{"id":"605a7567-d2b5-4b6f-8c76-942d5cb42d7d","arxiv_id":"1909.02003","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Four-dimensional N=8 supergravity loop integrands scale one power better at infinity than general-D power-counting predicts, and this homogeneous scaling combined with BCFW behavior uniquely fixes the integrand through three loops at four points and at one loop for all multiplicities.","lead":"This paper shows that gravity scattering amplitudes behave better than expected at very high loop momenta in four dimensions, and that this improved behavior can fix the full loop integrand of N=8 supergravity. A smart generalist might read it because these ultraviolet constraints hint at a hidden geometric structure for gravity amplitudes, analogous to the amplituhedron of planar N=4 super Yang-Mills theory.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Uniqueness of the reconstructed N=8 integrands rests on an unaudited 2757/2758 parameter count; Section 4.3 shows the same homogeneous-cut logic fails at seven loops, so the low-loop completeness claim needs independent rank verification.","rationale":"The paper has two main thrusts: (i) the four-dimensional improved scaling of gravity cuts (Eq. 3.9) and (ii) the reconstruction of integrands from homogeneous scaling constraints. Thrust (i) is supported by explicit symbolic computations through seven loops, though without released code; its D=4 mechanism is explicitly identified for L=2,3 as a vanishing Gram determinant, and the higher-loop data are consistent with that mechanism. Thrust (ii) is more delicate: the uniqueness up to an overall constant is established only by a parameter count in a finite ansatz, with no general proof of completeness. The paper is honest about the limits, explicitly warning in Sec. 4.3 that at seven loops and high multiplicity, cut-invisible terms appear and the homogeneous cut constraints are insufficient. That warning does not refute the low-loop claims, but it underscores that the low-loop completeness is an unproven empirical regularity. The most load-bearing concern is therefore not the scaling data themselves but the rank of the homogeneous system that underlies the uniqueness claim. A single independent nullspace computation on the three-loop ansatz would settle whether the 2757/2758 count is correct and whether the reconstructed integrand is truly unique. The reader's weakest assumption identified exactly this point, so I agree with the CONDITIONAL verdict; my proposed test would either elevate it to ACCEPT or force a downgrade to REJECT/UNVERDICTED depending on the outcome.","tokens_in":30131,"tokens_out":6651,"duration_ms":70126,"concrete_test":"Independently rebuild the three-loop four-point N=8 ansatz (83 diagrams, 2758 parameters) from the topology lists in Figs. 3 and 4, impose diagram symmetry and triangle power-counting, and evaluate the homogeneous constraints (multi-particle cut scaling F_i=0 for i>-5, iterated cut scaling, and BCFW-deformed cut scaling) at several independent rational kinematics. Compute the nullspace dimension and compare to 1. Additionally, verify the unique nonzero solution reproduces the known three-loop integrand of Bern et al. [88] by matching all maximal cuts. If the nullspace dimension exceeds 1, the uniqueness claim fails; if it is 1, the reconstruction is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that UV scaling conditions uniquely fix the N=8 integrand (Sec. 4.2) rests on the assertion that after imposing multi-particle cut scaling, iterated cut scaling, and BCFW-deformed cut scaling on a 2758-parameter triangle-power-counting ansatz, exactly one overall constant remains. This is presented only as a parameter count; no proof of completeness of the constraint set is given, and no code or ancillary data are provided to audit the count. The paper itself shows in Sec. 4.3 that the same type of homogeneous cut constraints is insufficient at seven loops (Eq. 4.26 introduces cut-invisible contact terms), so the low-loop uniqueness is an empirical, order-by-order fact rather than a structural theorem. Moreover, the constraints are checked by evaluating scaling conditions at 'generic' kinematic points; if the chosen points accidentally satisfy a Gram-determinant or spinor identity, the linear system can be degenerate and the nullspace can be larger than reported. Because the reconstruction claim is the main new result beyond the scaling observation itself, the absence of an independent rank/nullspace check is the weakest load-bearing point.