{"id":"b7a4e404-253d-4295-a04c-d19e5883601a","arxiv_id":"2511.01971","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"RG flow in scalar-fermion theories is gradient through four loops only when the 'beta shift' is included, via over a thousand scheme-independent constraints that hold wherever current data allow a check.","lead":"Physicists have long suspected that renormalisation-group flows can be written as downhill motion on a landscape, a property tied to irreversibility theorems. This paper tests that property in scalar-fermion quantum field theories to four-loop order and shows it works only after adding a subtle 'beta shift' term, while also finding that fixed points with such shifts dominate as particle content grows.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Four-loop gradiency of B is unverified: the beta shift S_ij/P_ab is never computed, so constraints (B.3)–(B.66) and (3.38) remain unchecked.","rationale":"The reader's weakest_assumption identifies exactly the most load-bearing gap: the four-loop beta shift is never computed, so the 64+2 constraints that are necessary for B to be gradient at 3/4 loops cannot be checked. I agree that this makes the central gradient claim conditional rather than established. The paper is transparent about the gap and provides substantial independent evidence—scheme-invariant constraints verified wherever data exist, the three-loop shift satisfying the 2/3-loop constraints, and the g0=24 factorization matching eF—but none of these supplies the missing four-loop beta-shift coefficients. The numerical-census concern is real but secondary: even if the beta-shift constraints were verified, the claim that beta-shift fixed points dominate would still require checking B=0 to higher order rather than only one-loop beta roots. Since the reader already assigned CONDITIONAL with moderate confidence, my stress-test does not move the verdict; it reinforces it.","tokens_in":26772,"tokens_out":5923,"duration_ms":59711,"concrete_test":"Perform an explicit four-loop calculation of the beta shift S_ij and P_ab for the general scalar-fermion theory, extending the three-loop method of [10,14,36], and substitute the resulting coefficients into each constraint in Appendix B and into (3.38). If any constraint fails, the four-loop gradiency claim is disproved; if all are satisfied, the central conditional claim becomes a verified result.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central four-loop claim—that B = β − (Sg) is Riemannian-gradient at 3/4 loops—depends on the four-loop beta shift satisfying 64 constraints in Appendix B plus two ε-dependent constraints (3.38). The paper does not compute this shift. In §3.1 and §3.2 the authors state that 'a true check would require the calculation of these coefficients at four loops,' and the list of individually fixed coefficients in (3.32) is explicitly contingent on that unperformed calculation. If any one of constraints (B.3)–(B.66) or (3.38) fails, the B-function is not gradient at four loops, so the paper's strongest gradient claim would be false. The three-loop shift satisfying (3.30)–(3.31) and the many lower-order constraints that are satisfied provide supporting evidence, but they do not remove the load-bearing nature of the unchecked four-loop coefficients. The numerical census in §4 and Table 3 is a separate limitation: it counts roots of the one-loop beta functions with non-zero S without verifying B=0 to O(ε^3), so the 'CFT space dominated by beta-shift fixed points' claim is also not fully established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the gradient property of the renormalisation-group vector field in general scalar-fermion theories, with the beta shift B = β − (S g) playing a central role. In d = 4 and d = 4 − ε, the authors enumerate all relevant diagrams through three/four loops, solve the gradient flow equation ∂A = G B order by order, and derive a large set of algebraic constraints on the beta-function coefficients and on the beta-shift coefficients. The lower-order constraints are checked against known MS results and against the known three-loop beta shift; the four-loop beta shift is not computed. The paper also matches the constructed A-function to the sphere free energy eF for the Gross–Neveu–Yukawa and Nambu–Jona-Lasinio–Yukawa CFTs, fixing one coefficient a3_4 = 8. Finally, a numerical search over one-loop fixed points for small numbers of scalars and fermions shows that a substantial and increasing fraction have non-zero beta shift.","tokens_in":27083,"tokens_out":4349,"duration_ms":45496,"significance":"If the four-loop beta-shift constraints are eventually verified, this would be a significant extension of the gradient-flow programme to scalar-fermion theories, giving further evidence that the proper RG vector field is B and that the beta shift is essential for Riemannian gradient flow, and providing an explicit construction of A up to three/four loops. The paper is transparent about the conditional nature of