Home·The theorem·The proof in one reading·Architecture·Units·Manuscript·Glossary·FAQ·Work items·Data·Method·Provenance·Changes

Where the ideas come from

Contents

1. Main conclusions 3

1.1 Scope and vocabulary 5

1.2 Complete unit accounting 6

2. The source toolkits 7

2.1 OpenAI's six papers 7

N1: exponential decay, not an exact ultraviolet RG construction 7

N2: the principal direct comparator for the ultraviolet engine 7

N3: extends the engine to continuum fields and stronger comparisons 9

N4: isolated mass from a massive continuum input 9

N5: an exact mass via a second regulator 10

N6: XY uses a different mechanism 10

2.2 Balaban's gauge toolkit 10

2.3 Balaban's N-vector work, Dimock and common constructive tools 12

3. Component-by-component provenance 13

01. Lattice, Haar measure and Wilson starting law — I.1 13

02. Gauge-invariant local fields, jets and engineering degree — II.1 13

03. Nilpotent markers and polymer/source algebra — II.2 14

04. Globally defined normalized block observation — III.1 (a) 14

05. Symmetric frame selection and integration gauges — III.1 (b) 15

06. Reference schedule, constant order and terminal shooting — III.2 15

07. Free quadratic action, covariances and localized rows — III.3 16

08. Sector change and localized Gaussian covariance — III.4 (a) 16

09. Local determinant ratios, Jacobians and joint Gaussian bounds — III.4 (b) 17

10. Absolute polymer sums and connected logarithms — III.5 17

11. Stationary nonlinear background with a uniform radius — III.6 18

12. Matching the Wilson energy through fourth order — III.7 (a) 18

13. Numerical one-loop slope and constructive matching — III.7 (b) 19

14. Connected activities, compulsory inventories and complete supports — III.8 19

15. Exact coupling reset and restoration of the activity class — III.9 20

16. Energy per seed and protection of rough sectors — III.10 (a) 20

17. Nonlinear contraction and bounds on irrelevant terms — III.10 (b) 21

18. Volume-uniform normalization and cancellation of gauge powers — III.11 (a) 21

19. Interacting/free rank and finite detector comparison — III.11 (b) 22

20. Reflection positivity, chessboard roots and derivative budgets — III.12 (a) 23

21. Fixed-volume convergence and analytic tuning — III.12 (b) 23

22. Extension to all labels and all finite orders — III.13 24

23. Iteration and selection of one construction with uniform bounds — III.14 24

24. Activity costs, local observable derivatives and analytic trace estimates — III.15–III.16 25

25. IV.1 — Infinite-volume extraction and Euclidean hierarchy 25

26. V.1 — Phase decomposition 26

27. V.2 — Reconstruction 27

28. V.3 — Reconstruction for every phase 27

29. VI.1 (a) — Ward identities and full rotation covariance 28

30. VI.1 (b) — Exterior flux at infinity 29

31. VI.2 — Maxwell reference theory and stress normalization 29

32. VII.1 — Short-distance freeness and finite detector systems 30

33. VII.2 — Trace anomaly and the all-radii asymptotic 31

34. VII.3 — Nonfreeness 31

35. VII.4 (a) — Separated OPE 32

36. VII.4 (b) — Contact extensions 33

37. VII.5 — Leading-log coupling asymptotics 34

38. VII.6 — Flexible scaling and arbitrary radius sequences 34

39. VIII.1 — Common-witness assembly 35

4. Cross-cutting comparisons 35

4.1 Where the ultraviolet construction ends 35

4.2 Reflection positivity and chessboard estimates 36

4.3 Composite insertions and observables 36

4.4 Running coupling, tuning and normalization 37

4.5 Order of constants and schedule growth 38

4.6 Correlations, full mesh limits and volume limits 38

4.7 Rotation restoration, phases and nonfreeness 39

5. What appears new, and what is routine 39

5.1 Ranked candidates for substantive new work 39

5.2 Routine or established ingredients that should receive ordinary credit 41

6. Open questions, limits and priorities for expert checking 42

6.1 Coverage and source reliability 42

6.2 Missing or incompletely accessed literature 42

6.3 Highest-value unresolved technical comparisons 43

7. Bibliography and source access 44

7.1 Lean source and orientation 44

7.2 Six OpenAI preprints 44

7.3 Balaban and related works: 24 sources 44

7.4 Other literature actually used 46

7.5 Background references and unread originals 49

Source keys 49

An analysis written by an AI model (GPT-6 Astra), from the Lean source and the literature listed at the end. It asks which parts of the proof follow Balaban's programme, which follow OpenAI's 2026 sigma-model papers, which are standard and which appear new. It identifies the closest antecedents among the sources read, not borrowing or priority. It has not been reviewed by the authors of the works it discusses.

Contents

Short keys such as B-RG-I (Bałaban's papers), N2 (OpenAI's O(4) paper) or GK85 are listed in the source keys at the end; full publication details and reading limits are in §7.

1. Main conclusions

N2 remains the closest inspected match for the particular ultraviolet density representation; the gauge analytic core has stronger direct antecedents in Bałaban, and several apparently distinctive polymer devices are already in Gawędzki–Kupiainen. The match to Sharp mass bounds for the two-dimensional O(4) model (N2), with developments in the canonical O(3) paper (N3), concerns a conjunction of formulas. It does not establish that the proof's ideas as a whole are closer to the NLSM papers. The five 1985 gauge papers directly supply nonlinear averages, controlled changes of gauge, analytic stationary backgrounds and regular background propagators. Magnen–Rivasseau–Sénéor supplies a further substantial four-dimensional gauge UV construction. Neither those antecedents nor the NLSM comparison supplies the exact stated all-composite, phase, stress and OPE endpoint. The converse description, “essentially Balaban, with the remainder just formalization,” is also unsupported. Confidence is high in the individual formula matches and older antecedents, medium in ranking the closest complete representation, and low in claims of worldwide novelty.

Four observations determine this verdict.

  1. The exact density format is unusually close to N2. The clipped kinetic term, masked square, regular errors with subtraction conditions, compulsory covering inventories, fixed reference thresholds, window inverse columns and exact reset occur together in N2 §5, pp.38–55. The corresponding objects occur in YM/RG/PathActions.lean:210–242, YM/RG/Activities.lean:42–129, YM/RG/History.lean:63–85, YM/SectorNext/GAUSSIAN_SECTOR_CHANGE/Windows.lean:65–74 and YM/AnalyticBridge/VERIFIED_RESET_STEP/Evaluation.lean:71–95. The conjunction is a close match, substantially adapted to gauge variables. Mandatory covering and selective exponentiation alone are older: GK85 p.212 Eq.(4.8) and pp.218–221 Eqs.(5.28)–(5.53) already separate bad-region polymers from the regular exponential.

  2. Several tempting NLSM attributions are actually older. Balaban already has symmetric nonlinear group averages and normalized gauge kernels, B-RG-I pp.252–254, Eqs.(0.4)–(0.15), as well as marginal Maxwell extraction, pp.258,292–297. The relevant Lean counterparts are YM/RG/Patch.lean:102–179 and YM/RG/Activities.lean:70–75. B-Gauge Theorem 2 p.83 and B-Var Theorem 1 p.279, Proposition 6 pp.295–296 directly establish controlled gauge changes and analytic stationary backgrounds. B-Back Theorems 3.1–3.4 pp.397–400 gives regular-background propagators. Compare YM/PCalc/Chart.lean:78–100 and YM/Work/F04.lean:62–86. GK85 p.222 and BY90 pp.360–363 also make the older order of constants and analytic RG control explicit.

