{"id":"06a6a82d-a10a-4df1-af47-57d05ed93c42","arxiv_id":"2608.04107","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"The paper reports closed-form and apparently exact formulas, involving Eisenstein series and Eichler integrals, for the nonperturbative completion of the ABJM twisted superpotential at levels 1, 2, and 4.","lead":"This paper pushes numerical data from a quantum field theory called ABJM to very high precision, then uses machine learning and number-theory algorithms to guess exact mathematical formulas for a function called the twisted superpotential. The proposed formulas, if proven, would give a complete and unusually simple description of the nonperturbative corrections that feed into supersymmetric black hole entropy in string theory.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (12) as printed is inconsistent with the modular completion (15): for k=1, m=2 it gives c_2=-100 while (15) gives -25/4; the stated odd-prime growth |c_ℓ|≍ℓ selects m^{-2}, not m^2.","rationale":"The reader's weakest assumption was the extrapolation from numerically reconstructed data to all N, which is real and well flagged in the paper. However, the more immediately decisive issue is internal: as printed, Eq. (12) is inconsistent with Eq. (15), the very identity claimed to be equivalent to it term by term, and with the paper's own stated growth |c_ℓ| ≍ ℓ for odd primes. This is not a question of proof versus numerical evidence; it is a coefficient-level algebraic check. If the printed m^2 is a typo for m^{-2}, then the intended modular completion is consistent and the conditional verdict stands. If it is not a typo, then the central formula (10)-(12) does not match the Eichler completion (15), and the claimed 10^{-797} agreement cannot be attributed to the printed equations. The concrete test isolates the first conflicting coefficient and settles the matter without new physics.","tokens_in":13966,"tokens_out":22777,"duration_ms":188384,"concrete_test":"Analytically evaluate the coefficient of q^2 in Eq. (15) for k=1. With eE_2(-q) = 240[L3(-q) - (1/2)L3(q^2)], the q^2 coefficient of (1+tD)Φ1 is -(25/8π)t - (25/16π). Equating this to (1/2π)c_2(t+1/2) forces c_2 = -25/4, whereas Eq. (12) as printed gives c_2 = -100. This single coefficient decides whether the printed exponent should be m^{-2}; the 800-digit data at small N can then confirm the correct rational by isolating the m=2 sector from several ranks.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central formula (10)-(12) is supposed to be equivalent term by term to the Eichler-integral representation (15), but as printed the two disagree. From (14), eE_L(Q) has coefficient 240[σ3(m)-L^2σ3(m/L)]/m^3 on Q^m. For k=1, Φ1 = -1/(96π)eE_2(-q), so (1+tD)Φ1 has q^m coefficient -2.5/π (-1)^m [σ3(m)-4σ3(m/2)] (t/m^2 + 1/m^3). Writing this as (1/2π)c_m(t+1/m) forces c_m = 5(-1)^{m+1}[σ3(m)-4σ3(m/2)]/m^2. For k=4 the same computation forces c_m = (-1)^{m+1}[σ3(m)-16σ3(m/4)]/m^2. Eq. (12) instead prints c_m = (-1)^{m+1} m^2 times those bracket combinations, a factor m^4 larger. Concretely, at k=1, m=2, Eq. (12) gives c_2 = -100, while (15) gives -25/4. The paper's own asymptotic sentence, that for odd primes ℓ one has |c_ℓ^{(k)}| ≍ ℓ, also selects m^{-2}, since with the printed m^2 one would get |c_ℓ| ≍ ℓ^5. Thus either the exponent in (12) is a typo for m^{-2}, or Eqs. (11)-(12) and (15) cannot simultaneously be true.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes the on-shell effective twisted superpotential of U(N)_k × U(N)_-k ABJM theory at the universal twist and the dominant Bethe vacuum, using high-precision numerical data (up to 800 digits) and machine-learning techniques (PSLQ integer-relation detection and sparse regression). It claims that the perturbative part terminates after a shifted-rank 3/2-power plus an N-independent constant map C(k) with closed-form expressions for every integer k (Eqs. (4)–(5), equivalently (6)), that the fixed-λ genus expansion has an explicit all-genus coefficient