{"id":"6df55000-99c9-472b-8e51-19fd2122f31e","arxiv_id":"1908.09887","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"TVID 2 computes all master integrals for planar three-loop self-energy diagrams with two closed fermion loops using at most two-dimensional numerical integrations.","lead":"This paper presents TVID 2, a program that evaluates a class of three-loop particle physics integrals with arbitrary masses much faster than existing general-purpose tools. Physicists computing high-precision predictions for future colliders could use it for diagrams that currently take hours or days.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The order-of-magnitude U8a imaginary-part disagreement with FIESTA at p^2=40 remains unresolved, and the paper supplies no third check to show TVID 2 is the correct side; this is the load-bearing gap in the 6–10 digit claim.","rationale":"The reader’s weakest assumption is exactly the point I would stress. The paper is an honest methods paper with publicly available code and broad benchmark agreement for roughly 25 integrals, which is genuine independent support. The U8a case is uniquely unresolved: real parts differ modestly, imaginary parts differ by an order of magnitude, and no third method breaks the tie. I considered whether the omitted O(epsilon) two-loop terms are more load-bearing; they are disclosed and argued to cancel in physical observables, and the benchmarks deliberately use modified master integrals that avoid them, so they do not undermine the benchmark claim as directly. I also considered the numerical differentiation used for U6m1, U6m3, U6n2, U7a1, and U7a2; it is disclosed with reduced precision estimates and the reported values are consistent, so it is a secondary limitation, not the principal risk. The U8a discrepancy is the single point where, if the paper is wrong, the central numerical claim fails, and the paper supplies no decisive evidence to choose between TVID and FIESTA. Since the reader already issued a CONDITIONAL verdict, my assessment does not change that verdict. A Cutkosky evaluation of Im U8a at the benchmark point would settle the issue: if it confirms TVID, the concern is resolved in favor of acceptance; if it agrees with FIESTA, the flagship U8a result is incorrect and the claim of 6–10 digit precision is not supported.","tokens_in":23892,"tokens_out":9729,"duration_ms":97507,"concrete_test":"Independently compute Im U8a at the benchmark point p^2=40, m_i^2=(1.1,1.2,1.3,1.4,1.5,1.6,1.7,1.8) using the Cutkosky rule: sum all two- and three-particle cuts that separate the U8a graph into two connected parts, replace each cut propagator with the corresponding on-shell delta function, fix the overall sign from the largest-time equation, and integrate the resulting phase-space integrals numerically. Compare the result with TVID’s -0.16344185(2). Agreement with TVID would resolve the concern in TVID’s favor; agreement with FIESTA’s -0.016361(3) would show that TVID’s U8a imaginary part is wrong. A useful sub-check is to evaluate separately the double-cut contribution through the two inserted propagators 4 and 5, since the missing double-cut term would most naturally account for the large difference.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that TVID 2 evaluates the planar three-loop self-energy masters to 6–10 digits. The benchmark support is agreement with FIESTA 4.1 in Tabs. 2–3. For the flagship topology U8a at p^2=40, the real parts roughly agree (0.01238717(2) vs 0.012353(3)), but the imaginary parts disagree by an order of magnitude: TVID gives -0.16344185(2), FIESTA gives -0.016361(3). The difference is about 5×10^4 times FIESTA’s quoted error. The paper’s only explanation is that FIESTA’s contour-deformed sector decomposition may be insufficiently smooth for 7+ propagators, but no independent method, analytic limit, or cross-check establishes that TVID is correct. This is especially important because U8a is the master topology from which all planar topologies descend and is the unique benchmark with two nontrivial propagator poles, x=m_4^2 and y=m_5^2 in Eq. (9). An error in the principal-value/residue treatment in Eq. (10), or in the handling of C0 branch cuts in the x,y variables, would affect exactly this topology. The broad agreement of the other integrals is not sufficient to certify U8a, because U8a is the only case exercising the double-pole x,y integration. Thus the load-bearing assumption—that TVID’s U8a value is right and FIESTA’s is wrong—is unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents TVID 2.0, a C/Mathematica package for the numerical evaluation of planar three-loop self-energy master integrals with arbitrary masses, focusing on the topology class