  3. The gauge comparison and continuum identification require separate treatment. The determinant budget at YM/Next/RANK/Determinant.lean:240–274 cancels gauge-dependent volume powers; it is not N3 §7's rough-region filling argument. The Ward/Maxwell/phase/trace route in YM/Next/ROTATIONS/Route.lean:114–123, YM/Phase/Mixture.lean:65–86 and YM/TraceExtension/Verse/Cape.lean:53–88 differs from N3 §§10–11,15. MRS93 §VI pp.368–369 already controls background-dependent gauge/ghost determinant normalization; the exact torus cancellation remains unmatched in the passages read, not established as novel. Much of reconstruction is standard under the corrected OS75 hypotheses; the potential contributions are narrower combinations and uniform estimates.

  4. The endpoints differ on both sides. B-Large-II, Theorem 1, p.355, expressly assumes small effective couplings and leaves physical observables for further work, p.356. MRS93 p.327 states a fixed-IR ultraviolet Schwinger-function limit and Slavnov identities, while expressly leaving a fully detailed proof unwritten; p.329 excludes the construction there of renormalized physical composite observables. N3 Theorems 1.1–1.2, pp.5–7, assert a canonical massive limit. ChallengeSpec.lean:177–188 instead permits a volume subsequence and does not assert a positive mass gap. Unique vacuum in each phase is not a mass-gap statement: compare PublicTheory.lean:270–278 with YM/EveryPhase/Decay.lean:173–192.

A defensible size comparison is possible, but not a measurement of intellectual ownership. The size comparison counts the union of source spans of declarations used by the main theorem within each of the 32 units. The total within the units is 410,299 lines; the global union, including declarations outside those units, is 410,820. A declaration's span includes statements, proof code and internal comments, so this is an implementation weight. It is neither the physical repository line count nor a count of mathematical ideas.

Closest organizational comparison Units Declaration-span lines Share of lines within the units Share of 32 units
NLSM-nearest exact UV format and iteration 7 135,357 33.0% 21.9%
Balaban-nearest gauge background/free/marginal engine 3 48,591 11.8% 9.4%
Mixed gauge comparison, all-label extension and observable bounds 4 90,943 22.2% 12.5%
Common lattice, algebra and RG tools 5 31,375 7.6% 15.6%
Continuum, identification and final assembly 13 104,033 25.4% 40.6%

These are disjoint groups of units, not exclusive provenance shares. The first group contains considerable Balaban and standard mathematics; the second contains standard Gaussian and perturbative mathematics. Neither the mixed gauge-comparison group nor the continuum group can be attributed wholesale to either author group. The unit membership is recorded in §1.2, and the evidence for each comparison follows in §3. Under a stricter definition, retaining only III.4, III.8, III.9 and III.10 as NLSM-nearest gives 24.1% of the implementation. Thus the conclusion survives excluding the more debatable block-observation, schedule and iteration comparisons. It does not justify saying that a majority of all mathematical ideas originated in the NLSM papers.

These conclusions concern closest mathematical antecedents among the sources read. They do not establish causal borrowing, author intentions, access histories, or priority over literature not read here. The September–October preprint dates precede the October Lean source version, but that chronology alone proves no dependence.

1.1 Scope and vocabulary

Claims about the proof's mechanisms rest on the Lean definitions and proofs cited below. The statement.md, summary.md and unit outlines provide orientation; their assertions about the literature are not evidence for these comparisons. The counts concern declarations used by the main theorem, as identified by the existing dependency information. This is a comparison of mathematical mechanisms, not a certification of the Lean theorem or of the six preprints.

“Same idea/device” means a fairly specific construction with the same role and structural constraints. “Same general method” means an established technique compatible with many implementations. “Superficial resemblance” means common terminology or a common goal without a matched mechanism. Those distinctions matter particularly for the words sector, Gaussian, annulus, locality, nonfree and continuum.

The classifications used below are:

A mixture's proportions describe rough mechanism weight within that component, rounded to broad increments. They are judgement calls, not fractions of proved originality; changing them by 15–20 percentage points is often reasonable. For a split component, the reported size is its parent unit's allocation unless an actual file size is stated. Repeated parent sizes must not be added. Confidence refers to the correspondence; confidence in universal novelty is lower throughout.

Lean links identify exact passages at commit 3456e5e75cf8d4cd558955a5c7638fba2d68f464. N1–N6 refer to the six OpenAI preprints; their printed and PDF page numbers agree. Balaban citations use printed journal pages. Dimock citations use the preprint pagination of the dated versions specified in §7, which differs from journal pagination. The bibliography distinguishes full-text access with selective reading, partial extracts, abstracts and references cited only for established background.

1.2 Complete unit accounting

The units locate material; they do not imply that a declaration belongs mathematically to only one component. Within each unit, overlapping declaration spans count once. Of the declarations used by the main theorem, 425 fall outside the units: 98 have usable source intervals and 327 do not. The intervals outside the units add 521 lines to the union. The older outline total, 413,085, is not the denominator for these percentages.

Unit Lines Mechanism/group of units
I.1 3,367 Lattice/Haar/Wilson setup; common
II.1 10,793 Labels and jets; common
II.2 3,613 Marker and polymer algebra; common
III.1 4,133 Block observation and gauges; NLSM-nearest format, mixed ancestry
III.2 18,129 Schedules and shooting; NLSM-nearest organization
III.3 8,966 Free covariances; Balaban-nearest gauge task
III.4 37,825 Sector Gaussian integration; NLSM-nearest implementation
III.5 9,671 Absolute cluster sums; common
III.6 19,698 Stationary background; Balaban-nearest gauge task
III.7 19,927 Wilson matching and marginal flow; Balaban-nearest gauge task
III.8 4,103 Connected activities and supports; NLSM-nearest format
III.9 7,162 Exact coupling reset; NLSM-nearest format
III.10 49,766 Nonlinear step; NLSM-nearest format
III.11 42,556 Gaussian/interacting comparison and rank; mixed
III.12 18,944 Chessboard and fixed-volume convergence; mixed
III.13 22,077 All labels and growing orders; mixed
III.14 14,239 One construction and iteration; NLSM-nearest organization
III.15 7,366 Observable cost and propagation; mixed
III.16 3,931 Analytic trace estimates; common
IV.1 7,336 Infinite-volume extraction; continuum
V.1 5,718 Phase decomposition; continuum
V.2 6,419 Reconstruction; continuum
V.3 4,063 Every-phase reconstruction; continuum
VI.1 7,689 Ward identities and rotations; continuum
VI.2 14,170 Maxwell comparison and stress; continuum
VII.1 8,004 Short-distance fields; continuum
VII.2 20,645 Trace anomaly; continuum
VII.3 3,303 Nonfree witness/free detection; continuum
VII.4 10,193 OPE and contacts; continuum
VII.5 232 Leading-log asymptotic; continuum
VII.6 4,205 Flexible scales; continuum
VIII.1 12,056 Common-witness assembly

The word “continuum” in this table includes identification and assembly, not a single antecedent class. III.11 and III.12 also contain continuum-facing arguments. Conversely some exterior-flux dependencies are included in upstream units, so line shares cannot locate all conceptual boundaries perfectly.

2. The source toolkits

2.1 OpenAI's six papers

The papers are distinct and must not be treated as a single “NLSM method”. The following descriptions concern their stated results and actual constructions; these descriptions do not certify those results.