the four-loop check and about the distinction between one-loop roots and B=0 fixed points. Strengths of the presentation include the explicit enumeration of diagrams, the scheme-invariance analysis, the detailed list of constraints in Appendix B, the matching of A to eF in two models, and the ancillary Mathematica file that allows independent inspection. The numerical observation that beta-shift fixed points become increasingly common is intriguing, but it is not yet a statement about solutions of B=0 to the order needed.","major_comments":[{"comment":"The central claim that the B-function is gradient through three/four loops is conditional on the four-loop beta shift satisfying the 64 constraints (B.3)–(B.66) and the two ε-dependent constraints (3.38). The manuscript states at §3.1 and §3.2 that 'a true check would require the calculation of these coefficients at four loops', and the fixed coefficient list in (3.32) is explicitly contingent on that calculation. Since S4 and P4 are never computed, the result is a set of necessary and sufficient constraints on an uncalculated object, not a verification of gradiency. The abstract and introduction should be reworded so that the four-loop conclusion is presented as a consistency condition with strong lower-loop support, or the four-loop shift must be computed and checked.","section":"Sec. 3.1/3.2, Appendix B"},{"comment":"The numerical census counts roots of the one-loop beta functions and labels any such root with non-zero S as a beta-shift fixed point. But a CFT fixed point requires B=0 to the appropriate order in ε; at order ε^3, where S first contributes, a one-loop root need not extend to a solution of B=0. The paper verifies B=0 explicitly only for the two analytic examples in Section 2, not for the hundreds of numerically found candidates. Therefore the conclusion that 'the space of CFTs is dominated by those with non-zero beta shift' is not established by this census. The candidates should be re-checked against the B=0 equations, or the claim should be restricted to one-loop beta-function roots with non-zero shift.","section":"Sec. 4, Table 3"},{"comment":"The treatment of odd-trace diagrams is an additional assumption: the text says they are included but their coefficients are 'set to zero as necessary'. If such coefficients are not independently computed, then the statement that the analysis covers general scalar-fermion beta functions is not fully supported. The assumption should be stated as a restriction on the class of theories (e.g., those with a four-dimensional Weyl uplift), or the odd-trace coefficients should be computed and included in the constraints.","section":"Sec. 3, diagram enumeration"}],"minor_comments":[{"comment":"The match of A to eF in the Gross–Neveu–Yukawa and NJL–Yukawa models is a significant check, but it uses one fitted coefficient a3_4 = 8. The paper correctly notes that this is a single coefficient independent of N; the text could state more explicitly that this is a consistency check of the eF-extension conjecture, not a proof.","section":"Sec. 3.4"},{"comment":"In the two analytic examples, the text says that the beta-function equations 'cannot be continued beyond two loops' and that B removes the obstruction. It would be helpful to display the non-zero constant term in the beta function of y2 at order ε^3, so the reader can see the cancellation explicitly without reconstructing it from the formulas.","section":"Sec. 2.1"},{"comment":"The scheme-invariance argument is clear, but the notation in equations (3.39)–(3.41), especially the sum over sub-tensor structures, is compressed. A one-sentence explanation of how T^n_m split under the action on β would improve readability.","section":"Sec. 3.3"},{"comment":"The list of constraints (B.3)–(B.66) would be more usable if the diagrammatic definitions of the individual S4_k and P4_k coefficients were available in the same document, rather than only via the ancillary file. At minimum, a table matching the coefficient labels to the drawing numbers would help.","section":"Appendix B"},{"comment":"There are minor typographical issues, e.g., 'N_S' vs 'Ns' in Section 2 and the label '0/1loop' in Table 1. These do not affect the results.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"This is a competent and useful paper with a large amount of explicit algebraic material. The main issue is that the headline four-loop gradient claim is conditional on an uncomputed object, and the numerical census does not verify the fixed-point condition that the paper itself defines. Both are fixable within the manuscript's scope by rephrasing the claims or by carrying out the missing checks. I see no concerns about citation practice or overlap; the paper builds on the authors' earlier work and on [10], [36] in a transparent way."