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the ultraviolet behavior of loop integrands in N=8 supergravity and pure gravity by examining the large-momentum scaling of multi-particle unitarity cuts. Its central claim is that in D=4 the leading scaling is improved by one power over the general-D expectation for L=2 through L=7, with N=8 supergravity cuts scaling as 1/t^5. The paper attributes this improvement to the vanishing of certain Gram determinants, makes this mechanism explicit for L=2 and L=3 in Eq. (3.10), and then uses the improved scaling together with iterated-cut scaling and BCFW scaling of external momenta as homogeneous constraints to reconstruct the four-point N=8 SUGRA integrand through three loops and the n-point MHV integrand at one loop, up to an overall constant. The final part introduces multi-line chiral shifts of tree amplitudes, derives empirical scaling rules, and proposes new recursion relations and bonus relations for gravity tree amplitudes. The paper is careful to flag its conjectural parts, in particular the non-cut-constructibility discussion in Sec. 4.3.","tokens_in":1200,"tokens_out":1361,"duration_ms":48443,"significance":"If the reconstruction claim is correct, the paper establishes a genuinely new property of four-dimensional gravity integrands: the amplitude is fixed by homogeneous vanishing conditions rather than by functional matching on cuts. This is a suggestive step toward a geometric formulation of gravity amplitudes along the lines of the amplituhedron program, and the explicit Gram-determinant mechanism in Eq. (3.10) is an elegant, falsifiable explanation of the D=4 special behavior. The KLT-based explanation of the tree-level scaling in Sec. 5 and the empirical scaling rules for multi-line shifts are also valuable contributions. The low-loop integrands are benchmarked against known results, and the authors are honest about the order-by-order nature of the construction. The main weakness is that the load-bearing uniqueness statement rests on an unaudited 2757/2758 parameter count with no code, data, or independent rank verification accompanying the paper.","major_comments":[{"comment":"The central claim that homogeneous scaling constraints uniquely fix the three-loop four-point N=8 integrand rests entirely on the statement that 2757 of 2758 ansatz parameters are fixed, leaving one overall constant. The manuscript does not provide the linear system, the rank computation, the nullspace dimension, or the kinematic points at which the series expansions were evaluated. Because the constraints are homogeneous, any accidental degeneracy of the sampled constraint matrix would enlarge the nullspace and invalidate the uniqueness claim. The authors should provide a reproducible verification, for example as ancillary data with the ansatz, the constraint equations, and rational-kinematics rank/nullspace computations, and should state explicitly that the result is stable under varying the generic kinematic points.","section":"Sec. 4.2, Eqs. (4.18)-(4.22)"},{"comment":"The paper itself demonstrates that the same homogeneous cut constraints are insufficient at seven loops, where integrals with no propagators in one loop enter the ansatz and vanish on all unitarity cuts. This means the three-loop uniqueness is an empirical, order-by-order fact rather than a structural theorem about N=8 SUGRA integrands. Given that the abstract and conclusion use the low-loop construction to suggest a new geometric picture, the authors should either identify a mechanism that protects the low-loop cases from the seven-loop obstruction or clearly restrict the geometric interpretation to the checked orders.","section":"Sec. 4.3, Eq. (4.26)"},{"comment":"The explicit Gram-determinant explanation is written out only for L=2 and L=3, while Eq. (3.9) claims improved 1/t^5 scaling for 2 <= L <= 7. The higher-loop entries in Fig. 2 are stated to come from numerical evaluation at generic kinematics, but no code, data, or analytic expressions are provided, and the text concedes that the Gram-determinant form has not been written out beyond L=3. Please either supply the reproducible computations for all seven loops or mark the higher-loop scaling entries as conjectural, with a clear statement of which entries are verified.","section":"Sec. 3.2, Eqs. (3.9)-(3.10) and Fig. 2"},{"comment":"For the ladder diagram, the text shows that after imposing cut scalings the numerator is reduced to three terms that all scale as t^-6 on the multi-particle cut, and states that BCFW scaling of external momenta fixes them. However, no explicit equations or coefficient-level computation are shown for this step, and it is not demonstrated that the BCFW condition is independent of the cut-scaling conditions on this topology. This is a second instance where the parameter count is the only evidence offered for completeness of the constraint set.","section":"Sec. 4.2, Eq. (4.21)"}],"minor_comments":[{"comment":"The text contains many typographical errors and LaTeX artifacts (for example, missing spaces such as 'four-dimensionaldeformations' in Fig. 2, 't−5' vs. '1/t^5' inconsistencies, and broken equation references in section titles). A careful proofread and consistent use of equation labels would improve readability.","section":"Throughout"},{"comment":"The figure is difficult to interpret without a legend that clearly separates the four theories and the D-dimensional versus D=4 cases. The text