N1: exponential decay, not an exact ultraviolet RG construction

Theorem 1.1, p.1, bounds the spin two-point function on every finite free-boundary subgraph of the square lattice, uniformly when ferromagnetic bond strengths are weakened or deleted. The result does not assert a full transfer gap; p.2 explicitly separates spin correlations from invariant observable sectors. Its distinctive analytic input is a fourth variation in two noncommuting rotation directions, Proposition 2.2, pp.5–8, Eq.(2.8). The finite-graph Ward identities permit arbitrary prescribed boundary spins. The commutator calculation is explicit: Eq.(2.10) is a change of variables, Eq.(2.11) gives the quadratic commutator field, and Eq.(2.12) cancels the fourth-order mixed coefficient for generator pattern \((a, b, a, b)\). The summed fourth derivative has a nonnegative-square contribution; it is not a perturbative effective-action calculation. Section 3, pp.9–12, uses signed frequency families on annuli: the commutator response gains \((\log N)^2\), whereas the error has size \(\log N\). The boundary twist partition-function ratio tends to one, uniformly under deleting bonds and vertices. Section 4 converts this into rare embedded-Ising sign-cluster crossings. Simultaneous boundary pinning allows disjoint annuli to be compared independently; a coarse-path count then gives exponential decay. The final reduction from n components to three conditions the remaining coordinates and uses uniformity under weaker bonds. Provenance implication: attributing the YM ultraviolet machinery to N1 on the basis of “non-Abelian fluctuations” would be a superficial resemblance. The N2/N3 retained-density constructions are the relevant comparators for that machinery.

N2: the principal direct comparator for the ultraviolet engine

The result, Theorem 1.1, p.3, is a full lattice transfer-gap bound of order \(\sqrt{\beta} \hspace{0.17em} e^{- \pi \beta}\). It is a lattice theorem; §11, p.81, separately discusses assumptions needed for a continuum spectral interpretation. Section 2, pp.9–11, turns a small rectangular partition-function defect into a full transfer-gap estimate by positivity and time/space doubling. Section 3, pp.12–33, proves preliminary all-observable mixing using quaternion current identities, stiffness reduction, disagreement propagation, and conditional Ising/rotator comparisons. This preliminary infrared input is logically separate from exact weak-coupling blocking. Section 4, pp.34–38, constructs scalar block precisions, conditional means and covariances, and a positive edge-stack identity. Section 5, pp.38–55, gives the exact nonlinear map, including rough configurations, and its comparison estimates. Section 6, pp.56–61, fixes the kinetic drift through finite-volume coefficient comparisons. Section 7, pp.61–65, admits trajectories and shoots to a fixed terminal kinetic coupling. Section 8, pp.65–77, compares complete positive densities with signed mandatory covering expansions. Section 9, pp.77–79, transfers a preliminary finite-box gap test from a fixed reference cutoff to all finer cutoffs. Section 10, pp.79–81, obtains an upper mass bound from correlated block observables. Appendices A–D supply the uniform free estimates, a fixed coefficient norm, complete supports of local sector factors, and joint Gaussian product majorants.

The most useful distinctive devices for comparison are these:

  1. Tagged normalized observation, Eq.(5.2), p.39: use the normalized weighted spin mean on its good branch and a weighted constituent spin on its fallback branch; normalize the output noise exactly.

  2. Clipped-plus-masked-square kinetic energy, Eq.(5.3), p.39: a quadratic direct term below threshold, a linear direct term above it, and a masked localized quadratic term.

  3. Fixed reference thresholds, Eq.(5.1), p.38: all geometry uses \(H_j\), while the actual running coupling varies in a band around it.

  4. Complete read supports, pp.39–41 and Appendix C: record masks, eligibility tests, negative queries, windows, and lifted winding paths, not merely visible field arguments.

  5. Two activity classes, Eqs.(5.7)–(5.10), pp.40–41: differentiable regular errors with zero affine Hessian; merely measurable signed large-gradient weights with nonempty inventories.

  6. Mandatory exact cover, Eq.(5.9), p.40: disjoint supports must cover precisely the bad-edge set. This is not an ordinary optional polymer gas.

  7. One Gaussian for a fixed sector, Eqs.(5.18)–(5.21), p.44: window inverse columns, symmetrized finite-range covariance, explicit density-ratio and determinant residuals.

  8. Core reserves and protector supports, Lemma 5.2, pp.43–44; Appendix C: energy per primary seed pays entropy, bad-region normalization, and derivatives through hard inputs.

  9. Joint products before expansion, Lemma 5.3, pp.47–48; Appendix D: overlapping factors use the same Gaussian; endpoint quadratic loads are controlled collectively.

  10. Inventory-preserving connected reassembly, §5.4, p.49: logarithm only for empty-inventory components; forbidden background polymers are opened and attached to seeded components.

  11. Exact kinetic reset, Eqs.(5.40)–(5.42), pp.54–55: remove constant/affine Hessian pieces and put mask-complement corrections into the compulsory gas.

  12. Cutoff matching before comparison, §7: the running marginal coordinate must agree before contracting all other endpoint data.

N3: extends the engine to continuum fields and stronger comparisons

N3 directly states the inherited scalar and density ingredients in §§3–4; §3.3, pp.23–28, explicitly defines the adapted retained-density class. Its target is \(S^2\), not the quaternion group \(S^3\); this changes the rotation algebra, physical tangent dimension, Jacobian, and drift. Theorem 1.1, pp.5–6, uses susceptibility and second-moment length to normalize the full continuum limit along all diverging real bare couplings. Theorem 1.2, pp.6–7, first constructs the theory along a terminal-coupling trajectory. Section 2, pp.9–21, proves preliminary mixing with a strict saving in \(e^{(4 \pi - \eta) \beta}\). Section 4, pp.29–39, includes ordinary and inclined first observations, their free kernels, and common history estimates. Section 5, pp.39–55, strengthens the regular contraction towards \(L^{- 2}\) using inversion/tangent-reflection cancellations. Its ambient completion, Lemma 5.2, pp.42–43, adjoins a normalized scalar Gaussian to the two-dimensional tangent field, producing an \(\mathbb{R}^3\)-valued common Gaussian without adding a physical degree of freedom. The scalar normalization cancels one color in the determinant, p.44. Sections 6–7, pp.55–69, shoot the terminal coupling and establish an integrated comparison against one positive reference law. Proposition 7.1, p.61, allows the comparison activities to be signed or complex; only the base density is required to be positive and reflection positive. Lemmas 7.3–7.4, pp.62–65, combine chessboard exponential moments, local Lipschitz fillings of rough regions, cap entropy, and retained energy gain. The covering identities keep the original region geometry fixed during replacements, §7.4, p.65. Section 8, pp.69–74, transfers a fixed-box estimate to a physical mass gap uniform in cutoff. Section 9, pp.74–82, carries independent local complex spin sources; normalizing their aligned linear response gives the product of source-normalization factors. Sections 10–11 construct Schwinger distributions, restore Euclidean rotations, reconstruct, and prove a separated nonzero fourth cumulant. Sections 12–15 remove the bounded bare-offset ambiguity. The key additional ingredients are summable sensitivity of the kinetic error (Lemma 12.2, pp.97–98) and two incommensurable offset periods arising from ordinary and inclined blockings (§15.2, pp.118–119). Not transferable without argument: these last uniqueness, mass-gap, and rotation mechanisms do not appear merely from recognizing the common ultraviolet density class.

N4: isolated mass from a massive continuum input

Theorem 1.1, p.2, asserts a positive isolated lowest atom of the spin-field Källén–Lehmann measure. The paper takes the constructed O3 field and its cutoff-uniform physical gap and factorial moments as inputs; Proposition 2.1, pp.4–5, records them. Section 3, pp.6–9, derives cylinder traces and torus-to-plane comparisons. Section 4, pp.9–14, bounds covariance of even spin polynomials quadratically in spin two-point correlations, separating even and odd transfer sectors. Section 5, pp.14–19, attaches quantitatively visible spin weight to low-energy states through embedded sign clusters and circuits. Section 6, pp.19–21, obtains a linear-in-circumference state count below the two-excitation threshold using trace doubling and positive polynomial energy cutoffs. Section 7, pp.21–25, combines momentum windows with the state count to prove that only finitely many masses lie near the bottom of spectral support. The proof does not obtain a particle pole from exponential decay alone. Provenance implication: its use of positive spectral measures is standard shared machinery; its two-dimensional cluster/circuit/state-count mechanism is not established as a YM antecedent by the inspected Lean.