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear X,\n\nThe one thing to know: this is a careful, honest paper whose central four-loop claim is, by the authors' own admission, unverified. The B-function being Riemannian-gradient at 3/4 loops in scalar-fermion theories is equivalent to 955+61 constraints, and all that can be checked with existing data are satisfied — but the four-loop beta shift S_ij/P_ab has not been computed, so the 64 constraints of Appendix B and the two ε-dependent constraints (3.38) remain predictions. If any one fails, the four-loop gradiency result collapses. The paper says this plainly; it is not hidden.\n\nWhat is new and good. The extension of the gradient program from multiscalar to scalar-fermion systems requires dealing with two coupling types, loop-order mixing, and the beta shift. The authors enumerate the vacuum graphs, derive a large set of scheme-invariant constraints, show that the ordinary β fails many of them while B passes at the orders where the shift is known, and construct an explicit A and metric. The example fixed points in §2, where β=0 cannot be continued beyond two loops but B=0 can, are concrete and convincing. They also match a single unfixed coefficient in A (a3_4=8) to the eF free energy in two independent GN(Y) and NJL(Y) models, which is a real check, not numerology. The ancillary Mathematica file is a plus.\n\nSoft spots. Two, both load-bearing. First is the four-loop shift. The authors call the constraints satisfied 'whenever the full set of results needed to check them is available' — that set is not available at four loops. The weight-of-evidence argument ('immense conspiracy') is reasonable but is not a computation. Second, the census in §4 counts roots of the one-loop beta functions with S≠0 as CFT fixed points without verifying B=0 to O(ε^3). The percentages are therefore about one-loop candidates, not established fixed points. The authors are not misrepresenting what they did, but the 'dominated by non-zero beta shift' claim is weaker than it sounds. Minor: the odd-trace coefficient zeroing in §3 is under-specified.\n\nThis is a serious paper by people who know the subject; it deserves a careful referee. The referee's main job is to push on the four-loop constraints and the census, and to make sure the authors present the conditional status as prominently as they do in the text. If the shift is eventually computed and the constraints hold, this will be a landmark. As it stands, it is a strong but incomplete result. If you work on RG gradient flows, ε-expansion CFTs, or the eF-theorem, read it.\n\nRecommendation: send to peer review.","headline":"Careful and honest paper, but its four-loop gradiency claim is conditional on an uncomputed beta shift; the lower-order results and the eF matching are genuinely solid.","tokens_in":27512,"tokens_out":2326,"would_cite":true,"duration_ms":20975,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81T17","81T40"],"pacs":["11.10.Hi"],"model":"deepseek-v4-flash","headline":"The renormalisation-group flow of scalar-fermion theories is a gradient flow through four loops — once the beta shift is subtracted from the beta function.","keywords":["renormalisation group","gradient flow","beta shift","scalar-fermion theories","epsilon expansion","conformal field theory fixed points","four-loop beta functions","scheme independence"],"falsifier":"Compute the four-loop beta-shift coefficients S_ij and P_ab in a general scalar-fermion theory and check the 64 equalities in Appendix B and the two in equation (3.38). A single violation would show B = β − Sg is not a Riemannian gradient at 3/4 loops; full agreement would confirm the paper's central claim. The list includes exact predictions such as S4_2 = 7/24 + π^4/120.","tokens_in":26647,"feed_emoji":"🌀","tokens_out":6515,"duration_ms":61732,"temperature":0.7,"pith_summary":"This paper asks whether the renormalisation-group flow of theories with scalars and fermions can be written as the gradient of a single function on coupling space, which would guarantee a quantity that decreases along every flow. It shows that the answer is yes through four loops in d=4 and d=4−ϵ, provided one uses the beta shift: the flow vector is B = β − Sg, not the standard beta function β. Satisfying the gradient equations turns out to be equivalent to 1016 scheme-independent constraints on the beta-function coefficients, and every constraint that can be checked with known results is satisfied. The paper also finds that among fixed points of these theories, those with nonzero beta shift — genuine CFTs with nonvanishing beta functions, whose couplings rotate along the flow — are common and become the majority as the number of fields grows. If right, this means the gradient structure of the RG survives the addition of fermions, and that a large class of conformal field theories would be missed by searches that set β=0.","feed_headline":"Subtracting the beta shift keeps RG flow gradient through four loops","feed_subtitle":"A shifted beta function passes 955 scheme-independent constraints; fixed points with nonzero shift dominate as fields grow.","key_machinery":"The central object is the beta shift S, a contribution to the trace