needed for three-loop self-energy diagrams with two closed fermion loops. The method separates each master integral into a known UV-divergent subtraction part and a finite remainder, which is written as a one- or two-dimensional integral over known one-loop B0/C0 functions and simpler two-loop or vacuum integrals. Two techniques are used: sub-loop dispersion relations for topologies with sub-bubbles, and a modified Ghinculov-type representation for topologies with sub-loop triangles. Benchmark comparisons against FIESTA 4.1 are presented for p^2 below threshold (Table 2) and for p^2=40 with physical thresholds and imaginary parts (Table 3). The paper claims general 6-10 digit precision with run times from under a second to about 20 minutes, while also listing explicit limitations: no IR-divergent inputs, omitted O(epsilon) terms of some two-loop functions, reduced precision from numerical mass differentiation, and double-precision evaluation of C0 functions.","tokens_in":24209,"tokens_out":6686,"duration_ms":70434,"significance":"If the central claims are correct, TVID 2.0 is a genuinely useful tool for precision calculations in the Standard Model and BSM theories, since it covers a large class of three-loop self-energy master integrals that are difficult to obtain analytically. The paper is commendably concrete: the derivations for representative topologies in Section 3 and the subtraction formulas in Appendix A are explicit, the source code is made available, and the benchmark tables permit independent scrutiny. The use of FIESTA 4.1 as an independent cross-check is appropriate, since no parameters of TVID are fitted to FIESTA. However, the benchmark evidence contains at least one order-of-magnitude disagreement for the flagship topology U8a, and one further sign discrepancy in Table 2, so the paper's central accuracy claim is not fully established. The stated limitations also narrow the practical scope of the 'arbitrary masses' claim. These issues are fixable and do not, in my view, invalidate the approach.","major_comments":[{"comment":"At p^2=40, the imaginary part of U8a is -0.16344185(2) in TVID 2.0 and -0.016361(3) in FIESTA 4.1, an order-of-magnitude discrepancy that is roughly 5e4 times FIESTA's stated error. The real parts also differ by more than the combined quoted errors (0.01238717(2) versus 0.012353(3)). The paragraph after Table 3 attributes the disagreement to insufficient smoothness of FIESTA's contour-deformed sector decomposition, but no third independent method, analytic limit, threshold behavior check, or alternative numerical integration is supplied to show that TVID's value is the correct one. This matters because U8a is the only benchmark that exercises the double propagator poles at x=m4^2 and y=m5^2 and the branch-cut structure of the C0 functions in Eq. (9); the residue/principal-value prescription of Eq. (10) is therefore not independently validated. The adjacent U7a row also shows differences larger than the quoted errors. I request a third cross-check for U8a, or a correspondingly qualified statement of the accuracy claim for this topology.","section":"Sec. 4, Table 2 (U4a1 row)"},{"comment":"The table lists TVID 2.0 result -1.4651121210(1) for U4a1, while FIESTA 4.1 gives 1.465(3), i.e. opposite sign and no overlap. The text states that agreement is 'excellent' for almost all master integrals and does not comment on this row. If the sign difference is due to a convention in defining the mass derivative, or to a known limitation of FIESTA for this doubled-propagator case, this should be stated explicitly; otherwise this is a second unexplained failure of the benchmark and must be resolved before the general accuracy claim can be accepted.","section":"Sec. 4, Table 2 (U4a1 row)"},{"comment":"The advertised scope is narrower than the phrase 'arbitrary masses' in the abstract may suggest. TVID 2.0 cannot handle IR-divergent inputs and does not check for them; several IR-finite but special parameter combinations are known to require treatment that is not yet implemented; the O(epsilon) terms T...,delta are omitted and are assumed to cancel in physical observables; and the mass-derivative masters U6m1, U6m3, U6n2, U7a1, U7a2 are evaluated by numerical differentiation, reducing their precision to 6-7 digits. These caveats are honestly stated, but the abstract and conclusions should reflect them, and the benchmark section should identify which master integrals and parameter regions are covered by the claimed 6-10 digit precision. As written, the accuracy claim applies only to a non-excluded, non-special subset.","section":"Sec. 4, limitations; Sec. 5"}],"minor_comments":[{"comment":"The sentence 'An simple method...' should read 'A simple method...'