{"id":"f185bcc1-4f9e-4f6b-b76f-04b7723f7795","arxiv_id":"2412.13043","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":2,"one_line_summary":"DiFfRG provides a general, code-generated numerical toolkit for fRG flows, demonstrated on O(N), Quark-Meson, Yang-Mills, and four-fermion systems.","lead":"DiFfRG is a new open-source C++ framework for solving functional Renormalisation Group flow equations, using finite elements for field dependences, large momentum-dependent vertex expansions, GPU support, and automatic Mathematica-based code generation. It ships four complete example applications, including a Yang-Mills calculation whose gluon propagator matches lattice data.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The framework's 'very general, quantitatively reliable' claim rests on Eq. (10) and on stable FEM discretisations of field dependences, but the paper provides no convergence test for the FEM sector; only the vertex/variable sector is quantitatively checked against lattice data.","rationale":"I read the paper as a software framework paper whose central claim is that DiFfRG provides a general, reliable, and reproducible toolkit for fRG flows, with finite elements for field dependences as a stand-out feature. The shipped code, four complete examples, and the lattice comparison for Yang-Mills are genuine evidence and should be credited. The reader's weakest assumption correctly identified Eq. (10) as the gatekeeper for the generality claim. My stress-test sharpens that concern: the paper does not just fail to stress-test Eq. (10) across many truncations; it also provides no numerical convergence evidence for the FEM sector at all. Since the FEM field-dependence machinery is the paper's primary novelty and the basis for 'quantitatively reliable computations of field dependences', the absence of a single mesh/convergence study is a load-bearing gap. I do not think this warrants rejection: the framework may well work, and prior works cited by the authors give supporting evidence. But the reliability claim should be conditional until such tests are supplied. I therefore leave the reader's CONDITIONAL verdict unchanged, with partial agreement on the precise location of the weakness.","tokens_in":46038,"tokens_out":6310,"duration_ms":71549,"concrete_test":"Run the shipped O(N) finite-temperature example (Examples/ONfiniteT) at fixed physical parameters and compare the final effective potential at k = 0.01 GeV for increasing FE polynomial order (fe_order = 1, 2, 3) and mesh refinement (level = 0, 1, 2, 3) in the CG, dDG, and LDG variants. Compute the L2 error of the mass function against a high-resolution finite-difference reference solution. If the error does not decrease at the expected convergence order for all three assemblers, the 'quantitatively reliable' claim for field-dependent flows is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that DiFfRG is a comprehensive framework for 'very general truncation schemes' with 'quantitatively reliable computations of field dependences' hinges on the Ansatz in Eq. (10) (Section 2.2) and its concrete implementation in Eq. (25) (Section 5.2.2). There, the paper states: 'We make a very general Ansatz and assume that the flow equation of u in field space has the shape ...'. If some fRG flows, e.g. higher-order derivative expansions with fourth-order field derivatives or non-trivial multi-field invariants, cannot be stably cast into this local convection-diffusion form, then the advertised generality is overstated. More importantly, even for flows that do fit Eq. (10), the paper never demonstrates that the supplied CG, dDG, and LDG discretisations converge at the expected FE rates. The only quantitative external benchmark, the SU(3) Yang-Mills gluon propagator comparison to lattice data (Figure 4), uses the variable/vertex sector, not the FEM field-dependence sector. The O(N) and Quark-Meson FEM examples show qualitative phase diagrams and running couplings but contain no mesh-refinement study, no polynomial-order convergence check, and no error budget. Thus the abstract's 'quantitatively reliable' assertion is unsupported for the FEM part of the framework.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces DiFfRG, an open-source C++ framework for solving functional Renormalisation Group flows, and is written as a Computer Physics Communications software paper. The central features are finite-element discretisations of field-dependent quantities, large momentum-dependent vertex expansions, implicit and explicit time-stepping, Mathematica-based automatic code generation, and optional GPU support. The manuscript describes the theoretical setup, the structure of the code, and four included examples: finite-temperature O(N) theory, a four-Fermi QCD-motivated system, SU(3) Yang-Mills theory, and a Quark-Meson model in LPA'. The paper claims that DiFfRG is a comprehensive, extendable framework for very general truncation schemes and that its FEM treatment of field dependences is quantitatively reliable, with the Yang-Mills gluon propagator compared to lattice data and scaling exponents quoted as consistent with earlier work.","tokens_in":46272,"tokens_out":5277,"duration_ms":51308,"significance":"If