N5: an exact mass via a second regulator

Theorem 1.1, p.3, states the exact prefactor \(32 e^{\pi / 4 - 1 / 2}\) for the full O4 transfer gap in original lattice time units. Sections 2–3, pp.6–16, compare the spin-visible mass with the full spectrum and a rectangular partition-function rate. Sections 4–6, pp.16–55, introduce a Brownian U(2) regulator, a particle expansion, a commuting contact Hamiltonian, integral equations, finite-circle Bethe states and a spectral maximization over particle parities. These are distinctive integrability and regulator-specific steps; no basis was found for importing them into the YM comparison. Sections 7–9, pp.55–97, adapt the retained-density engine to four kinetic coefficients: spatial and temporal stiffnesses for each of the quaternion and circle fields, a real circle lift, complex weights, and integer curvature. The observation in §7.4, p.61, uses normalized mean/fallback noise on flat blocks and a medoid with randomized symmetric ties when curvature is present. Section 7.7, pp.65–68, retains the same window Gaussian, determinant residuals and joint-factor logic. Section 9.2, pp.87–88, shoots four parameters rather than a single coupling. Sections 9.3–9.6 compare the possibly complex Brownian expression by integration against the positive flat reference measure, including growing rectangular aspect ratios. Section 10, pp.98–104, determines a finite coupling shift by a helicity response, separately from the partition-error estimate. Section 11, pp.104–107, extracts odd/even traces, removes the additional circle field, and transfers the auxiliary mass. Provenance implication: the generalized exact-density engine is relevant; Bethe states, U(2) loops, winding-sector parity and exact mass extraction are not evidenced as ingredients of this Lean construction.

N6: XY uses a different mechanism

Theorem statements in §1, pp.2–4, concern the critical spin field as an imaginary exponential of a Dirichlet GFF, normalized by exact center magnetization and a diagonal Green factor. Section 2, pp.6–12, imports critical height and pin limits and relates spins to magnetic height sectors. Section 3, pp.12–20, proves fixed-positive-radius-hole sector comparisons by Gaussian variational estimates. Section 4, pp.20–24, controls flat pins on long annuli. Section 5, pp.24–28, glues conditional interface kernels while summing the integer constant mode. Lemma 5.4, p.27, is a useful specific device: kernels in \((0, 1]\) with small average loss contract centered densities, so an arbitrarily poor but positive initial core eventually produces a last marginal bounded by two. Section 6, pp.28–35, cancels identical microscopic annular normalizations against the center observable and obtains separated correlations. Section 7, pp.35–43, uses Lee–Yang moments, integrable two-point bounds, collision removal and local Sobolev tightness. Provenance implication: this is not the retained-action/kinetic-reset construction of N2–N3. Calling any annulus or Gaussian comparison in YM an XY adaptation would be superficial without further evidence.

2.2 Balaban's gauge toolkit

Balaban's programme starts with compact-group link variables and preserves a complete effective density while integrating short scales. Its main analytic tasks are gauge fixing, regular localized propagators, nonlinear background fields, marginal extraction, and control of large-field regions through many iterations. The pertinent sources span 1983–1989; they are not a single theorem that can be invoked as an unconditional four-dimensional construction.

Block averaging and gauge fixing. B-Prop-I §A, pp.18–20, Eqs.(1.8)–(1.17), gives contour averages and their linear versions after axial fixing. B-RG-I pp.252–254, Eqs.(0.4)–(0.15), uses route symmetrization, analytic nonlinear group means and gauge-invariant kernels. The mean respects inversion, permutation, left/right multiplication and linearization. B-Average pp.19–20 Eqs.(10)–(15), Proposition 3 p.36 and Proposition 7 p.43 directly develop nonlinear averages, quadratic remainders and analytic background dependence. B-Gauge Theorem 2 p.83 and Proposition 7 p.100 establish controlled axial/Landau gauge transfer and its inverse. Thus nonlinear averaging, gauge reconciliation and regular slices are older mechanisms; neither symmetry averaging nor a soft normalized kernel distinguishes N2 by itself. B-Prop-I §C, pp.20–26, changes to a divergence-gauge formulation with projection operators to obtain regular propagators. The Lean comparison is YM/RG/Patch.lean:102–179, YM/RG/Block.lean:49–72, and YM/PCalc/Chart.lean:52–67; the formulas and division of gauge roles differ.

Variational problem and backgrounds. The nonlinear stationary background is built from a linear lift plus corrections satisfying local analytic equations. B-RG-II §1, pp.2–3, Eqs.(1.2)–(1.5), recalls those equations and localizes their propagators. B-Var directly supplies the gauge comparison: Theorem 1 p.279 gives the unique small critical orbit and regularity; Proposition 6 pp.295–296 gives the analytic small solution; Proposition 9 p.309 gives dependence on coarse averages and external background, with localized derivative bounds (190), p.308. Its fixed-average constrained problem differs from the Lean soft-kernel problem. The derivative calculations preceding (189) are expressly omitted on p.308; this is a comparison of stated mechanisms, not certification of those estimates. YM/PCalc/Chart.lean:78–100 and YM/Push/STATIONARY_UNIFORM/ImplicitRadius.lean:24–43, YM/Push/STATIONARY_UNIFORM/ImplicitRadius.lean:80–92 implement the corresponding task for the actual soft-block tree chart.

Propagators and random walks. B-Green §2, pp.574–585, patches local inverses by generalized random-walk expansions. B-Prop-I §F, pp.35–40, applies regularity/locality estimates to gauge propagators. B-Prop-II handles irregular domains and restrictions at several scales; see p.229 and Eq.(2.86), p.238. The traced Lean free rows instead arise from analytic symbols, Laurent coefficients and finite-period Fourier identities, YM/Work/F07.lean:34–81, YM/Work/F07.lean:109–112. B-Back Theorems 3.1–3.4 pp.397–400 supplies exponential decay, derivative/Hölder regularity and analytic background dependence; p.399 makes the constants independent of the admissible domain sequence. The difference is implementation and required geometry: Balaban himself uses Fourier and Laurent formulas elsewhere.

Small/large fields and localization. B-Conv §2, pp.254–262, represents complete densities by nested region histories with coupling-dependent cutoffs. B-Large-I pp.175–177 explains why one rare-region penalty cannot simply be carried unchanged through arbitrarily many four-dimensional steps. Its R operation replaces exhausted large-field densities by normalized local ones, with quotient corrections preserving the integral, Eqs.(0.2)–(0.6). B-Large-II §1 localizes and bounds these corrections. Lean's inspected interfaces retain forced inventories and explicit seed reserves, YM/RG/Activities.lean:42–55, YM/RG/Activities.lean:113–129 and YM/Pressure/Seeds.lean:34–68. The exact aged-region R replacement has not been identified as a separate step in that route. This is a bounded negative finding, not proof that no analogous local replacement occurs anywhere in the repository.

Localization, polymers and marginal flow. B-RG-II §§1–2, pp.2–22, makes nonlinear background dependence local and exponentiates connected corrections. B-RG-I p.258, Eqs.(0.28)–(0.30), subtracts the gauge-action marginal direction. Its §4 differentiates gauge Ward identities through fourth order; §5, pp.292–297, identifies the transverse Maxwell quadratic coefficient and higher-derivative remainder. The same gauge tasks appear at YM/PCalc/Chart.lean:104–109, YM/RG/Activities.lean:70–75 and YM/Next/STATIONARY/Level/Transition.lean:86–93. The particular compulsory-inventory gas and masked reset in Lean more closely match N2's later retained format.