of the energy-momentum tensor from operator mixing with non-conserved spin-one currents. It corrects the beta function to B = β − (Sg). The gradient flow equation ∂A/∂g = G B, with A the would-be monotone and G a Riemannian metric on coupling space, is expanded order by order using tensor structures built from quartic and Yukawa couplings. All vacuum graphs through O(λ^5, …, y^10) are enumerated, and the resulting linear system is reduced to the constraint equations. A primitive-diagram family of constraints fixes the fermion-metric normalization g0 = 24, which also makes A match the sphere free-energy quantity at low orders","core_discovery":"The paper claims that for the most general scalar-fermion theory in d=4 and d=4−ϵ, the RG flow generated by B = β − (Sg) is a Riemannian gradient through 3/4 loop order. This holds if and only if a set of 955 (d=4) plus 61 (d=4−ϵ) scheme-independent constraints on the beta-function and beta-shift coefficients are satisfied; all constraints that can be verified with existing four-loop data are satisfied. The standard beta function β alone fails once the shift can contribute, starting at 2/3 loop order. At fixed points of B with β ≠ 0, the couplings trace limit cycles under flavour rotations, and these are genuine CFTs. The paper exhibits two analytic examples and shows numerically that such f","pith_inferences":["Inference: If the four-loop beta shift is ever computed, the 64 constraints in Appendix B plus the two ε-dependent constraints provide an unambiguous binary test — one violated equality would show the gradient picture breaks at this order, while full agreement would extend it to the deepest order claimed.","Inference: Systematic searches for infrared fixed points in the ε expansion that solve β=0 will miss most of the conformal theories once N_s ≥ 3 and N_f ≥ 2; such searches should instead solve B=0 at order ε^3.","Inference: The constraint-counting method should transfer to general gauge-Yukawa theories, where a beta shift also exists; verifying the analogous constraints there would broaden the gradient property beyond the scalar-fermion class.","Inference: The matching of A to the sphere free energy with a single coefficient suggests eF may admit a gradient-flow extension; a concrete test would be computing both quantities at a third, less symmetric fixed point."],"forward_implications":["Fixed points of the RG are zeros of B = β − Sg, not of β; at such points the couplings undergo flavour rotations, and the physically relevant anomalous-dimension matrix is γ + S.","Gradient flow at 3/4 loops is equivalent to 1016 scheme-independent constraints; the constraints on β alone are violated in the minimal-subtraction scheme, and the beta shift repairs them.","In the ε expansion, fixed points with nonzero beta shift already exist for two scalars and one Weyl fermion, and their proportion grows sharply with field content (to roughly two-thirds for N_s=4, N_f=2).","The A-function constructed as the gradient potential can be matched at low order to the sphere free energy with a single field-independent coefficient choice, extending the eF-conjecture toward a full gradient-flow statement."],"fun_headline_variants":["Shifted beta keeps RG flow gradient through four loops","955 constraints pass: shifted beta makes RG flow gradient","Gradient RG flow survives four loops with beta shift","Beta shift enforces gradient flow to four loops","Beta shift preserves gradient flow; nonzero-shift CFTs dominate"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The whole gradient-flow conclusion rests on the as-yet-uncomputed four-loop coefficients of the beta shift satisfying the 64 constraints listed in Appendix B and the two ε-dependent constraints; if even one fails, the B function is not gradient at 3/4 loops.","fun_headline_variants_meta":{"raw":{"variants":["Shifted beta keeps RG flow gradient through four loops","955 constraints pass: shifted beta makes RG flow gradient","Gradient RG flow survives four loops with beta shift","Beta shift enforces gradient flow to four loops","Beta shift preserves gradient flow; nonzero-shift CFTs dominate"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000725,"raw_usage":{"total_tokens":3068,"prompt_tokens":706,"completion_tokens":2362,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":450,"completion_tokens_details":{"reasoning_tokens":2284}},"tokens_in":450,"tokens_out":2362,"duration_ms":15955,"temperature":1.0,"reasoning_tokens":2284,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T06:51:21.182741+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the four-loop beta-shift coefficients S_ij and P_ab in a general scalar-fermion theory and check the 64 equalities in Appendix B and the two in equation (3.38). A single violation would show B = β − Sg is not a Riemannian gradient at 3/4 loops; full agreement would confirm the paper's central claim. The list includes exact predictions such as S4_2 = 7/24 + π^4/120.","supporting_citations":[],"review_version":1}