{"id":"38d64adf-a4da-43ca-a657-414e908cbfcb","arxiv_id":"2501.09717","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"One-loop massless corrections modify the null constraints of shift-symmetric scalar EFTs, and IR-finite combinations of them reproduce tree-level bounds in the weak-coupling limit while deforming bounds in five and six dimensions.","lead":"Massless quantum loops change the crossing-symmetry relations that positivity bounds use to constrain effective field theories, so loop-corrected rules are needed. The authors derive these corrected rules for one scalar field and compute the resulting allowed ranges for couplings in five and six spacetime dimensions.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Strongest bounds rely on ε=1 one-loop truncation with O(1) loop corrections at the reported g2 values.","rationale":"The reader's weakest_assumption identifies the truncation of the one-loop amplitude as the main risk, and I agree. The explicit acknowledgements in §3.2 and §6 confirm that the strongest ε=1 bounds are not guaranteed to be free of neglected two-loop and higher-derivative contributions. My stress-test therefore focuses on that single issue rather than on the internal structure of the moment-problem analysis, which appears consistent: the numerical LP and analytic moment-problem results match (Figs. 5–6), and the 4D weak-coupling limit reproduces tree-level bounds (Fig. 7), giving independent support to the framework. I considered whether the 4D log(µ) dependence in n46 (Eq. (32)) or the undocumented '8g4' notation indicated a deeper inconsistency, but these appear to be typographical and not load-bearing for the central claim. The two-loop test I propose would settle whether the O(1) loop corrections at the reported bounds invalidate the numbers. Since the paper itself flags the assumption, the appropriate verdict remains CONDITIONAL; no change to the reader's assessment is needed.","tokens_in":30144,"tokens_out":17276,"duration_ms":172544,"concrete_test":"Compute the two-loop order-g2³ correction to the 5D n4 null constraint and a2 sum rule (Eqs. (33)–(34)) using the two-loop massless scalar amplitude. Evaluate it at the Table 1 maximum g2 ≈ 2375.58 (g3=0, ε=1) and compare with the retained one-loop terms 3g2²/(2304π²) and 119g2²/(14336π²). If the two-loop term is not suppressed by at least an order of magnitude, the ε=1 bounds are not controlled by the one-loop truncation and Table 1 should be presented as ε-dependent.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's quantitative claims (Table 1 and Figs. 4–6) use modified null constraints derived from the one-loop amplitude Eq. (5), which keeps only g2² and g2g3 loop corrections. The n6 constraint (Eq. (30)) explicitly drops two-loop g2³ terms ('Assuming the weak coupling limit g2 → 0, we neglect these two-loop contributions'), and all β-functions of higher-order operators are omitted. The manuscript itself states in §3.2 that choosing ε=1 'provides the strongest bounds, however, in order to trust them, one has to assume that infinite number of contributions from further β-functions and multiloop corrections indeed can be neglected,' and in §6 that the upper bounds 'rely on assumption that the higher-derivative terms in EFT are suppressed.' At the reported maxima this assumption fails parametrically: Eq. (52) yields g2_max ≈ 2376 in 5D for g3=0, where the one-loop correction 119g2²/(14336π²) equals the tree term 2g2, so the loop expansion is O(1). The same applies to the 6D bound. Thus the central conceptual claim (massless loops deform null constraints) is credible, but the specific numbers are conditional on an unverified truncation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript extends the EFT-hedron / null-constraint program to one-loop order for a single massless shift-symmetric scalar. Starting from the one-loop amplitude in Eq. (5), it derives loop-modified null constraints, e.g. Eqs. (22), (29), (30) in four dimensions and Eqs. (33), (37) in five and six dimensions, in which tree-level zero combinations become proportional to beta-function-dependent quantities. Using extremal moment-problem methods and linear programming, it computes allowed regions for g2, g3 and g3/g2, g4/g2 in d=5,6 and constructs IR-finite combinations n45 and n46 to recover tree-level bounds in the weak-coupling limit in d=4. The central conclusion is that massless one-loop effects cannot be absorbed solely into running couplings, because they deform the crossing-symmetry relations.","tokens_in":30430,"tokens_out":9330,"duration_ms":100000,"significance":"If correct, the loop-deformed null constraints are a genuine extension of the positivity program and are likely to be useful for other EFTs with massless states. The paper's internal checks are strong: the 4D weak-coupling limit reproduces the known tree-level bounds (Fig. 7), and the analytic moment-problem results agree with the numerical linear-programming results in §5.1. Appendix A and B provide a careful distributional justification for the IR-finite combinations, addressing a real technical obstruction. The reason I do not recommend acceptance as is is