The precise ultraviolet endpoint. B-Large-II Theorem 1, p.355, assumes all effective couplings lie in a sufficiently small interval. The next paragraph explicitly says that the argument removing the condition had not yet been published, so the theorem retains the assumption. Equation (0.1), pp.355–356, provides upper and small-field lower density bounds uniform in scale, torus and configuration. The same page leaves physical-observable analysis, including loops, for further work. B-RG-I Theorem 2, p.259, states the tuned SU(2) running objective while postponing its proof. These scope limits are explicit in the sources; they are not inferred from the absence of a later reconstruction chapter. The Lean has its own coefficient calculation and transition proof, YM/Next/STATIONARY/FullPlanned.lean:17–52, followed by observable and reconstruction work. None of this source comparison certifies that the latter work is mathematically correct.

What the comparison does not identify as literally used. The traced Lean route does not invoke Balaban's stability theorem as a black box. Its global nonlinear observation is a tagged soft kernel, not the hard block delta constraint of the initial linear Propagators formulation. Its free-row construction is not the same irregular multiscale-domain inverse construction. Its seed/activity interface is not demonstrably the same separate R replacement. Its stage schedule allows increasing orders and factors rather than simply fixing all such data once. The source pairs establishing these differences are given in components 04–08 and 16–17. They are departures from particular formulas, not evidence that localized integration, gauge symmetry or large-field control is new.

Gaussian coordinates and Haar corrections. B-UV3 pp.260–262, Eqs.(17)–(24), linearizes the nonlinear average, retains its Jacobian and the even Haar density, and separates Gaussian and nonlinear terms. Equation (63), p.272, localizes logarithmic determinants. Compare YM/SectorNext/GAUSSIAN_SECTOR_CHANGE/Ratio.lean:96–105 and YM/AnalyticBridge/RELATIVE_LOGDET_TRACE/Sector.lean:54–71. These are direct older gauge antecedents. B-UV3 Theorem 1 p.257 is three dimensional and superrenormalizable; it does not supply the four-dimensional marginal coefficient transfer in YM/Next/STATIONARY/Level/Transition.lean:153–178.

2.3 Balaban's N-vector work, Dimock and common constructive tools

A Balaban-versus-NLSM opposition is historically misleading if “Balaban” means only his gauge papers. B-RG-I p.260 already says the method also applies to two-dimensional group-valued chiral models. His later N-vector programme develops nonlinear backgrounds, RG equations, localization, fluctuation integration and large-field renormalization in a spin setting. B-N-II p.676 insists that localized outputs depend only on localized inputs and avoids artificial torus dependence by formulating the flow globally. B-N-Loc-I §§2–4 treats backgrounds, covariances and nested nonlinear compositions. B-N-Large §4 treats a last-step generating functional. B-N-Corr p.493 and its main results concern spontaneous magnetization and transverse Goldstone behavior in dimensions at least three. These are not the two-dimensional massive O(3)/O(4) conclusions of N2–N3, but they refute the idea that Balaban's interests stopped at vacuum partition functions.

B-Corr pp.1–2 presents a general effective-action method for correlations, with massive Gross–Neveu as its worked example. It is neither a pure-gauge observable construction nor one of the classical N-vector papers despite its inclusion in the same bibliography. The bibliography also includes a Higgs lower-bound paper and Cammarota's polymer paper; neither should be misidentified as a Balaban pure-gauge article.

Dimock I–III explain the Balaban-style small-field, large-field and convergence architecture for scalar \(\varphi^4\) in three dimensions. They are excellent sources for Gaussian normalization, domains and quantifier order, but cannot supply missing non-Abelian identities. D-I preprint p.10, Eqs.(74),(78), exhibits determinant normalization; D-II preprint p.29 states a hierarchy of choices. D-III preprint p.33, Theorem 2 and Corollary 1, establishes partition-function stability and remarks that correlations require further work.

Bry §§3–4, pp.138–148, supplies older tree bounds and interpolation formulas common to both programmes. B-RG-I pp.250–251 explicitly credits Gawędzki–Kupiainen's analytic effective-action viewpoint. GK85 supplies a direct comparison. Its §4 p.212 Eq.(4.8) requires disjoint polymers to cover the bad region, with each touching a bad component; §5 pp.218–221 exponentiates the outside components and restores the same density class after marginal extraction. These are specific older antecedents for YM/RG/Activities.lean:42–55, YM/RG/Activities.lean:113–129 and YM/AnalyticBridge/VERIFIED_RESET_STEP/Evaluation.lean:71–95, beyond generic polymer language. Its result is infrared asymptotic freedom of positive-coupling scalar lattice φ⁴₄, not a gauge UV theorem.

BY90 pp.353–356, Eqs.(1.7)–(1.10), uses an exact disjoint-set polymer exponential/logarithm, explicitly distinguishing its algebra from GK's hybrid small/large-field representation. Theorems A/B pp.360–363 extract local constant and quadratic gradient terms, shift the covariance and control regulator losses; Theorem B′ pp.383–384 treats complex covariances. Its analytic norm already sums derivatives of all orders, p.359 Eq.(1.13). The Lean's masked, compulsory-inventory density is not literally BY's whole gas. These older constructions reduce the independent NLSM evidentiary weight of exact reblocking, analytic families, marginal subtraction and all-derivative control.

3. Component-by-component provenance

There are 39 components below, covering all 32 units. Large units are split by mechanism; several mechanisms cross unit boundaries.

01. Lattice, Haar measure and Wilson starting law — I.1

Function and size. 3,367 lines. This establishes periodic links, compact-group integration, plaquette energy, partition function and normalized lattice expectations. It is the exact finite-dimensional starting probability model.

Evidence. YM/Wilson/Law.lean:29, YM/Wilson/Law.lean:55–99 defines energy, prior, bare density, partition and law. B-RG-I p.252, Eq.(0.2), starts with the same Wilson action and attributes it to Wilson. Wilson 1974, Confinement of quarks, is the original lattice-gauge reference; its broad role is cited from established knowledge, not a fresh full-paper reading.

Classification. STANDARD, high confidence; effectively 100% at the level of ideas assessed here. The abstract compact matrix group and formal symmetry bookkeeping are implementation generality. Sharing the starting model discriminates neither Balaban nor NLSM provenance.

02. Gauge-invariant local fields, jets and engineering degree — II.1

Function and size. 10,793 lines. Finite jets of radial-gauge potentials provide a common algebra of local composite labels. A weighted degree makes each bounded-degree space finite dimensional and supports later quotient and detector arguments.

Evidence. YM/Labels/Jet.lean:19–39 gives derivative weight and radial constraints. YM/Labels/Polynomial.lean:60–112 gives degree, finite dimensionality, invariant space and labels. The closest broad precedent is the local composite-operator framework: W69 §II pp.1500–1501 explicitly includes composite fields and is read in part; Z73 remains abstract-only, as specified in §7. MRS93 p.329 distinguishes such renormalized physical composites from its elementary gauge-fixed field insertions. N3 §9.1, p.75, instead introduces complex sources for a particular spin field, not this gauge-jet algebra. B-RG-I §4, pp.281–291, uses gauge Taylor identities but does not supply the all-label construction shown here.

Classification. Roughly STANDARD/ADAPTED-OTHER 75%, candidate-NEW organization 25%; medium confidence. The polynomial algebra is standard; the chosen common finite-dimensional system and its later compatibility requirements are specific to this construction. There is no evidence for an exclusive NLSM origin of the local-field concept.

03. Nilpotent markers and polymer/source algebra — II.2

Function and size. 3,613 lines. Truncated commuting source variables retain finitely many insertions and allow coefficient bounds for products and exponentials. This supplies bookkeeping shared by vacuum and observable expansions.

Evidence. YM/Marker/Truncated.lean:10–58 constructs nilpotence, finite exponential passage and coefficient budgets. N3 §9.1, p.75, likewise carries independent local complex sources through regular and covering lists. Bry §§3–4, pp.138–148, provides the older connected/polymer expansion setting; source differentiation is a standard extension of generating functions.