formula (7)–(9) and is asymptotic, and that for k=1,2,4 the entire finite-N remainder is an absolutely convergent divisor-sum q-series (10)–(12), equivalent term-by-term to the Eichler-integral representation (15) of a weight-four Eisenstein series on Γ0(2) or Γ0(4). The claimed numerical precision is extremely high: residuals reach ~10^-797 across 297 data points. The paper is candid about the status of its statements: the constant map and divisor coefficients are arithmetically reconstructed, and the proof of the all-level integral reduction is deferred to a companion paper.","tokens_in":14480,"tokens_out":5095,"duration_ms":44606,"significance":"If the central claims hold, the result is significant: it would show that the nonperturbative completion of the ABJM twisted superpotential at special levels is a single absolutely convergent Eisenstein–Eichler q-series rather than an asymptotic multi-instanton transseries, and it would provide closed-form constant-map and all-genus data relevant to supersymmetric AdS4 black-hole entropy. The paper's strengths include the extraordinary numerical agreement (800 digits, systematically cross-checked), the independent numerical audit of the Clausen, Euler, and integral representations using no Bethe data (Supplemental Material, Eq. (25)), and the explicit admission, in the text, of which statements are derived and which are reconstructed. These strengths make the conjectures credible, but they do not replace the missing derivations.","major_comments":[{"comment":"","section":"Exact modular completion, Eq. (12) vs Eq. (15)"},{"comment":"","section":"Constant map and the deferred proof"},{"comment":"","section":"Exactness for all N and the status of Eq. (12)"}],"minor_comments":[{"comment":"","section":"Comparison with literature"},{"comment":"","section":"Supplementary Figure 3 caption"},{"comment":"","section":"Data availability"},{"comment":"","section":"Conventions for q and Q"}],"recommendation":"major_revision","confidential_remarks":"The internal inconsistency between Eq. (12) and Eq. (15) is the most urgent technical issue; it is almost certainly a typographical exponent error (m^2 vs m^{-2}) rather than a fundamental flaw, but it affects the central formula and must be corrected before publication. The two deferrals (the root-of-unity evaluation and the derivation of (12)) are candidly stated, but they mean the paper's headline claims are partly conjectural. The numerical evidence is unusually strong, so I would recommend major revision rather than rejection: the authors should correct (12), re-verify the numerics with the corrected formula, and adjust the abstract/title to accurately reflect the proven versus reconstructed status. I would also note that the companion paper [24] is essential for the 'every integer level' claim, and the editor may wish to monitor whether it actually appears with the promised proof."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is worth taking seriously. The constant-map formulas, the all-genus closed coefficients, and the special-level modular completion are genuinely new, and the 800-digit agreement across 297 data points is the strongest numerical evidence I have seen in this kind of Bethe-vacuum reconstruction. The paper is also admirably explicit about what is derived and what is reconstructed. That said, the central exact-completion claim as printed is internally inconsistent, and this is not a minor nit. In Eq. (12) the coefficient c_m is printed with a factor m^2 multiplying the divisor combination. In Eq. (15), the same object follows from the Eichler integral plus the (1+tD) operator, which forces a factor 1/m^2. Concretely, for k=1, m=2, Eq. (12) gives c_2 = -100, while Eq. (15) gives -25/4. That is a factor of m^4. The paper's own asymptotic statement—|c_ℓ| ~ ℓ for odd primes ℓ—also selects the m^{-2} version, since the printed m^2 would give ℓ^5. So either Eq. (12) has a typo (likely a missing 1/m^4 or an exponent -2 rather than +2), or the claimed term-by-term equivalence