.","section":"Sec. 3.2"},{"comment":"'In princple' should be 'In principle'.","section":"Sec. 3.3"},{"comment":"'The numerical of TVID uses quadruple precision...' appears to be missing the word 'part'; should read 'The numerical part of TVID...'.","section":"Sec. 4"},{"comment":"'TVID 2 contains also contains all the basic elements...' contains a duplicated verb; should read 'TVID 2 also contains...'.","section":"Sec. 5"},{"comment":"The example directory is named 'vaccum.m' in the text; this appears to be a typo for 'vacuum.m'. The notation pm[n] for integration errors in UCallE output should be defined explicitly.","section":"Appendix B.2"},{"comment":"The asymptotic cutoff parameter scut = c * p^2 is said to be 'suitably chosen' with c depending on the function, but no values or convergence tests are reported. Please provide the chosen constants or a reproducibility test for the choice of scut.","section":"Sec. 4"}],"recommendation":"major_revision","confidential_remarks":"The paper reports a useful and potentially citable numerical package, and the overall construction is credible. The main blocking issue is the unresolved order-of-magnitude imaginary-part disagreement for U8a at p^2=40: the authors need either a third independent check (for example, direct numerical integration in a different variable set, an analytic threshold limit, or a dispersion-relation consistency test) or a clearly qualified accuracy claim for that topology. The U4a1 sign mismatch in Table 2 also needs a one-line explanation. I found no issues with novelty or attribution; the references and prior work are appropriately acknowledged."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this is a genuinely useful methods paper with one unresolved discrepancy in its flagship benchmark. The authors have built a public C/Mathematica program that evaluates a complete set of planar three-loop self-energy master integrals with arbitrary masses, using dispersion relations for sub-loop bubbles and a variant of Ghinculov's method that avoids complex contour deformation via a residue-plus-principal-value split. For almost all master integrals, the numbers in Tables 2 and 3 agree with FIESTA to the quoted precision, often in seconds to minutes rather than hours. That is a real advance for precision electroweak and Higgs calculations.\n\nWhat's new: this is the first public implementation covering this set of three-loop self-energy masters with two closed fermion loops, including doubled-propagator variants. The derivations in Sec. 3 and Appendix A are explicit for representative cases, and the limitation list in Sec. 4 is honest: no IR divergences, some integrals evaluated by numerical differentiation at 6-7 digits, O(eps) two-loop terms omitted because they should cancel in physical observables. The code is publicly available, which makes the paper stronger than a pure formalism paper. The citation pattern is fine; the reliance on the authors' own TVID 1 and Bauberger's thesis is natural since the methods genuinely build on those.\n\nThe soft spot is U8a at p^2=40. TVID returns an imaginary part of -0.16344185(2), FIESTA returns -0.016361(3), an order of magnitude apart, while the real parts roughly agree. The paper suggests FIESTA's contour-deformed sector decomposition may be unable to smooth the integrand for 7+ propagators, which is plausible, but no third independent method, analytic limit, or other cross-check is presented. U8a is the parent topology and the only benchmark that exercises the double-pole x,y integration, so the '6-10 digit precision' claim is not yet supported for the hardest case. I don't see this as a deal-breaker, but the authors should either add an independent cross-check for U8a or soften the claim to something like '6 digits for U8a, pending further verification.' The rest of the benchmark agreement is strong enough that I'd trust the program for the other topologies.\n\nFor whom: anyone computing three-loop self-energies in models with arbitrary masses, and anyone working on numerical multi-loop methods. The U8a discrepancy is also a useful cautionary tale about benchmarking against a single external code.