Classification. STANDARD 80%, ADAPTED-NLSM organizational affinity 20%; high confidence for the standard part, medium for the immediate comparison. The Lean's finite algebra is not identical to an analytic complex-source neighborhood. Its value is exact coefficient tracking, not a new physical source principle.

04. Globally defined normalized block observation — III.1 (a)

Function and size. Within III.1's 4,133 lines. Gauge-covariant fine-link data produce coarse observations in every configuration, including rough ones. Tags choose frames and fallback predictors; compact-group noise is normalized exactly.

Evidence. YM/RG/Block.lean:39–79, YM/RG/Block.lean:146–163 defines connector, likelihood, noise, density and kernel normalization. YM/RG/Patch.lean:159–179 implements the mean/fallback predictor. N2 §5.1, p.39, Eq.(5.2), has the normalized mean on the good branch and a weighted constituent-spin fallback on the bad branch. B-RG-I pp.252–254, Eqs.(0.4)–(0.15), already has symmetric routes, nonlinear group means and gauge-invariant kernels. Fed87, pp.18–20, also describes near-identity group averaging; its result is classical compatibility, not a quantum measure construction.

Classification. Approximately ADAPTED-NLSM 45%, ADAPTED-BALABAN 40%, STANDARD 15%; medium-high confidence. The specific total tagged construction is the close N2 match; normalized kernels or nonlinear means alone are not. Link transport between independently chosen frames is the substantive gauge adaptation missing from the spin formula.

B-Average pp.19–20 Eqs.(10)–(15), Proposition 3 p.36 and Proposition 7 p.43 directly strengthen the nonlinear analytic-average antecedent. They concern a local logarithmic average; YM/RG/Patch.lean:169–179 supplies the different total minimizing/fallback predictor. Confidence remains high in this concrete distinction, medium in the relative attribution of the whole observation.

05. Symmetric frame selection and integration gauges — III.1 (b)

Function and size. Also within III.1. A symmetric frame selector belongs to the observation, whereas bounded-height tree slices supply coordinates for integration. Keeping these roles distinct matters: a minimizing observation frame is not automatically the Gaussian integration gauge.

Evidence. YM/RG/Patch.lean:109–155 defines the minimizing regime, eligible frames and their masses. YM/Tree/Pines.lean:7–12, YM/Tree/Pines.lean:24–29 explicitly separates the minimizing frame and arbitrary rooted forests. YM/PCalc/Chart.lean:52–67 uses Lie-algebra coordinates on non-tree links. B-RG-I pp.253–254, Eqs.(0.11),(0.14), symmetrizes axial fixing and uses exponential gauge factors. B-Prop-I §C, pp.20–26, Eqs.(1.22)–(1.47), performs a different divergence-gauge change with projections.

Classification. Predominantly ADAPTED-BALABAN and STANDARD for controlled gauge transfer, with a specialized observation/forest implementation; high confidence in the older mechanism, medium in the implementation comparison. B-Gauge Theorem 2 p.83, Eqs.(1.36)–(1.39), gives a unique small axial-to-Landau transformation with regularity bounds; Proposition 7 p.100, Eqs.(1.139)–(1.145), gives the reverse correspondence. Generic “gauge synchronization” should therefore not carry candidate-new credit. The narrower distinction is reconciling this particular minimizing/fallback observation with an independently chosen bounded-height forest. YM/Tree/Pines.lean:7–12, YM/Tree/Pines.lean:24–29 separates them; YM/Work/F04/Choices.lean:78–86 compares tested germs under the branch hypotheses. No literal use of Balaban's complete gauge-change sequence is established by these passages.

06. Reference schedule, constant order and terminal shooting — III.2

Function and size. 18,129 lines. A reference inverse coupling determines thresholds and localization radii while the actual coupling is varied. Successive finite stages accommodate increasing derivative orders and choose a trajectory reaching a prescribed terminal band.

Evidence. YM/RG/History.lean:43–118 separates stage choices, hatted reference couplings, geometry and schedule conditions. YM/Shoot/Reach.lean:18–49, YM/Shoot/Reach.lean:112–140 uses surviving intervals and continuity to shoot the trajectory. N2 Eq.(5.1), p.38, and §7, pp.61–65, have the same fixed-reference geometry and terminal matching purpose. BBS19 §8.3, preprint pp.122–123, Proposition 8.3.1, provides an older surviving-interval construction of critical trajectories. D-II Theorem 3.1, preprint p.29, makes an older constructive order of choices explicit.

Classification. ADAPTED-NLSM 40%, STANDARD/ADAPTED-OTHER 40%, candidate-NEW growing-order organization 20%; medium-high confidence. N2 fixes a block factor and finite order; Lean lets factors/orders grow under one common run. The threshold monitor is \(e^{(\log \log \beta)^2}\), not N2's fixed power of \(\log H\). Neither shooting by continuity nor ordering constants is itself new.

GK85 p.211 Eq.(4.1) uses the numerical reference \(n_0 + n\) for small-field domains, p.221 Eq.(6.2) bounds the running coupling by inverse-reference envelopes, p.222 orders the constants, and p.246 uses nested intervals to tune the critical mass. MRS93 p.330 Eqs.(II.11)–(II.12) distinguishes tentative and exact effective couplings. BY90 pp.360–363 controls successive regulator losses with fixed block size. Thus reference envelopes, tuning intervals and constant schedules are older. N2 is closest to the particular terminal-coupling arrangement, while Lean's increasing stages and common gauge-composite obligations remain a specialized extension. Confidence in that closest-source ranking is medium; none of the older scalar couplings has the gauge beta-function normalization.

07. Free quadratic action, covariances and localized rows — III.3

Function and size. 8,966 lines. This constructs the free gauge reference, stationary lifts, square-root rows and exponential locality uniformly in finite period. These are the linear estimates needed before the nonlinear background can be controlled.

Evidence. YM/PREC/TreeMatch.lean:76–103 gives the coarse quadratic square-completion identity. YM/Work/F07.lean:34–81, YM/Work/F07.lean:109–112 constructs analytic-collar roots and all-period Fourier identities. YM/Next/FREE_KERNELS/Ground.lean:54–59, YM/Next/FREE_KERNELS/Ground.lean:69–120 identifies the concrete rows and uniform bounds. B-Prop-I §D–F, pp.26–40, especially Eq.(1.64)–(1.67), has the stationary Gaussian lift and comparison to Maxwell energy. B-Green §2, pp.574–585, and B-Prop-II §B, p.229, use generalized random walks to patch local inverses on irregular multiscale domains. N2 §4, pp.34–38, gives scalar block precisions and covariances, a useful but less gauge-specific comparator.

Classification. ADAPTED-BALABAN 45%, STANDARD spectral/Gaussian tools 40%, later adaptation 15%; high confidence in the mechanism distinction. The Lean's translation-invariant row construction is Fourier/Laurent based, with sector restrictions handled later. Balaban also uses Fourier and Laurent methods; the difference is the construction's organization and domains, not possession of Fourier analysis.

The direct regular-background antecedent is B-Back Theorems 3.1–3.3 pp.397–399 and Theorem 3.4 p.400: exponential decay, derivative/Hölder bounds and analytic continuation, with constants independent of admissible nested domains (p.399). This directly supports the Balaban background-regularity attribution; it leaves the Fourier/Laurent versus domain-local random-walk organization distinct.

08. Sector change and localized Gaussian covariance — III.4 (a)

Function and size. Within III.4's 37,825 lines. The exact integral is rewritten in tree-chart coordinates, shifted around a background and expressed against a localized Gaussian. Masks, tags, inactive variables and residual quadratic terms remain in the equality.

