window predicted in [3,12], produces the interpolating Tracy-Widom family, and adds the joint Curie-Weiss/external-field transition. The main theorems look right in broad strokes, and the machinery is genuinely new: the resolvent-entry convergence in Proposition 3.3 is the key technical step, and it goes beyond the L^2 resolvent convergence in [18,4]. That alone is a solid contribution.\n\nThe paper earns credit for building the limit objects from eigenfunctions of the stochastic Airy operator rather than fitting finite-N data. The special cases h=0 and fixed h fall out, so the family is not reverse-engineered. The citation pattern is honest; the paper clearly states what was known and what was conjectured.\n\nNow the soft spots, and they are exactly where the reader put them. Theorem 1.3 as printed has an inconsistent scaling: it states N^{2/3}(2 - Crit_{N-k,N}(H^beta_{N,mu,h})) with N^{1/2}(1 - mu_N) -> w, but the proof and the reference [4] use N^{1/3}(1 - mu_N) -> w. This is a typo-level issue but it is load-bearing for anyone trying to state the theorem; it must be fixed before publication.\n\nMore serious: the proof of Lemma 3.7 applies Lemma 3.8 to a limit f^{[1]} = f' - yf that is only known to lie in H^1_loc, while Lemma 3.8 is stated for C^1 limits and its proof uses sup|f'|. That is a genuine gap in the printed argument. The stress-test note is right that it is repairable: a standard compactness argument (weak L^2 convergence plus the discrete inequality |g_N - f_N| <= m_N^{-1}|g_N'| and Morrey embedding) gives uniform convergence for H^1 limits. But the manuscript needs to state and prove that weaker lemma. Since Proposition 3.3 depends on this step, the proof as written is not complete.\n\nNeither issue undermines my belief in the result. The strategy is coherent, the technical difficulties are engaged head-on, and the gaps look like repairs, not holes. For a reader working on random matrix edge statistics or spherical spin glasses, this is worth serious engagement. For a referee, I would send it out, with instructions to demand the scaling fix and the H^1 version of Lemma 3.8.","headline":"The paper proves the conjectured critical fluctuations for the spherical SK model in an external field, with a repairable regularity gap in Lemma 3.7 and a scaling typo in Theorem 1.3.","tokens_in":23795,"tokens_out":1717,"would_cite":true,"duration_ms":15901,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60B20","82B44","60K35","82B27"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves a critical external-field window in which spherical SK ground-state fluctuations interpolate between Tracy–Widom and Gaussian laws, via the stochastic Airy operator.","keywords":["spherical Sherrington-Kirkpatrick model","Tracy-Widom distribution","external field","critical fluctuations","stochastic Airy operator","beta ensemble","Weyl-Titchmarsh m-function","random matrix theory"],"falsifier":"Construct a sequence $f_N$ in the discrete quasi-derivative setting such that $f_N$ and $D_N f_N$ converge weakly in $L^2$ to an $H^1_{\\rm loc}$ limit $f$ with $f'\\notin C^0$, yet the interpolating remainder $f_N-g_N$ does not vanish compact-uniformly; this would disprove the compact-uniform convergence used in Proposition 3.3. Alternatively, check whether Lemma 3.8 remains true for $H^1_{\\rm loc}$ limits: if it does, the gap is repairable, and if it does not, the limiting point-process identification lacks a proof.","tokens_in":22635,"feed_emoji":"🎲","tokens_out":12862,"duration_ms":114438,"temperature":0.7,"pith_summary":"External fields change the way the spherical Sherrington–Kirkpatrick model's ground-state energy fluctuates, and this paper pinpoints the transition. When the field is scaled as $N^{1/6}h_N\\to h$, the paper proves that $N^{2/3}(1+\\tfrac12 h_N^2-E^\\beta_{N,h_N})$ converges in law to a one-parameter family $-TW^h_\\beta$ that reduces to the Tracy–Widom law at $h=0$ and connects to the Gaussian regime for fixed fields. The same approach proves a stronger statement: the $N^{2/3}$-rescaled point process of the top critical values converges to an explicit limit $\\Lambda^{h,k}_\\beta$. A parallel theorem treats a joint critical regime with a Curie–Weiss ferromagnetic term, producing a three-parameter family $TW^h_{\\beta,w}$. Both theorems are proved at the level of the $\\beta$-ensemble analogue, so they cover the real, complex, and quaternionic