the framework performs as advertised, DiFfRG would be a genuinely useful contribution to the open numerical fRG ecosystem: it combines FE discretisations of field space with vertex expansions, automated code generation, GPU acceleration, and four runnable examples. Important strengths are the concrete reproducibility claims, the availability of the full code, the automated Mathematica workflow, and the independent lattice comparison for the gluon propagator in Figure 4. The significance is tempered, however, by the fact that the quantitative reliability claim is not backed by convergence or error-budget tests for the FEM sector, and by the partly tuned or same-group nature of the other benchmarks. The paper is better supported as a description of a flexible and feature-rich framework than as a demonstration of quantitative precision across all advertised use cases.","major_comments":[{"comment":"The central quantitative claim for the FEM sector is not demonstrated. The abstract advertises 'quantitatively reliable computations of field dependences', but none of the FE examples (O(N) in Section 3.1, Quark-Meson in Section 3.4) includes a mesh-refinement study, a polynomial-order convergence check, or an error budget. The only independent quantitative benchmark, the Yang-Mills gluon propagator comparison in Figure 4, exercises the variable/vertex sector rather than the FEM field-dependence sector. A revision should add a systematic convergence test for at least one FEM flow, e.g. a manufactured solution or a known O(N)/Yukawa benchmark with h- and p-refinement, and report the resulting discretisation error together with the production parameters used for Figures 1 and 8.","section":"Section 2.2, Section 3.1, Section 5.2.2"},{"comment":"The 'very general Ansatz' is an assumption, not an established characterisation. The text states: 'We make a very general Ansatz and assume that the flow equation of u in field space has the shape ...', and this structure underlies all four FEM assemblers. It is not shown that every truncation the paper intends to support, in particular derivative expansions beyond LPA' with fourth-order field derivatives or potentials over several coupled invariants, can be cast into this local convection-diffusion form and solved stably with the supplied CG/dDG/LDG discretisations. The revision should either soften the generality claims or add an explicit analysis or classification of the equations covered, including a concrete example that probes the boundaries of the Ansatz.","section":"Section 2.2, Eq. (10); Section 5.2.2, Eq. (25)"},{"comment":"Several quantitative-sounding statements rest on tuned or same-group input. The four-Fermi example is explicitly called qualitative in the text and in the Figure 3 caption, and it uses vacuum input data from reference [9] without thermal or density corrections; the Quark-Meson initial conditions are tuned to m_pi = 140 MeV and m_q = 350 MeV; and the Yang-Mills gluon mass gap is tuned to the scaling solution. The independent lattice comparison in Figure 4 is a genuine strength, but it covers only the vertex sector. To support the abstract's overall claim of quantitative reliability, the paper should state for each example what is calibrated, what is independently verified, and what remains qualitative.","section":"Sections 3.2, 3.3, 3.4"}],"minor_comments":[{"comment":"The word 'extremly' should be 'extremely'.","section":"Program Summary"},{"comment":"The word 'consituting' should be 'constituting'; similar typos such as 'straight-forward' and 'adpative' in Appendix C should be corrected.","section":"Section 5.2.2"},{"comment":"The phase diagram is not accompanied by the numerical parameters used, such as mesh size, polynomial order, and tolerances; providing these in the text or in the shipped parameter files would improve reproducibility.","section":"Section 3.1, Figure 1"},{"comment":"Because the text states that the four-Fermi results are 'at best qualitative', the abstract's phrase 'four-Fermi flows in the QCD phase diagram' should make this caveat visible or label the example as schematic.","section":"Section 3.2, Figure 3"},{"comment":"The statement that the integration error of a medium-sized quadrature rule is of order 10^-4 should specify the test integral, the quadrature order, and the comparison method against which the error is measured.","section":"Section 5.3.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a good fit for Computer Physics Communications and the code appears genuine, with four reproducible examples and a valuable lattice comparison. The main obstacle is the mismatch between the abstract's claims of comprehensive, quantitatively reliable FEM computations and the verification actually supplied: no FE convergence or error-budget tests are presented, and the independent benchmark is in the vertex sector. This is fixable within the scope of a revision by adding targeted convergence studies and by making the generality and calibration caveats explicit. I therefore recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThis is a real software contribution, not a physics-results paper. What's new is the integrated combo: Mathematica-to-C++ code generation for fRG flows, FEM discretisation of field dependences, implicit-explicit time-steppers for stiff symmetry-breaking flows, and GPU support, all in an open GPLv3 library with four complete, runnable examples. The Yang-Mills example reproduces the gluon propagator against lattice data and the scaling exponents agree with [33]; that's a genuine end-to-end check of the machinery. The four-Fermi example is explicitly qualitative, and the O(N) and Quark-Meson examples are illustrative, which is acceptable in a toolkit paper.