breaks down. As printed, the central formula cannot be true. I do not think this is fatal to the overall idea: the numerical data and the Eisenstein structure almost certainly mean the m^{-2} version is correct and the printed exponent is a transcription error. But the paper needs to be fixed before it can be accepted. Beyond that, my concerns are the ones the authors themselves flag: the proof of C(k) = C_int(k) is deferred to a companion paper, and the divisor structure is reconstructed, not derived. Code and data are not released, which matters for a paper whose method is high-precision reconstruction. None of this changes my judgment that the paper deserves a serious referee: the discovery is significant and the evidence, once the typo is corrected, would be strong. But my verdict would be conditional on the typo being fixed and the companion proof or independent reproduction appearing. This is a paper for the ABJM and supersymmetric index community; a reader who cares about exact modular completions will find it useful even in this form, mainly because the machinery is clearly explained. For my own work, I would not cite the exact formula until the inconsistency is resolved.","headline":"A strong numerical discovery with a central typo: Eq. (12) and Eq. (15) are off by m^4, so the paper as printed is not internally consistent.","tokens_in":14943,"tokens_out":2129,"would_cite":false,"duration_ms":19996,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11F11","81T13"],"pacs":[],"model":"deepseek-v4-flash","headline":"For ABJM theory at $k=1,2,4$, the nonperturbative remainder of the on-shell effective twisted superpotential is exactly one absolutely convergent divisor-sum $q$-series built from a weight-four Eisenstein–Eichler integral, reproducing…","keywords":["ABJM theory","twisted superpotential","Eisenstein series","Eichler integral","divisor-sum q-series","constant map","integer-relation detection","Bethe vacua"],"falsifier":"Compute the on-shell $\\operatorname{Im}W$ for a rank $N>100$ at $k=1,2,4$ with the same 800-digit precision and compare with (10)–(12); any residual above the data rounding floor, about $5\\times10^{-797}$, would falsify the exact completion. Alternatively, a direct derivation from the Bethe equations that produces a correction to (12) which is exponentially small on the computed ranks but nonzero for larger $N$ would also settle the claim.","tokens_in":13734,"feed_emoji":"🔢","tokens_out":12918,"duration_ms":100398,"temperature":0.7,"pith_summary":"This paper claims that the nonperturbative sector of a central quantity in three-dimensional supersymmetric field theory is simpler than the standard transseries picture suggests. In ABJM theory at Chern–Simons levels $k=1,2,4$, the exponentially small tail of the effective twisted superpotential is exactly a single absolutely convergent divisor-sum $q$-series. The paper also fixes the full perturbative part for every level and gives a closed all-genus coefficient formula. If the claims are right, the complete coefficient of $1/\\omega$ in the Cardy limit of the topologically twisted index is known in closed form at those levels, including finite-$N$ nonperturbative corrections, and the same exponentials become testable microscopic data for the $\\mathrm{AdS}_4$ black hole entropy function. The evidence is high-precision numerical discovery, reproducing 800-digit Bethe-vacuum data to about $10^{-797}$.","feed_headline":"One convergent series completes the ABJM superpotential tail","feed_subtitle":"A divisor-sum Eisenstein–Eichler series reproduces 800-digit Bethe data to 10^{-797}.","key_machinery":"The carrying object is the first-order operator $(1+tD)$ acting on Eichler integrals, where $D=Q\\,\\mathrm{d}/\\mathrm{d}Q$ multiplies the $m$-th Fourier coefficient by $m$. Weight-four Eisenstein combinations $E_4(Q)-L^2 E_4(Q^L)$ on $\\Gamma_0(L)$, with $L=2$ for $k=1,2$ and $L=4$ for $k=4$, have Fourier coefficients $\\sigma_3(m)-L^2\\sigma_3(m/L)$. Their coefficientwise Eichler primitive, the