spherical SK models.","feed_headline":"External field bridges Tracy–Widom and Gaussian SK fluctuations","feed_subtitle":"For $N^{1/6}h_N\\to h$, the ground-state energy fluctuates at order $N^{-2/3}$ with an interpolating law.","key_machinery":"The engine is the stochastic Airy operator $$A^\\$\\beta$ = -\\frac{$d^{2}$}{$dx^{2}$}+x+\\frac{2}{\\sqrt{\\$\\beta$}}B'_x$$ on $L^2(\\mathbb R_+)$, with $w$-Robinson boundary condition $w\\phi(0)=\\phi'(0)$ (and $\\phi(0)=0$ when $w=\\infty$); its eigenvalues are the limiting edge eigenvalues of the rescaled $\\beta$-Hermite ensemble. The paper's new input is to control the resolvent entry appearing in the Lagrange-dual formula for the ground-state energy. The central object is the Weyl solution $\\phi^w_\\lambda$, the unique meromorphic family solving $A^\\beta\\phi=\\lambda\\phi$ with normalization $w\\phi^w_\\lambda(0)+1=(\\phi^w_\\lambda)'(0)$, whose boundary value is the Weyl–Titchmarsh $m$-function. The discrete quasi-derivatives $D_N^{[1]},D_N^{[2]}$ and the compactness lemma that convert weak $L^2$ convergence of discrete resolvent columns into compact-uniform convergence of their first entries carry the proof; Proposition 3.3 is the analytic heart.","core_discovery":"The central claim is that the critical fluctuations of the spherical SK ground state are governed by the stochastic Airy operator, not just through its eigenvalues but through a boundary object: the Weyl solution. For each $\\lambda$ off the spectrum, the relevant quantity is the boundary value or boundary derivative of the unique solution $\\phi^w_\\lambda$ of $A^\\beta\\phi=\\lambda\\phi$ with $w$-Robinson data, and the limiting fluctuation law is obtained by optimizing $\\tfrac12(\\lambda-h^2\\phi^w_\\lambda(0))$ over $\\lambda$ below the first eigenvalue. The paper proves the compact-uniform convergence of the corresponding discrete resolvent entries to these Weyl data, and from that convergence derives the point-process limits in Theorems 1.1 and 1.3. In the Curie–Weiss case the same machinery yields the three-parameter family $TW^h_{\\beta,w}$, subsuming the two-parameter family of the spiked random-matrix transition.","pith_inferences":["Extension beyond the paper: the resolvent-entry convergence method suggests that the same Weyl–Titchmarsh data control rank-one perturbations of general spiked tridiagonal ensembles at the edge, so the limiting laws should be universal across ensembles with the same stochastic Airy limit.","Testable extension: for $h_N=hN^{-1/6}$, empirical quantiles of $N^{2/3}(1+h_N^2/2-E_{N,h_N})$ should track the quantiles of $-TW^h_\\beta$; this prediction is sharp enough for a direct simulation check.","Conjectural connection: the three-parameter family $TW^h_{\\beta,w}$ plausibly interpolates between the spiked-random-matrix phase transition and the Tracy–Widom law, making the Curie–Weiss critical window a spherical analogue of that transition."],"forward_implications":["For every $\\beta>0$ and every $h$, the ground-state fluctuation law $-TW^h_\\beta$ interpolates between the Tracy–Widom law at $h=0$ and the Gaussian fluctuations of the fixed-field regime.","The $N^{2/3}$ scaling applies to the entire top end of the critical-value landscape: the point process of the $k$ largest critical values has an explicit limiting law $\\Lambda^{h,k}_\\beta$, not just its minimum.","With a critical Curie–Weiss term, the limiting fluctuation is a three-parameter distribution $TW^h_{\\beta,w}$, extending the two-parameter family previously known without an external field.","Because the theorems are proved for the $\\beta$-ensemble analogue, they apply simultaneously to the real ($\\beta=1$), complex ($\\beta=2$), and quaternionic ($\\beta=4$) spherical SK models."],"supporting_citations":[{"why":"Supplies the stochastic Airy operator and the convergence of the rescaled beta-Hermite matrices to its spectrum, the backdrop for the resolvent-entry result.","marker":"[18]"},{"why":"Introduces the spiked tridiagonal ensembles and the two-parameter limiting family that the Curie-Weiss theorem generalizes.","marker":"[4]"},{"why":"Makes the prediction for critical fluctuations with external field that Theorem 1.1 confirms.","marker":"[3]"},{"why":"Predicts the ground-state fluctuation crossover that Theorem 1.1 confirms.","marker":"[12]"},{"why":"Establishes the zero-field Tracy-Widom limit that the new one-parameter family