that the headline quantitative bounds are stated as unconditional unitarity bounds although they rely on a truncation that the paper itself identifies as an assumption; the abstract and §4 overstate the robustness of Table 1 and Figures 4–6.","major_comments":[{"comment":"The headline bounds are obtained at ε=1, where the loop expansion and the EFT arc computation are both uncontrolled. The arc radius is ε²Λ² and the figure/§3.1 require ε<1 for the EFT computation to be valid, yet §4.1 sets ε=1 to maximize the bounds. At the reported 5D boundary, Eq. (52) gives g2,max ≈ 2375.58 for g3=0; inserting this value, the one-loop term 119g2²/(14336π²) is equal to the tree-level term 2g2 (and the same parametric failure occurs in 6D). Thus Table 1 and the associated 'upper bound on g2' statements are conditional on a regime in which the one-loop-truncated amplitude is not a reliable expansion. The paper acknowledges the general issue in §3.2 and §6, but the numerical results are still presented as unitarity bounds; they should be reframed or supplemented with results at smaller ε and an estimate of neglected higher-loop/higher-derivative terms.","section":"§3.1, §4.1, Eq. (52), Table 1"},{"comment":"The modified null constraints used for the strongest bounds are truncated: Eq. (30) explicitly drops two-loop g2³ contributions, and the computation omits the beta functions of all higher-order EFT coefficients. In §3.2 the paper states that only 'small enough ε' suppresses these contributions for granted, while ε=1 is then chosen in §4.1. The central conceptual claim — that null constraints acquire beta-function-dependent right-hand sides — does not depend on this truncation, but the quantitative allowed regions in Figures 4–6 do. Please separate the robust structural result from the numerical bounds, and state explicitly that the latter are illustrative under a weak-coupling/higher-order-suppression assumption.","section":"§3.2, Eq. (30); §6"},{"comment":"The 4D recovery of tree-level bounds is demonstrated only at g2=0.01. The combinations n45 and n46 are constructed after dropping two-loop g2³ terms, and they involve the very couplings (g3, g4) being bounded; their use as constraints is therefore a consistency check in the weak-coupling slice rather than a derivation valid at finite g2. Since §6 identifies two-loop log²(−t) terms as the obstruction to strong-coupling 4D bounds, the paper should state that the finite-combination method has been validated only where those terms are negligible.","section":"§5.2, Eqs. (31)–(32), Fig. 7"}],"minor_comments":[{"comment":"The printed expression contains apparent typesetting corruption with factors like '8g4' and 'g2 · 8g4'; please replace with the intended algebraic expression.","section":"Eq. (32)"},{"comment":"The cross-reference 'Appendix??' should point to Appendix C.","section":"§5.1"},{"comment":"The word 'maximun' should be 'maximum'.","section":"Table 1"},{"comment":"The claim that the moment-problem and linear-programming bounds 'match with almost no differences' would be easier to verify if the plot showed the difference or error bars.","section":"Fig. 5(b)"},{"comment":"The notation ε is used both for the dimensional-regularization parameter in Eq. (6) and for the arc-radius parameter beginning in §3.1; although the paper notes this, the double use makes equations like Eq. (30) difficult to read. A different symbol for the arc parameter would help.","section":"§3.1 and Eq. (6)"}],"recommendation":"major_revision","confidential_remarks":"I have no suspicion of misconduct; the manuscript is honest about its limitations. The main concern is that the abstract and main-text claims go beyond what the truncation supports. The paper fits the journal; with a revision that separates structural results from conditional bounds, it would be publishable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The central claim here is solid and worth taking seriously: massless one-loop effects do not just renormalize Wilson coefficients, they change the null constraints themselves, turning zeroes into beta-function-dependent combinations. The construction of IR-finite combinations n45 and n46, and the demonstration that they reproduce tree-level bounds in the weak-coupling limit, is a genuine step forward. The smearing appendix is also a useful methodological contribution for handling forward-limit singularities in dispersion relations. I would send this to a serious referee on that basis alone.\n\nThe soft spot is exactly where the stress test lands. The strongest bounds in Table 1 and Figs. 4–6 use epsilon = 1, and the one-loop truncation of the amplitude keeps only g2^2 and g2 g3 corrections. At the reported maxima, the one-loop correction is the same size as the tree term — for g2 ≈ 2376 in 5D, 119 g2^2/(14336 pi^2) ≈ 2 g2. That is not a small correction; it is an O(1) effect. The authors actually acknowledge this in §3.2 and §6, but the framing of the paper still presents these as the bounds. The honest characterization is that the conceptual mechanism is established, while the quantitative results are conditional on an unverified assumption that all higher-order beta functions and multiloop terms can be neglected. That is not a fatal flaw in the idea, but it should be front and center if this is published.