third antiderivative, sends $a_m Q^m$ to $a_m m^{-3}Q^m$. Applying $(1+tD)$ then turns $a_m/m^3$ into $a_m(t/m^2+1/m^3)$, reproducing the series in (11). Absolute convergence follows from $\\sigma_3(m)/m^2=O(m^{1+\\epsilon})$ and from the subtraction terms; the paper shows $\\limsup_m |c_m^{(k)}|^{1/m}=1$, so the physical remainder converges absolutely for $0<q<1$. This object carries the argument by packaging the entire exponentially small tail into classical modular data.","core_discovery":"The paper's central claim is that the on-shell effective twisted superpotential of $U(N)_k\\times U(N)_{-k}$ ABJM theory at the universal twist and the dominant Bethe vacuum, $\\operatorname{Im} W(N,k)$, decomposes exactly as $\\frac{\\pi^2}{3}\\sqrt{k/2}\\,\\hat b_N^{3/2}+C(k)+W_{\\mathrm{np}}(N,k)$. The perturbative part closes after two terms: the shifted-rank $3/2$ power with $\\hat b_N=N-k/24+2/(3k)$, and an $N$-independent constant map $C(k)$ given by parity-dependent finite Clausen sums at every integer level. At $k=1,2,4$ the entire finite-$N$ remainder is the absolutely convergent divisor-sum series $W_{\\mathrm{np}}(N,k)=\\frac{k^2}{2\\pi}\\sum_{m\\ge1} c_m^{(k)}(t+1/m)\\,q^m$, with $c_m^{(k)}$ given by divisor sums $\\sigma_3(m)$ with subtraction terms, equivalently by $(1+tD)\\Phi_k(-q)$, where $\\Phi_k$ is the Eichler integral of a weight-four Eisenstein combination on $\\Gamma_0(2)$ or $\\Gamma_0(4)$. This closed form reproduces 800-digit Bethe-vacuum data to about $10^{-797}$ for every rank $2\\le N\\le 100$ at those levels. The paper also gives a closed coefficient formula for the asymptotic type-IIA genus expansion at arbitrary genus, involving binomial coefficients and Bernoulli numbers.","pith_inferences":["If the Eisenstein–Eichler organization is not a numerical accident, the same operator $(1+tD)$ may appear in the nonperturbative completion of other twisted-superpotential observables at these levels, with the universal twist acting as a projection onto a modular subsector; the paper raises this as an interpretation and does not claim a transformation law for $(1+tD)\\Phi_k$.","Because the divisor coefficients are reconstructed rather than derived, a proof from the Bethe equations would likely reveal how root-of-unity dilogarithm identities assemble into Eisenstein coefficients; absent that proof, the sharpest test is to compute new ranks $N>100$ at $k=1,2,4$ and require the residual to stay on the $10^{-797}$ floor.","One could probe the specialness of $k=1,2,4$ by running the same discovery approach on other twists or on the topologically twisted index, where the paper notes a parallel constant map and a modular subsector appear in the companion work; if that modular structure persists, it would suggest a broader arithmetic organization rather than an accident of the universal twist."],"forward_implications":["At $k=1,2,4$, the nonperturbative instanton sector of $\\operatorname{Im}W$ is not an asymptotic expansion that must be stopped at an optimal order; the remainder converges absolutely for $0<q<1$, so the finite-$N$ answer can be evaluated to any precision by truncating (11).","The full Cardy-limit coefficient of the superconformal index at the universal twist is determined at those levels, including its exponentially small finite-$N$ tail, not just the leading $N^{3/2}$ term.","The series provides explicit exponential corrections that a quantum-corrected $\\mathrm{AdS}_4$ black hole entropy function would have to reproduce, giving a concrete microscopic target.","The constant map $C(k)$ is fixed in closed form for every positive integer level, so the previously undetermined $N$-independent constant in the on-shell twisted superpotential is no longer a free parameter.","The parity-dependent Clausen formulas and the level-dependent modular groups ($\\Gamma_0(2)$ for $k=1,2$ and $\\Gamma_0(4)$ for $k=4$) give arithmetic constraints that any exact derivation from the Bethe equations will have to explain."],"supporting_citations":[{"why":"Supplies the