\n\nThe soft spots are real but not fatal. The abstract's claim of 'quantitatively reliable computations of field dependences' outruns the evidence: the only quantitative external benchmark, Yang-Mills, uses the vertex/variable sector, not the FEM sector. There is no mesh-refinement or polynomial-order convergence study for CG/dDG/LDG, and no error budget. The generality of the framework rests on the convection-diffusion ansatz in Eq. (10); the paper does not stress-test how far that form extends beyond the examples. I also found no mention of an automated test suite in the manuscript, which matters when much of the code is auto-generated and depends on external libraries. The reader's suggestions—document test coverage, pin a commit hash, add convergence tests—are exactly the right remedies.\n\nI would send this to a serious referee rather than desk-reject. With a revision that adds an FEM error budget and testing documentation, this belongs in CPC. My own verdict would be conditional accept.","headline":"A genuinely useful open-source fRG toolkit with real end-to-end benchmarks; the FEM error-budget evidence is thinner than the abstract suggests, but it deserves a proper referee.","tokens_in":46861,"tokens_out":2941,"would_cite":true,"duration_ms":29460,"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":"DiFfRG is an open framework for solving general functional renormalisation group flows.","keywords":["functional renormalisation group","finite element method","effective potential","vertex expansion","QCD phase diagram","spontaneous symmetry breaking","code generation","GPU computing"],"falsifier":"Take a truncation with an analytic benchmark, such as the large-N O(N) effective potential with a flat regulator, run it with DiFfRG, and test whether the finite-element propagation matches the exact convexity-restoration front; if the front lags, oscillates, or fails to converge under mesh refinement, the claimed general stability of the convection-diffusion discretisation is falsified.","tokens_in":45775,"feed_emoji":"⚛️","tokens_out":8457,"duration_ms":75002,"temperature":0.7,"pith_summary":"DiFfRG is a C++ framework that solves functional Renormalisation Group (fRG) flows in very general truncation schemes. The paper's core claim is that fRG flows can be treated as convection-diffusion equations in field space, and that finite element methods, combined with efficient time-stepping and automatic code generation from a symbolic derivation step, make large fRG simulations straightforward to set up and run reliably. This matters because fRG studies of QCD phase structure and competing orders require both fully field-dependent quantities, such as the effective potential, and large momentum-dependent vertex expansions, which usually force researchers into writing dedicated, hard-to-test codes. The authors ship four complete, reproducible examples, namely finite-temperature O(N) theory, a quark-meson model, SU(3) Yang-Mills theory, and four-Fermi flows, and report a GPU speedup of up to 4.6 times on the Yang-Mills example.","feed_headline":"Finite-element toolbox tames general fRG flows","feed_subtitle":"Four reproducible examples show O(N), quark-meson, Yang-Mills, and four-Fermi flows running on one code base.","key_machinery":"The central object is the general PDE Ansatz of Eq. (10): $m_i(\\partial_t u,u,t,x)+\\partial_{x_j}F_{ij}(u,t,x,\\dots)+s_i(u,t,x,\\dots)=0$, where $u$ holds the field-dependent quantities, $F_{ij}$ is the flux, $s_i$ is the source, and $m_i$ is the mass function. The paper casts fRG flows for field-dependent quantities into this convection-diffusion form and discretises field space with Legendre-basis finite elements of four types: continuous, discontinuous, direct discontinuous, and local discontinuous Galerkin. RG-time integration uses implicit differential-algebraic solvers for stiff flows and explicit multistep or Runge-Kutta methods for large momentum-dependent systems, with jacobians generated by automatic differentiation. A Mathematica-side package derives the flow equations and exports C++ integration kernels with both CPU and GPU back-ends.","core_discovery":"The paper's central claim is that DiFfRG, a modular open-source C++ framework, lets users build fRG simulations in very general truncation schemes: full field-dependent effective potentials handled by finite elements, large fully momentum-dependent vertex expansions, mixtures of both, and automatic code generation from a symbolic derivation step. This claim is supported by four shipped examples with complete code, namely finite-temperature O(N) theory, a quark-meson model, SU(3) Yang-Mills theory, and four-Fermi flows in the QCD