\n\nMinor issues: there is a missing cross-reference in §5.1 and a few equations have typographical corruption (e.g., the repeated factors in n46). These are cosmetic and easily fixed.\n\nThe paper is internally consistent, reproduces known tree-level limits, and tackles an open problem that affects many published positivity bounds. It deserves peer review, but the referee should push the authors to either quantify the truncation error or explicitly reframe the numerical bounds as illustrative under a stated assumption. I would not cite the numerical values without that caveat.","headline":"The deformation of null constraints by massless loops is a real, credible result, but the quantitative bounds in 5D/6D rest on a truncation that fails parametrically at the reported maxima.","tokens_in":30932,"tokens_out":1399,"would_cite":true,"duration_ms":16159,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"One-loop massless corrections deform the null constraints of the shift-symmetric scalar EFT, changing unitarity bounds on dimension-8 and dimension-10 couplings.","keywords":["positivity bounds","null constraints","one-loop amplitudes","massless scalar EFT","crossing symmetry","dispersion relations","EFT-hedron","moment problem"],"falsifier":"Compute the two-loop contribution to the $n=6$ combination for the same shift-symmetric scalar: if the $g_2^3\\log^2(-t)$ term in $n_6$ is not negligible relative to the one-loop $g_2g_3$ terms at $\\epsilon=1$, then $n_{46}$ gains an uncancelled infrared divergence and the reported 4D weak-coupling bounds fail. A second check is to evaluate Eq. (22) at fixed negative $t$ in an explicit UV completion with known spectral density and compare the arc-integral side with the moment side; any mismatch would show the truncation used for the null-constraint right-hand side is incomplete.","tokens_in":29962,"feed_emoji":"🌀","tokens_out":8563,"duration_ms":85690,"temperature":0.7,"pith_summary":"The paper studies a single massless scalar field with shift symmetry and asks how one-loop self-interactions change the positivity bounds that follow from analyticity, causality, and unitarity. Its central claim is that the crossing-symmetry relations known as null constraints, which previous positivity bounds set to zero, acquire non-vanishing right-hand sides proportional to one-loop beta functions. That means massless loops cannot be absorbed just by renormalizing Wilson coefficients: the allowed region for dimension-8 and dimension-10 couplings shifts, and in five and six spacetime dimensions the upper bounds on g2 and g3 change, with the allowed region becoming dependent on g2. In four dimensions, infrared logarithms force the construction of special finite combinations of null constraints; in the weak-coupling limit these reproduce the tree-level bounds on g3/g2 and g4/g2. A sympathetic reader would care because most realistic EFTs have massless states, so the constraints used to test whether an EFT can have a UV completion need a loop-level version.","feed_headline":"Massless loops rewrite the null constraints behind EFT bounds","feed_subtitle":"One-loop corrections make null constraints beta-function dependent and shift unitarity bounds in 5D and 6D.","key_machinery":"The load-bearing object is the null constraint, a linear relation between positive moments of the UV spectral density that full crossing symmetry imposes on the amplitude; at tree level these relations are homogeneous equalities. The machinery is the dispersion-relation identity $M_n(t)=B_n(t)$ between EFT arc integrals and spectral integrals, applied at one-loop level to the shift-symmetric scalar amplitude. From this identity the paper extracts the moment combinations $n_4$, $n_5$, $n_6$; their tree-level zeroes become $\\beta$-function-dependent expressions. Because $t$-derivatives of the loop amplitude introduce $\\log(-t)$ and $1/t$ terms in $d=4$, the paper forms the IR-finite linear combinations $n_{45}=2g_2^2 n_5+36g_2g_3 n_4$ and $n_{46}=6(24g_2g_4+7g_3^2)n_4-7g_2^2 n_6$, whose forward limits are well defined and can be matched to the corresponding moment brackets. For the numerical bounds on $g_2$ and $g_3$ in $d=5,6$ it uses the L-moment problem with full unitarity $0<\\rho_j<2$, whose allowed region is obtained as a Minkowski sum over angular momenta.","core_discovery":"The discovery is that the null constraints of the EFT-hedron are deformed by massless one-loop physics rather than destroyed. At tree level, full crossing symmetry forces certain brackets of partial-wave moments to vanish, for example the n=4 constraint $J^2(J^2-8)/\\mu^5$ with $J^2=j(j+1)$. At one loop the same combination equals $2b_2\\log(\\epsilon^2)-2b_2\\log(-t)-3b_2+2c_2\\epsilon^2$, with $b_2=g_2^2/(240\\pi^2)$ and $c_2=-g_2g_3/(240\\pi^2)$. In $d=5$ and $d=6$ the analogous n=4 constraints become $g_2\\sqrt{\\epsilon^2}(3g_2-g_3\\epsilon^2)/(2304\\pi^2)$ and $g_2\\epsilon^2(2g_2-g_3\\epsilon^2)/(13440\\pi^3)$. Because these right-hand sides involve the couplings themselves, the