should state explicitly that the dashed lines are conjectural for the higher-loop D-dimensional scaling, not only for the continuous part.","section":"Sec. 3.2, Fig. 2"},{"comment":"The table would benefit from a caption stating the precise shift definition and the relation between the L-loop label and the number of external legs. Currently the caption is missing, and the text refers to the table only as 'Tab. 2' without explaining how to read the rows and columns.","section":"Sec. 5.1, Table 2"},{"comment":"The authors should define what is meant by 'triangle power-counting' when applied to reducible numerators, and clarify the statement that the ansatz includes terms beyond triangle power-counting at three loops; the current footnote 9 is ambiguous.","section":"Sec. 4.2, around Eq. (4.13)"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the headline: this paper finds a genuinely new property of gravity loop integrands — a D=4-specific improvement in the UV scaling of multi-particle cuts (1/t^5 vs 1/t^4 in general D) that persists through seven loops and is traced to a Gram determinant vanishing. It then shows that homogeneous vanishing conditions — the improved scaling plus BCFW scaling of external momenta — uniquely fix the two- and three-loop four-point N=8 integrands and the one-loop n-point MHV integrand, up to an overall constant. This is real content, not numerology.\n\nWhat the paper does well: the scaling computations are explicit, the Gram-determinant mechanism is an independent explanation rather than just curve-fitting, and the authors are unusually honest about what is conjectural. They flag the tree-level scaling law (Eq. 5.8) as empirical up to eight points, and they themselves show in Sec. 4.3 that the same homogeneous-cut logic fails at seven loops due to cut-invisible contact terms. The reconstruction is non-circular: the constraints come from tree-level building blocks, not from the ansatz coefficients.\n\nThe soft spots are mostly about verifiability. The central uniqueness claim for the three-loop integrand rests on a parameter count — 2757 of 2758 parameters fixed — and no code or data files are provided to audit that count or the scaling results through seven loops. The stress-test concern about degeneracy at generic kinematic points is real: if the chosen evaluation points accidentally satisfy a Gram identity, the nullspace could be larger than reported. That said, this is a standard worry in this area, and the authors' explicit mention of generic values suggests they checked more than one point. Still, an independent rank check would settle it.\n\nThe seven-loop failure in Sec. 4.3 is not a flaw; it is a stated limitation. It undercuts any claim that the homogeneous construction is a general theorem, but the paper does not make that claim. It says the constraints work through three loops and conjecturally more generally, with exceptions starting at seven loops. That is a fair statement.\n\nMinor quibbles: the m-line shift scaling (5.8) is presented as 'experimental evidence' — fine as data, but a derivation or sharper conjecture would be welcome. The KLT-based explanation in Sec. 5 is suggestive and consistent with the data.\n\nWho should read this: anyone working on the UV structure of gravity, integrand reconstruction, or recursion relations. It extends the program from [34] and [30] meaningfully. It deserves a serious referee. My recommendation: send it to review, and ask the authors to provide the ancillary files or at least a detailed description of the linear system, the evaluation points, and the nullspace.","headline":"A genuinely new D=4 UV scaling property for gravity cuts, with a Gram-determinant mechanism and a non-circular integrand reconstruction that deserves peer review despite the unaudited parameter-count step.","tokens_in":30877,"tokens_out":2692,"would_cite":true,"duration_ms":27790,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Four-dimensional gravity cuts are one power softer at infinity, and that softness fixes the integrand.","keywords":["N=8 supergravity","pure gravity","loop integrands","ultraviolet behavior","unitarity cuts","Gram determinant","BCFW scaling","integrand reconstruction"],"falsifier":"A concrete check: compute the complete three-loop four-point N=8 supergravity integrand in $D=4$ by an independent method, for example direct double-copy or unitarity matching on a spanning set of cuts, and compare with the reconstruction; any nonzero difference among the 2758 ansatz coefficients would falsify the claim that the homogeneous constraints are complete. A second, cheaper check: evaluate the $t^{-4}$ coefficient of the four-loop four-point multi-particle cut in $D=4$; if it does not vanish, the improved scaling of Eq. (3.9) breaks down at $L=4$, and if it vanishes but is not proportional to $(\\mathrm{Gram}[q_1 q_2 p_1 p_2 p_3])^2$, the Gram-determinant mechanism is not the full explanation.","tokens_in":29877,"feed_emoji":"","tokens_out":9271,"duration_ms":85751,"temperature":0.7,"pith_summary":"This