Evidence. YM/SectorNext/GAUSSIAN_SECTOR_CHANGE/Change.lean:63–71, YM/SectorNext/GAUSSIAN_SECTOR_CHANGE/Change.lean:95–103, YM/SectorNext/GAUSSIAN_SECTOR_CHANGE/Change.lean:125–135 records the actual retained factors and integral identity. YM/SectorNext/GAUSSIAN_SECTOR_CHANGE/Windows.lean:65–74 builds inverse columns on separate windows, then symmetrizes them. N2 p.44, Eqs.(5.18)–(5.21), uses exactly this window-column/symmetrization arrangement and retains the inverse defect in the density ratio. B-RG-II p.12, Eq.(2.5), is an older Gaussian-conditioning/localization predecessor, but less close to this exact finite-range covariance formula.

Classification. ADAPTED-NLSM 55%, STANDARD Gaussian identities 25%, ADAPTED-BALABAN 20%; high confidence. The symmetrized matrix is not asserted to be the true principal inverse; the residual is essential. Gauge null directions, active non-tree links and regions held fixed require new model-specific estimates. N3's ambient normal scalar completion, pp.42–44, differs from the inactive-coordinate padding in YM/SectorNext/GAUSSIAN_SECTOR_CHANGE/Completion.lean:34–45, YM/SectorNext/GAUSSIAN_SECTOR_CHANGE/Completion.lean:149–180.

B-UV3 pp.260–262, Eqs.(17)–(24), supplies a direct older nonlinear-coordinate change, Jacobian and Gaussian fluctuation decomposition. It strengthens the Balaban part of the comparison. The close N2 match is specifically the separate inverse-window columns and their symmetrization with a retained defect, not Gaussian change of variables or background expansion in isolation.

09. Local determinant ratios, Jacobians and joint Gaussian bounds — III.4 (b)

Function and size. Also within III.4. Log determinants and inverse defects are decomposed into localized chains; several factors are bounded under one Gaussian before regrouping. The Haar chart contributes an even, quadratic-order Jacobian correction.

Evidence. YM/AnalyticBridge/RELATIVE_LOGDET_TRACE/Sector.lean:54–71 retains both local determinant data and residuals. YM/AnalyticBridge/ANNULUS_RATIO_INSERTIONS/Arc.lean:121–166, YM/AnalyticBridge/ANNULUS_RATIO_INSERTIONS/Arc.lean:176–201 combines quadratic loads, polynomial insertions and rarity under a common law. YM/SectorNext/GAUSSIAN_SECTOR_CHANGE/Ratio.lean:96–105, YM/SectorNext/GAUSSIAN_SECTOR_CHANGE/Ratio.lean:125–145 establishes the quadratic Jacobian error. N2 p.44, Eqs.(5.19)–(5.21), Lemma 5.3, pp.47–48, and Appendix D provide the closest combined formulation. B-Large-II pp.356–358, Eqs.(1.2),(1.10), already retains determinant normalization; D-I preprint p.10, Eqs.(74),(78), displays the familiar determinant factor.

Classification. ADAPTED-NLSM 45%, STANDARD 40%, ADAPTED-BALABAN 15%; medium-high confidence. The same-Gaussian treatment of overlapping factors is the specific affinity; ordinary determinant calculus and Haar inversion symmetry are standard. Treating overlapping factors as independent would lose the actual content of the comparison.

B-UV3 p.260 explicitly identifies the Haar density as analytic, positive and even; p.261 Eqs.(19)–(22) retains the nonlinear Jacobian, and p.272 Eq.(63) localizes the log determinant. These are direct gauge antecedents for the Jacobian and determinant pieces. BY90 Theorem B p.362 and §8 pp.378–381 also bounds Gaussian integration of interacting local functionals collectively. The N2 affinity remains the particular inserted-factor estimates in its retained-density setting, not a claim that joint Gaussian control first appeared there.

10. Absolute polymer sums and connected logarithms — III.5

Function and size. 9,671 lines. Connected graph sums are bounded by trees, summed with support weights, and exponentiated back to the gas. Absolute convergence supplies marked-site and tail estimates usable in later source constructions.

Evidence. YM/Push/ABSOLUTE_SUMS/Graph.lean:113–148 controls graph sums by a rooted-tree organization. YM/Push/ABSOLUTE_SUMS/Log.lean:114–149 proves the logarithm/exponential identity. YM/Push/ABSOLUTE_SUMS/Clusters.lean:224–243 supplies the geometric connected-cluster tail. Bry §3, pp.138–143, Theorem 3.1 and Corollary 3.2, is the direct older tree-control reference; Cam §3, pp.525–528, covers polymer convergence. B-RG-II §2, pp.11–22, and N2 §5.4, p.49, both use this general constructive machinery.

Classification. STANDARD/ADAPTED-OTHER about 85%, application-specific support format 15%; high confidence. Counting the whole unit as exclusively Balaban or exclusively NLSM would confuse use of a method with its origin. The special exact inventory algebra belongs to the surrounding representation, not to the tree inequality itself.

BY90 pp.354–355 Eqs.(1.7)–(1.9) gives the disjoint-set algebra and finite polymer exponential/logarithm; pp.364–366 Eqs.(2.1)–(2.10) gives relevant-term extraction and connected reblocking. GK85 p.218 Eq.(5.28) gives the connected-graph exponentiation in a hybrid representation. These direct older formulas reinforce the STANDARD/ADAPTED-OTHER classification; BY's algebraic logarithm should not be conflated with an ordinary convergent logarithm on every field configuration.

11. Stationary nonlinear background with a uniform radius — III.6

Function and size. 19,698 lines. A critical branch for the actual classical action is constructed in gauge-fixed coordinates. The subsequent estimates make its analytic radius and inverse-Hessian control usable uniformly in volume.

Evidence. YM/PCalc/Chart.lean:78–100 defines the tested action, critical branch and Hessian condition. YM/Work/F04.lean:62–86 constructs the local branch by an implicit-function argument. YM/Push/STATIONARY_UNIFORM/ImplicitRadius.lean:24–43, YM/Push/STATIONARY_UNIFORM/ImplicitRadius.lean:80–92 prices an inverse without a spurious volume factor from changing norms. B-RG-II §1, pp.2–3, Eqs.(1.2)–(1.5), recalls the nonlinear background equations and their propagator dependence. B-Var Theorem 1 p.279, Proposition 6 pp.295–296 and Proposition 9 p.309 directly supply the gauge antecedent: a unique small critical orbit, an analytic stationary solution, and regularity/localized dependence on the background. Equation (190), p.308, states derivative decay, although its preceding detailed derivative calculation is omitted there. N2 §5.2, pp.42–44, uses a localized affine Gaussian shift with an explicit residual; this is a related but weaker role than solving nonlinear stationarity.

Classification. Predominantly ADAPTED-BALABAN with STANDARD analytic tools and a specialized soft-kernel implementation; high confidence in the direct local-branch and regularity antecedent. Generic analytic stationary branches and background regularity are established older mechanisms. The branch is for a soft penalty on tree-chart variables, not literally Balaban's constrained minimizer on nested irregular domains. The implicit-function theorem alone is routine; uniformity in the appropriate norms is the substantive work.

12. Matching the Wilson energy through fourth order — III.7 (a)

Function and size. Within III.7's 19,927 lines. The chosen coarse action reproduces the relevant low Taylor coefficients and separates the marginal curvature-square direction from irrelevant terms. This is where gauge symmetry and the explicit cubic/quartic structures matter.

Evidence. YM/PCalc/Chart.lean:104–109 defines SameJet and ClassicalMatch through order four. YM/RG/PathActions.lean:236–242 retains principal, cubic, quartic and loop terms. YM/Next/STATIONARY/Level/Transition.lean:86–93 packages matching with the quantitative transition. B-RG-I §4, pp.281–291, differentiates gauge invariance through fourth order; §5, pp.292–297, extracts the marginal transverse Maxwell form. N2 Eq.(5.7), p.40, and Eq.(5.40), p.54, impose the later activity-class affine-Hessian subtraction.