\n\nRecommendation: send to peer review. The paper is substantial, reproducible, and the one open discrepancy is a normal referee issue, not desk-reject material. I'd ask for an independent check of U8a before acceptance.","headline":"A useful, honest methods paper whose 6-10 digit claim for the flagship U8a topology is not yet supported due to an order-of-magnitude imaginary-part disagreement with FIESTA that lacks a third cross-check.","tokens_in":24767,"tokens_out":3163,"would_cite":true,"duration_ms":29102,"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":"The authors show that every planar three-loop self-energy master integral needed for two closed fermion loops can be evaluated as a one- or two-dimensional numerical integral, with 6–10 digit precision for arbitrary masses, by combining…","keywords":["three-loop self-energy","master integrals","dispersion relations","numerical integration","arbitrary masses","planar topologies","TVID","dimensional regularization"],"falsifier":"Evaluate $U_{8a}$ for $p^2=40$ and squared masses $m^2_1=1.1,\\ldots,m^2_8=1.8$ using a third independent method, such as differential equations in $p^2$ matched to the known $p^2\\to 0$ limit, and compare the imaginary part: one method gives $-0.16344185(2)$, the other gives $-0.016361(3)$. The value reproduced by the independent method settles which code, if either, is correct.","tokens_in":23649,"feed_emoji":"🧮","tokens_out":14697,"duration_ms":125399,"temperature":0.7,"pith_summary":"This paper presents TVID 2, a program that numerically evaluates the planar-type three-loop self-energy master integrals needed for three-loop self-energy diagrams with two closed fermion loops, for arbitrary masses. The central claim is that every integral in this class can be reduced to a one- or two-dimensional numerical integral over analytically known functions, by combining sub-loop dispersion relations with a modified triangle-based method. If the claim is correct, precision calculations in the Standard Model and beyond gain a fast, automated way to compute these difficult three-loop contributions, avoiding the heavy resource demands of general-purpose sector-decomposition codes. The paper reports 6–10 digit precision in run times from under a second to about twenty minutes, with the caveats that IR-divergent parameter choices are excluded and that one benchmark comparison with the independent code FIESTA shows a large discrepancy for the most complex topology at large momentum.","feed_headline":"Planar three-loop self-energy masters become 2D integrals","feed_subtitle":"TVID 2 evaluates them with arbitrary masses to 6-10 digits in seconds to minutes.","key_machinery":"The two load-bearing devices are the sub-loop dispersion relation, which replaces each self-energy bubble by an integral over its discontinuity $\\Delta B_0(s)$ weighted by $1/(s-q_2^2-i\\varepsilon)$, and a modification of the triangle-based method of Ref. [15] that casts $U_{8a}$ (and $U_{7a}$) as a double integral over the one-loop vertex function $C_0(p^2,y,x,\\ldots)$ times a phase-space factor $\\sqrt{\\lambda(x,y,p^2)}$, with the propagator denominators $x-m_4^2$ and $y-m_5^2$ split into residue plus principal value. These devices reduce every master integral to at most two dimensions and avoid complex contour deformation.","core_discovery":"The paper claims that all master integrals descending from the planar three-loop self-energy topology $U_{8a}$ that are needed for three-loop self-energy diagrams with two closed fermion loops can be evaluated numerically as one- or two-dimensional integrals over known functions, with arbitrary masses. Topologies with sub-loop self-energies are handled by dispersion relations that replace each sub-bubble by an integral over its discontinuity; topologies without sub-loop self-energies ($U_{7a}$, $U_{8a}$, and their doubled-propagator variants) are handled by writing the integral in terms of the one-loop triangle function $C_0$ in the variables $x=q^2$ and $y=(q+p)^2$, splitting each propagator pole into a residue plus principal value. UV divergences are removed by subtracting known vacuum and two-loop integrals, and the finite remainders are integrated with adaptive quadrature in C. The paper reports 6–10 digit precision for most master integrals, in run times from under a second to about twenty minutes, and documents agreement with FIESTA for almost all cases except the $U_{8a}$ imaginary part at large $p^2$.","pith_inferences":["If a third method confirms TVID's value for $\\mathrm{Im}\\,U_{8a}$ at $p^2=40$, the discrepancy would point to a systematic failure of contour-deformed sector decomposition above threshold and would motivate re-benchmarking such codes on imaginary parts.","The residue/principal-value split separates dispersive and absorptive parts cleanly, so it could be combined with differential equations in the masses or in $p^2$ to produce semi-analytic results for the same master integrals.","The same two-technique template (dispersion for sub-bubbles, $C_0$ double integrals for the rest) may extend to the missing non-planar and Mercedes-star three-loop self-energy topologies.","Replacing the numerical mass differentiation used for $U_{6m1}$, $U_{6m3}$, $U_{6n2}$, $U_{7a1}$ and $U_{7a2}$ with analytic derivatives of the