relation between EFT coefficients and positive moments becomes nonlinear, and unitarity bounds on $g_2$, $g_3$, and $g_4/g_2$ are modified; in the weak-coupling limit the tree-level constraints are recovered. The paper establishes this by computing the one-loop amplitude, building IR-finite combinations of arc integrals, and solving the resulting moment problem with full unitarity in $d=5$ and $d=6$, while in $d=4$ it shows that the finite combinations $n_{45}$ and $n_{46}$ reproduce the tree-level bounds when $g_2$ is small.","pith_inferences":["Going beyond the paper: if one-loop deformations of null constraints are generic, similar beta-function terms should appear in other massless EFTs, where they would link unitarity bounds to the running of dimension-8 operators.","Going beyond the paper: the 4D construction suggests a general recipe of taking IR-finite linear combinations of null constraints before matching to moments, which could be tested against explicit UV completions with heavy resonances where the spectral density is known.","Going beyond the paper: whether the $\\epsilon=1$ bounds survive depends on infinite higher-order beta functions; a numerical evaluation of two-loop terms at fixed arc radius would settle whether the reported $g_2$ maxima are a truncation effect."],"forward_implications":["In $d=5$ and $d=6$, the upper bound on $g_2$ from full unitarity becomes finite and $g_2$-dependent; for $g_3=0$ the paper finds max $g_2\\approx 2375.6$ in 5D and $\\approx 27739.4$ in 6D, with different values for $g_3=4g_2$ and $g_3=-10g_2$.","In the weak-coupling regime, the loop-level bounds on $g_3/g_2$ and $g_4/g_2$ coincide with tree-level bounds; in 5D, for $g_2=100$ the loop-corrected region is nearly identical to the tree-level one.","In 4D, the forward limit of $n_4$ contains $\\log(-t)$, so the raw null constraint cannot be used; the two IR-finite combinations $n_{45}$ and $n_{46}$ are sufficient, with $g_2=0.01$, to reproduce the tree-level bound on $g_3/g_2$ and $g_4/g_2$.","When $g_2$ is larger, the allowed region in the $g_3/g_2$--$g_4/g_2$ plane deviates from the tree-level region and becomes non-convex, because the map from moments to Wilson coefficients is nonlinear at loop level."],"supporting_citations":[{"why":"Defines the null constraints from full crossing symmetry at tree level that this paper deforms at loop level.","marker":"[5]"},{"why":"Establishes the positive-moment and arc-integral dispersion relation formalism that the loop-level analysis extends.","marker":"[3]"},{"why":"Shows that loops can prevent absorbing IR effects purely as running of Wilson coefficients, the starting point for the massless analysis.","marker":"[37]"},{"why":"Provides the de-projected EFT-hedron and Minkowski-sum method with full unitarity used to bound $g_2$.","marker":"[10]"},{"why":"Supplies the one-loop amplitude of the shift-symmetric scalar used as the low-energy input.","marker":"[70]"},{"why":"Gives the analytic moment-problem bounds on $g_3/g_2$ and $g_4/g_2$ that are reproduced in the weak-coupling limit.","marker":"[9]"},{"why":"Discusses full unitarity and moments, used here as a comparison for the non-projective loop bounds.","marker":"[56]"}],"fun_headline_variants":["One-loop corrections bend the EFT-hedron's null constraints","Massless loops make EFT bounds loop-dependent","Null constraints get a one-loop twist in EFT","Loop-level EFT bounds go beyond weak coupling"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the one-loop amplitude with only the beta functions of $g_2$ and $g_3$ included describes the low-energy side of the dispersion relations well enough at the chosen arc radius; if two-loop corrections or the running of higher-order couplings are not negligible, the modified null constraints and the bounds built on them change.","fun_headline_variants_meta":{"raw":{"variants":["One-loop corrections bend the EFT-hedron's null constraints","Massless loops make EFT bounds loop-dependent","Null constraints get a one-loop twist in EFT","Loop-level EFT bounds go beyond weak coupling"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000367,"raw_usage":{"total_tokens":2081,"prompt_tokens":1165,"completion_tokens":916,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":781,"completion_tokens_details":{"reasoning_tokens":853}},"tokens_in":781,"tokens_out":916,"duration_ms":9834,"temperature":1.0,"reasoning_tokens":853,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T19:43:32.279620+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the two-loop contribution to the $n=6$ combination for the same shift-symmetric scalar: if the $g_2^3\\log^2(-t)$ term in $n_6$ is not negligible relative to the one-loop $g_2g_3$ terms at $\\epsilon=1$, then $n_{46}$ gains an uncancelled infrared divergence and the reported 4D weak-coupling bounds fail. A second check is to evaluate Eq. (22) at fixed negative $t$ in an explicit UV completion with known spectral density and compare the arc-integral side with the moment side; any mismatch would show the truncation used for the null-constraint right-hand side is incomplete.","supporting_citations":[],"review_version":1}