paper tries to establish that gravity loop integrands are surprisingly tame in exactly four spacetime dimensions: on multi-particle unitarity cuts, as the cut loop momenta are sent to infinity, the maximally supersymmetric N=8 supergravity integrand falls off as $1/t^5$ rather than $1/t^4$, and pure gravity as $t^3$ rather than $t^4$, for two through seven loops. The mechanism is located in the vanishing of a Gram determinant that is nonzero in general dimension but identically zero in $D=4$. The paper then shows that this improved scaling, combined with the known $1/t^2$ behavior of the cut under BCFW deformations of external momenta, uniquely fixes the four-point loop integrand of N=8 supergravity through three loops and the $n$-point MHV integrand at one loop, up to an overall constant. Because every constraint is homogeneous, demanding that an ansatz vanish at particular points rather than match nonzero cut data, the construction offers a possible template for a geometric, amplituhedron-like picture of gravity amplitudes.","feed_headline":"Four-dimensional gravity cuts soften by one power at infinity","feed_subtitle":"The vanishing at infinity alone pins down the N=8 supergravity loop integrand through three loops.","key_machinery":"The engine of the argument is the multi-particle unitarity cut $F(\\ell_k,p_j)$, the residue of the $L$-loop integrand on the $L+1$ on-shell conditions $\\ell_1^2=\\cdots=\\ell_{L+1}^2=0$ with $\\sum \\ell_k = -(p_1+p_2)$, evaluated in the limit where the on-shell loop momenta are sent to infinity along a chiral shift $\\tilde\\lambda_{\\ell_k}\\to\\tilde\\lambda_{\\ell_k}+t z_k \\tilde\\eta$ with $\\sum_k z_k \\lambda_{\\ell_k}=0$. In general dimension the cut falls off like the worst-behaved contributing integral, but in $D=4$ the leading term is proportional to the square of a Gram determinant, $(\\mathrm{Gram}[q_1 q_2 p_1 p_2 p_3])^2$, which vanishes identically and leaves the improved scaling. This improved scaling is then used as a homogeneous constraint on an integrand ansatz built in a triangle-power-counting basis, together with the requirement that the cut scale as $1/t^2$ under BCFW shifts of external momenta and, at one loop, vanishing on forbidden cuts; the vanishing conditions fix the numerator degrees of freedom without any functional matching of the amplitude on cuts.","core_discovery":"The central claim is that four-dimensional cuts of gravity loop integrands improve by one power of $t$ at large loop momenta compared with their general-dimensional counterparts. Writing the multi-particle unitarity cut as a product of tree amplitudes and deforming the on-shell loop momenta chirally, $\\tilde\\lambda_{\\ell_k} \\to \\tilde\\lambda_{\\ell_k} + t z_k \\tilde\\eta$ with $\\sum_k z_k \\lambda_{\\ell_k}=0$, the paper finds $F_{\\rm SUGRA}\\sim 1/t^5$ and $F_{\\rm GR}\\sim t^3$ for $2\\le L\\le 7$; in general $D$ the corresponding scalings are $1/t^4$ and $t^4$. The extra power of softness is traced to the leading $1/t^4$ term being proportional to $(\\mathrm{Gram}[q_1 q_2 p_1 p_2 p_3])^2$, which vanishes identically in $D=4$. The paper further claims that these improved scaling conditions are homogeneous constraints that, together with the $1/t^2$ BCFW scaling of the cut under external shifts, uniquely fix the N=8 supergravity integrand in nontrivial cases: the two- and three-loop four-point amplitudes, with 2757 of 2758 ansatz parameters set by the constraints, and the one-loop $n$-point MHV amplitude, up to one overall constant.","pith_inferences":["If the Gram-determinant mechanism is the whole story, the improved scaling should persist at all loop orders in $D=4$, since it relies on a kinematic identity rather than on the specific integrand data checked through seven loops; this could be tested by computing the $L=4$ cut coefficient of $t^{-4}$ explicitly.","The observation that the KLT representation of gravity trees inherits good large-$t$ behavior from Yang-Mills while individual BCJ terms do not suggests that the improved UV scaling is not a property of any local diagrammatic expansion; a fully nonlocal or geometry-based formulation may be the natural language for these cancellations.","A concrete extension would be to apply the same homogeneous-constraint program to non-MHV one-loop N=8 amplitudes: the paper notes that the improved $1/t^3$ two-particle cut does not add independent information for MHV but might be necessary for higher MHV degree, so applying it there is a direct test of the method's scope.","If the homogeneous reconstruction extends beyond three loops, it would provide evidence for a gravity analog of the amplituhedron picture in which the integrand is defined by its zeros rather than by factorization data; the seven-loop obstruction marks where that geometric story would have to be modified."],"forward_implications":["In four dimensions the multi-particle unitarity cuts of N=8 supergravity scale as $1/t^5$ and of pure gravity as $t^3$ for $2\\le L\\le 7$, one power better than their general-dimensional counterparts.","The leading-order cancellation is kinematic: the $1/t^4$ term in N=8 supergravity is $(\\mathrm{Gram}[q_1 q_2 p_1 p_2 p_3])^2$, which vanishes only because four-dimensional Lorentzian kinematics makes that Gram determinant zero.","The same vanishing conditions that improve the scaling also pin the integrand: they fix 2757 of 2758 parameters in the three-loop four-point N=8 supergravity ansatz, and together with BCFW scaling leave only the overall constant, giving a unique integrand without matching cuts to nonzero values.","At one loop the procedure extends to all multiplicities for MHV N=8 supergravity: the homogeneous constraints (improved scaling, BCFW behavior, and vanishing on forbidden cuts) fix the unique integrand, consistent with the known chiral-box representation.","The method is not universal: at seven loops, and at two loops for sufficiently many external legs, diagrams with no propagators in one loop can appear, so cuts alone no longer determine the integrand and new constraints are needed."],"supporting_citations":[{"why":"the prior study of UV cancellations in gravity loop integrands that this paper generalizes and explains","marker":"[34]"},{"why":"the general-dimensional integrand-versus-integral analysis that found no cancellations, contrasted with the D=4 improvement","marker":"[30]"},{"why":"supplies the known two-loop four-point N=8 supergravity integrand that the homogeneous reconstruction reproduces","marker":"[87]"},{"why":"supplies the known three-loop four-point N=8 supergravity integrand used as the reconstruction target","marker":"[88]"},{"why":"provides four-loop N=8 supergravity integrand data used in the scaling analysis","marker":"[89]"},{"why":"KLT relations used to build gravity tree amplitudes from Yang-Mills trees when computing the unitarity cuts","marker":"[91]"},{"why":"provides the multileg one-loop gravity amplitude that the one-loop MHV reconstruction reproduces","marker":"[93]"},{"why":"defines the triangle-power-counting integrand basis and the completeness assumptions underlying the ansatz","marker":"[58]"},{"why":"establishes poles at infinity and non-cut-constructibility for high-multiplicity amplitudes, limiting the method's scope","marker":"[108]"},{"why":"shows gravity integrands vanish in collinear regions, supporting the forbidden-cut constraints used at one loop","marker":"[68]"}],"fun_headline_variants":["Gravity loop integrands get one extra UV power in 4D","Improved UV scaling fixes N=8 supergravity integrands","Four-dimensional gravity cuts softer by one power at infinity","UV softness uniquely determines N=8 supergravity loop integrands","Gravity's UV scaling gains an extra soft power at infinity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the small set of vanishing-at-infinity conditions (multi-particle cuts, iterated cuts, and BCFW-deformed cuts) is complete enough to separate every numerator degree of freedom in the triangle-power-counting ansatz; the paper verifies this case-by-case by counting parameters, with 2757 of 2758 fixed, but offers no general proof, and Section 4.3 shows the same completeness fails at seven loops.","fun_headline_variants_meta":{"raw":{"variants":["Gravity loop integrands get one extra UV power in 4D","Improved UV scaling fixes N=8 supergravity integrands","Four-dimensional gravity cuts softer by one power at infinity","UV softness uniquely determines N=8 supergravity loop integrands","Gravity's UV scaling gains an extra soft power at infinity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000451,"raw_usage":{"total_tokens":2307,"prompt_tokens":1016,"completion_tokens":1291,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":632,"completion_tokens_details":{"reasoning_tokens":1205}},"tokens_in":632,"tokens_out":1291,"duration_ms":9761,"temperature":1.0,"reasoning_tokens":1205,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T05:03:18.104598+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete check: compute the complete three-loop four-point N=8 supergravity integrand in $D=4$ by an independent method, for example direct double-copy or unitarity matching on a spanning set of cuts, and compare with the reconstruction; any nonzero difference among the 2758 ansatz coefficients would falsify the claim that the homogeneous constraints are complete. A second, cheaper check: evaluate the $t^{-4}$ coefficient of the four-loop four-point multi-particle cut in $D=4$; if it does not vanish, the improved scaling of Eq. (3.9) breaks down at $L=4$, and if it vanishes but is not proportional to $(\\mathrm{Gram}[q_1 q_2 p_1 p_2 p_3])^2$, the Gram-determinant mechanism is not the full explanation.","supporting_citations":[{"cited_title":"Amplitudes at Infinity","cited_arxiv_id":"1812.11185","evidence_quote":"establishes poles at infinity and non-cut-constructibility for high-multiplicity amplitudes, limiting the method's scope"}],"review_version":1}