interpolates from.","marker":"[22]"},{"why":"Provides the beta-Hermite tridiagonal models and Householder reduction that make the beta-ensemble generality and the real/complex/quaternionic corollaries possible.","marker":"[8]"},{"why":"Gives the norm-resolvent convergence of the discrete operators to the continuum operator used in the proof.","marker":"[14]"},{"why":"Supplies the Lagrange-duality correspondence between critical points of the quadratic Hamiltonian and of its dual function.","marker":"[11]"},{"why":"Provides the Weyl-Titchmarsh theory for Sturm-Liouville operators with distributional potentials from which the Weyl solutions and m-function are taken.","marker":"[9]"}],"fun_headline_variants":["SK fluctuations in a field: critical regime proven","External field interpolates SK fluctuations from TW to Gaussian","Three-parameter family captures SK critical fluctuations in field","Stochastic Airy operator yields new SK critical laws","Criticality in spherical SK with field: exact fluctuation laws"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing step is the compact-uniform convergence of the discrete resolvent entry, and that step applies a compactness lemma (Lemma 3.8) to a limit function whose first derivative is only known to be locally square-integrable, while the lemma is stated for continuously differentiable functions; if that regularity gap cannot be closed, the identification of the limiting point process would need a new argument.","fun_headline_variants_meta":{"raw":{"variants":["SK fluctuations in a field: critical regime proven","External field interpolates SK fluctuations from TW to Gaussian","Three-parameter family captures SK critical fluctuations in field","Stochastic Airy operator yields new SK critical laws","Criticality in spherical SK with field: exact fluctuation laws"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000411,"raw_usage":{"total_tokens":2081,"prompt_tokens":851,"completion_tokens":1230,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":467,"completion_tokens_details":{"reasoning_tokens":1154}},"tokens_in":467,"tokens_out":1230,"duration_ms":11917,"temperature":1.0,"reasoning_tokens":1154,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:07:35.712486+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct a sequence $f_N$ in the discrete quasi-derivative setting such that $f_N$ and $D_N f_N$ converge weakly in $L^2$ to an $H^1_{\\rm loc}$ limit $f$ with $f'\\notin C^0$, yet the interpolating remainder $f_N-g_N$ does not vanish compact-uniformly; this would disprove the compact-uniform convergence used in Proposition 3.3. Alternatively, check whether Lemma 3.8 remains true for $H^1_{\\rm loc}$ limits: if it does, the gap is repairable, and if it does not, the limiting point-process identification lacks a proof.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the stochastic Airy operator and the convergence of the rescaled beta-Hermite matrices to its spectrum, the backdrop for the resolvent-entry result."},{"cited_title":"and Vir´ ag, B","cited_arxiv_id":null,"evidence_quote":"Introduces the spiked tridiagonal ensembles and the two-parameter limiting family that the Curie-Weiss theorem generalizes."},{"cited_title":"and Lee, J","cited_arxiv_id":null,"evidence_quote":"Makes the prediction for critical fluctuations with external field that Theorem 1.1 confirms."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Predicts the ground-state fluctuation crossover that Theorem 1.1 confirms."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the zero-field Tracy-Widom limit that the new one-parameter family interpolates from."},{"cited_title":"and Edelman, A","cited_arxiv_id":null,"evidence_quote":"Provides the beta-Hermite tridiagonal models and Householder reduction that make the beta-ensemble generality and the real/complex/quaternionic corollaries possible."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the norm-resolvent convergence of the discrete operators to the continuum operator used in the proof."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Lagrange-duality correspondence between critical points of the quadratic Hamiltonian and of its dual function."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Weyl-Titchmarsh theory for Sturm-Liouville operators with distributional potentials from which the Weyl solutions and m-function are taken."}],"review_version":1}