phase diagram. The Yang-Mills example reproduces the expected gluon and ghost dressings and shows a GPU speedup of 4.6 times on the tested full truncation, and the framework's earlier results on the quark-meson effective potential become reproducible with the release.","pith_inferences":["Beyond the paper, the convection-diffusion formulation points to a direct transfer of the same finite-element machinery to real-time and Keldysh fRG flows, which also develop steep fronts in field space.","Beyond the paper, the implicit-explicit splitting suggests that momentum-dependent vertices could be stepped implicitly as well, potentially stabilising stiff mixed systems beyond the examples shown here.","Beyond the paper, extending the automatic code generation to emit complete application skeletons would lower the C++ entry barrier to near zero, widening access to fRG methods."],"forward_implications":["Full effective potentials with spontaneous symmetry breaking can be integrated reliably with implicit finite-element time stepping, including shock formation in field space.","Large momentum-dependent vertex expansions can be run on their own or simultaneously with field-dependent quantities, using extractors to pass equation-of-motion data between the two sectors.","The four shipped examples are claimed to be fully reproducible from the included code, making earlier DiFfRG-based physics results testable by any user.","GPU execution accelerates the flow evaluation by up to a factor of 4.6 in the tested Yang-Mills vertex expansion, with the same code running on CPU when no GPU is present.","The Mathematica-to-C++ pipeline automates the most error-prone parts of fRG work, so new truncations can be turned into running simulations without hand-deriving large systems."],"supporting_citations":[{"why":"Introduces finite-element discretisations of fRG effective potentials, the method DiFfRG extends.","marker":"[4]"},{"why":"Supplies the symbolic derivation of functional flow equations used in code generation.","marker":"[5]"},{"why":"Performs the Lorentz, Dirac, and group traces needed to project flows onto tensor structures.","marker":"[6]"},{"why":"Provides the tensor bases and projectors used to set up truncations.","marker":"[7]"},{"why":"Supplies the vacuum glue and quark data and the diagrammatic rules used in the four-Fermi example.","marker":"[9]"},{"why":"Demonstrates that discontinuous Galerkin methods capture field-space shocks in fRG flows.","marker":"[12]"},{"why":"Analyses numerical RG-time integration of the effective potential and motivates the implicit time-stepping strategy.","marker":"[18]"},{"why":"Supplies the finite-element mesh, basis, and assembly infrastructure DiFfRG builds on.","marker":"[21]"},{"why":"Provides the implicit and implicit-explicit time integrators used for stiff flows.","marker":"[22]"},{"why":"Defines the Yang-Mills vertex-expansion setup that DiFfRG reproduces and benchmarks.","marker":"[33]"}],"fun_headline_variants":["DiFfRG: finite-element fRG flows in any truncation","Open-source DiFfRG runs O(N), quark-meson, Yang-Mills, four-Fermi","GPU-accelerated fRG with DiFfRG's finite-element toolbox","DiFfRG: reproduce fRG flows from code, no black boxes","Finite elements plus vertex expansions: DiFfRG framework"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper's 'very general' applicability stands on the assumption that the fRG flow equations of interest can be written in the specific convection-diffusion form with mass, flux, and source terms, and that the included finite-element methods stay stable on those equations.","fun_headline_variants_meta":{"raw":{"variants":["DiFfRG: finite-element fRG flows in any truncation","Open-source DiFfRG runs O(N), quark-meson, Yang-Mills, four-Fermi","GPU-accelerated fRG with DiFfRG's finite-element toolbox","DiFfRG: reproduce fRG flows from code, no black boxes","Finite elements plus vertex expansions: DiFfRG framework"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000483,"raw_usage":{"total_tokens":2374,"prompt_tokens":924,"completion_tokens":1450,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":540,"completion_tokens_details":{"reasoning_tokens":1360}},"tokens_in":540,"tokens_out":1450,"duration_ms":12207,"temperature":1.0,"reasoning_tokens":1360,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T13:28:10.168742+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a truncation with an analytic benchmark, such as the large-N O(N) effective potential with a flat regulator, run it with DiFfRG, and test whether the finite-element propagation matches the exact convexity-restoration front; if the front lags, oscillates, or fails to converge under mesh refinement, the claimed general stability of the convection-diffusion discretisation is falsified.","supporting_citations":[{"cited_title":"QMeS-Derivation: Mathematica package for the symbolic derivation of functional equations","cited_arxiv_id":"2102.01410","evidence_quote":"Supplies the symbolic derivation of functional flow equations used in code generation."},{"cited_title":"Geißel, J","cited_arxiv_id":null,"evidence_quote":"Provides the tensor bases and projectors used to set up truncations."}],"review_version":1}