dispersion integrands would test whether the reduced 6–7 digit precision is a limitation of the implementation or of the representation."],"forward_implications":["Planar three-loop self-energy diagrams with two closed fermion loops can be evaluated numerically with 6–10 digit precision for arbitrary masses in run times of seconds to minutes.","The threshold region, where the integrals develop imaginary parts, is handled without complex contour deformation, so the reported precision does not degrade when $p^2$ lies above physical thresholds.","The $O(\\epsilon)$ parts of the two-loop subtraction functions cancel against counterterms in physical observables and therefore do not need to be evaluated explicitly.","The method currently excludes IR-divergent parameter choices and the non-planar and Mercedes-star three-loop self-energy topologies, which would need separate treatments.","For most master integrals the results agree with the independent code within integration errors, but for $U_{7a}$ and $U_{8a}$ the discrepancies exceed the reported errors, so the flagship topology's result needs independent confirmation before being used in predictions."],"supporting_citations":[{"why":"Supplies the sub-loop dispersion-relation technique for self-energy bubbles that is used for all topologies with sub-loop self-energies.","marker":"[13]"},{"why":"Extends the dispersion method and is cited together with [13] as the basis for the two-loop self-energy treatment.","marker":"[14]"},{"why":"Provides the triangle-based representation for integrals whose sub-loops are triangles, which Secs. 3.2–3.3 adapt to avoid complex contour deformation.","marker":"[15]"},{"why":"The predecessor paper supplies the subtraction scheme for UV divergences, the residue/principal-value prescription of eq. (10), and the three-loop vacuum integrals used in the subtracted terms.","marker":"[10]"},{"why":"Supplies the independent numerical results used for benchmarking in Tabs. 2 and 3.","marker":"[8]"},{"why":"Provides the analytic expressions for the one-loop functions $B_0$ and $C_0$ that appear as integrands in the dispersion and triangle representations.","marker":"[19]"},{"why":"FIRE 5 is used for the integration-by-parts reduction that defines the set of irreducible master integrals.","marker":"[17]"}],"fun_headline_variants":["TVID 2: 3-loop self-energy masters now with any masses","Planar 3-loop self-energies: now 2D integrals for any mass","Arbitrary-mass planar 3-loop self-energies in seconds","Evaluating 3-loop self-energy masters with arbitrary masses","3-loop self-energies: reduced to 2D integrals, any mass"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The accuracy claim for the flagship topology $U_{8a}$ stands on the assumption that TVID's imaginary part is right and the independent code's is wrong at $p^2=40$, a choice the paper makes without a third independent cross-check.","fun_headline_variants_meta":{"raw":{"variants":["TVID 2: 3-loop self-energy masters now with any masses","Planar 3-loop self-energies: now 2D integrals for any mass","Arbitrary-mass planar 3-loop self-energies in seconds","Evaluating 3-loop self-energy masters with arbitrary masses","3-loop self-energies: reduced to 2D integrals, any mass"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001119,"raw_usage":{"total_tokens":4636,"prompt_tokens":904,"completion_tokens":3732,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":520,"completion_tokens_details":{"reasoning_tokens":3632}},"tokens_in":520,"tokens_out":3732,"duration_ms":28972,"temperature":1.0,"reasoning_tokens":3632,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T10:58:55.076454+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate $U_{8a}$ for $p^2=40$ and squared masses $m^2_1=1.1,\\ldots,m^2_8=1.8$ using a third independent method, such as differential equations in $p^2$ matched to the known $p^2\\to 0$ limit, and compare the imaginary part: one method gives $-0.16344185(2)$, the other gives $-0.016361(3)$. The value reproduced by the independent method settles which code, if either, is correct.","supporting_citations":[{"cited_title":"On the evaluation of three--loop scalar integrals in the massive case","cited_arxiv_id":"hep-ph/9604333","evidence_quote":"Provides the triangle-based representation for integrals whose sub-loops are triangles, which Secs. 3.2–3.3 adapt to avoid complex contour deformation."},{"cited_title":"Three-loop vacuum integrals with arbitrary masses","cited_arxiv_id":"1609.09159","evidence_quote":"The predecessor paper supplies the subtraction scheme for UV divergences, the residue/principal-value prescription of eq. (10), and the three-loop vacuum integrals used in the subtracted terms."}],"review_version":1}