Bethe-equation form of the effective twisted superpotential and the dominant-vacuum assignment used to generate all data.","marker":"[26]"},{"why":"Supplies the perturbatively exact shifted-rank scaling that the parent formula (3) sharpens and extends.","marker":"[14]"},{"why":"Supplies earlier numerical estimates of the constant term at k=1,2,3,4 that the closed constant-map formulas reproduce.","marker":"[34]"},{"why":"Provides the integer-relation detection that certifies the constant-map and divisor-coefficient reconstructions.","marker":"[22]"},{"why":"Identifies the weight-four Eisenstein spaces on $\\Gamma_0(2)$ and $\\Gamma_0(4)$ that contain the divisor coefficient sequences.","marker":"[31]"},{"why":"Defines the Eichler-integral operation that converts weight-four Fourier coefficients into the $m^{-3}$ coefficients of the completion.","marker":"[32]"},{"why":"Fixes the worldsheet-instanton action scale that the exponent $t$ matches, and supplies the sphere instanton transseries the $q$-series contrasts with.","marker":"[11]"},{"why":"Supplies the Fermi-gas/Airy completion of the sphere partition function whose multi-sector transseries structure is contrasted with the single-series completion.","marker":"[13]"},{"why":"Is the companion paper to which the analytic root-of-unity proof of the all-level constant-map formula (6) is deferred.","marker":"[24]"}],"fun_headline_variants":["Divisor-sum series completes ABJM superpotential","Exact ABJM superpotential from Eichler integral","Eisenstein–Eichler series reproduces 800-digit ABJM data","Closed form for ABJM superpotential's non-perturbative tail"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The exactness claim rests on the assumption that the numbers reconstructed from high-precision data — the rank shift $\\hat b_N=N-k/24+2/(3k)$, the constant plateau $C(k)$, and the divisor-sum coefficients fitted for $N\\le 100$ — are exactly correct for all $N$ and all ranks, rather than merely extraordinarily accurate numerical coincidences on the computed range.","fun_headline_variants_meta":{"raw":{"variants":["Divisor-sum series completes ABJM superpotential","Exact ABJM superpotential from Eichler integral","Eisenstein–Eichler series reproduces 800-digit ABJM data","Closed form for ABJM superpotential's non-perturbative tail"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000541,"raw_usage":{"total_tokens":2680,"prompt_tokens":1120,"completion_tokens":1560,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":736,"completion_tokens_details":{"reasoning_tokens":1488}},"tokens_in":736,"tokens_out":1560,"duration_ms":10662,"temperature":1.0,"reasoning_tokens":1488,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:43:23.843084+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the on-shell $\\operatorname{Im}W$ for a rank $N>100$ at $k=1,2,4$ with the same 800-digit precision and compare with (10)–(12); any residual above the data rounding floor, about $5\\times10^{-797}$, would falsify the exact completion. Alternatively, a direct derivation from the Bethe equations that produces a correction to (12) which is exponentially small on the computed ranks but nonzero for larger $N$ would also settle the claim.","supporting_citations":[{"cited_title":"Ferguson, D","cited_arxiv_id":null,"evidence_quote":"Provides the integer-relation detection that certifies the constant-map and divisor-coefficient reconstructions."},{"cited_title":"Diamond and J","cited_arxiv_id":null,"evidence_quote":"Identifies the weight-four Eisenstein spaces on $\\Gamma_0(2)$ and $\\Gamma_0(4)$ that contain the divisor coefficient sequences."},{"cited_title":"Paşol and A","cited_arxiv_id":null,"evidence_quote":"Defines the Eichler-integral operation that converts weight-four Fourier coefficients into the $m^{-3}$ coefficients of the completion."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Is the companion paper to which the analytic root-of-unity proof of the all-level constant-map formula (6) is deferred."}],"review_version":1}