Classification. ADAPTED-BALABAN 55%, STANDARD gauge power counting 25%, ADAPTED-NLSM format 20%; high confidence. Zero affine Hessian is not a distinctive NLSM invention: the gauge-specific subtraction has a direct Balaban antecedent. The exact path-based retained format still requires a separate preservation proof in Lean.

B-Var §G pp.305–307, Eqs.(174)–(180), expands the stationary background around its linear lift; B-UV3 p.261 Eq.(19) separates quadratic and higher action terms. These strengthen the background-Taylor antecedent. Neither alone identifies the four-dimensional marginal coefficient: B-RG-I §§4–5 remains the direct gauge symmetry/marginal comparator for YM/Next/STATIONARY/Level/Transition.lean:86–93.

13. Numerical one-loop slope and constructive matching — III.7 (b)

Function and size. Also within III.7. A vector/ghost calculation identifies the universal logarithmic coefficient, and a bounded-error comparison attaches it to the actual blocked action. These are different obligations: computing a continuum loop integral does not by itself identify the exact RG update.

Evidence. YM/Next/STATIONARY/OneLoop/Bubbles.lean:14–26, YM/Next/STATIONARY/OneLoop/Bubbles.lean:48–61, YM/Next/STATIONARY/OneLoop/Bubbles.lean:125–141 derives color traces and the vector/ghost integrand. YM/Next/STATIONARY/OneLoop/MomentumIntegral.lean:326–338, YM/Next/STATIONARY/OneLoop/MomentumIntegral.lean:346–408 integrates the pole and controls the remaining terms. YM/Next/STATIONARY/FullPlanned.lean:34–52 connects the action data to the auxiliary logarithmic estimate. YM/Next/STATIONARY/Level/Transition.lean:153–178 then uses pole_reduced to transfer that estimate to the stationary coefficient; this is a separate comparison. B-RG-I p.297, Eqs.(5.37)–(5.42), extracts the marginal coefficient, but Theorem 2, p.259, postpones the promised running-coupling proof; B-Large-II p.355 retains the assumption. Abbott82 §3.3, p.48, Eqs.(3.3.1)–(3.3.5), directly displays the older vector/ghost coefficient; Gross–Wilczek and Politzer 1973 supply its historical asymptotic-freedom setting. N2's different spin coefficients in §6, pp.56–61, are not the numerical input for Yang–Mills.

Classification. STANDARD/ADAPTED-OTHER perturbation theory 45%, ADAPTED-BALABAN extraction 30%, candidate-NEW controlled matching 25%; medium-high confidence. The usual beta-function coefficient \(b_0 = 11 C_A / (48 \pi^2)\) is not new physics. Lean's \(\kappa_G = 2 b_0 = 11 C_A / (24 \pi^2)\) is the corresponding slope for the inverse-square coupling; these conventions must not be confused. Completing its uniform linkage to this particular exact transformation is a substantial constructive obligation; a complete mathematical verification of that linkage remains outside the scope of this comparison.

MRS93 p.376 also describes one-/two-loop flow calculations with next-order remainder control and higher-order truncation of gauge changes; §VIII pp.376–378 ties coupling relations to Slavnov identities. This is an older four-dimensional gauge marginal-control programme. It does not identify the matrices or normalization of the particular Lean transfer at YM/Next/STATIONARY/Level/Transition.lean:153–178; that bridge remains a separate technical question.

14. Connected activities, compulsory inventories and complete supports — III.8

Function and size. 4,103 lines. Only empty-inventory connected components may be exponentiated into regular errors. Seeded components retain mandatory coverage, and dependence supports include negative status queries and all geometric choices.

Evidence. YM/RG/Activities.lean:42–55, YM/RG/Activities.lean:77–96, YM/RG/Activities.lean:113–135 records inventory, eligibility, whole-label transport and exact coverage. YM/SectorNext/VACUUM_NORMALIZATION/Gas.lean:213–240 gives the unforced catalog and the exact deleted-support gas correction by omitted logarithmic pieces, supplying the reassembly evidence beyond those definitions. N2 Eqs.(5.7)–(5.10), pp.40–41, §5.4, p.49, and Appendix C have the closest full representation and reassembly discipline. B-RG-II pp.2–4, Eqs.(1.2)–(1.8), and B-N-Loc-I §§2–4 already localize nonlinear background dependence. B-N-II p.676 explicitly demands that local outputs depend only on local inputs.

Classification. Mixed ADAPTED-OTHER (GK), ADAPTED-BALABAN and ADAPTED-NLSM, with STANDARD support logic. High confidence in the specific older covering device; medium in N2 as the closest complete format. GK85 p.212 Eq.(4.8) already sums disjoint supports covering the bad region, each meeting it in a nonempty union of components. Its pp.218–221, especially p.219 Eqs.(5.34)–(5.39), groups components touching that region and exponentiates the outside sum. Mandatory coverage and this selective reassembly therefore cannot independently distinguish N2. Lean adds explicit inventory labels, eligibility predicates and exact read supports; the complete clipped/masked/tagged format remains closer to N2. Complete read supports are an exact independence requirement, not a new principle of locality. Gauge routes, frames and windows enlarge those supports beyond spin-field arguments.

15. Exact coupling reset and restoration of the activity class — III.9

Function and size. 7,162 lines. The marginal and vacuum pieces are absorbed into new scalar parameters, and all remaining corrections are redistributed into the admissible regular/forced classes. Exact density equality is stronger than matching two test values.

Evidence. YM/UVNext/COUPLING_RESET/Advance.lean:47–51, YM/UVNext/COUPLING_RESET/Advance.lean:78–103 separates scalar parameter matching from ShapeControl. YM/AnalyticBridge/VERIFIED_RESET_STEP/Evaluation.lean:17–29, YM/AnalyticBridge/VERIFIED_RESET_STEP/Evaluation.lean:71–95 proves equality of reset and pre-reset densities. N2 pp.54–55, Eqs.(5.40)–(5.42), removes constant/affine terms and sends mask-complement corrections into the compulsory gas. B-RG-I pp.258–259, Eqs.(0.28)–(0.30), and B-N-II §§4–5, pp.698–718, supply older marginal extraction and renormalization conditions.

Classification. Mixed older constructive RG and ADAPTED-NLSM for the particular mask-complement reset; high confidence in exact-format matches, medium in closest-source ranking. GK85 pp.218–221, Eqs.(5.32)–(5.53), already removes Taylor terms, reassembles the bad-region density and restores the original interaction class with new coupling and wave-function normalization. BY90 pp.360–361 Eqs.(1.17)–(1.23) and pp.364–366 Eqs.(2.1)–(2.10) gives another exact extraction/reblocking identity. Exact density restoration itself is older; the N2 affinity is the specific redistribution of clipped/masked action terms into the regular and compulsory classes. YM/RG/History.lean:25–33 explicitly treats this as an action-representation coordinate rather than an independently defined observable coupling.

16. Energy per seed and protection of rough sectors — III.10 (a)

Function and size. Within III.10's 49,766 lines. Energy reserved for primary rough seeds pays entropy, normalization and derivative costs without repeatedly spending the same reserve. Protecting complete neighborhoods prevents local expansions from erasing the predicates that selected the sector.

Evidence. YM/Pressure/Seeds.lean:22–58, YM/Pressure/Seeds.lean:124–131 records three inventories and the energy lower bound per share. YM/Next/FLUCTUATIONS/Grants.lean:24–36 supplies those bounds and controls their costs in the stage order. N2 Lemma 5.2, pp.43–44, p.45 and Appendix C use primary cores, protector supports and energy reserves in the same retained-density context. B-Conv §2, pp.254–262, and B-Large-I pp.175–177 give the older need to keep and repeatedly control large-field regions.

Classification. ADAPTED-NLSM 45%, ADAPTED-BALABAN 30%, specialized adaptation 25%; medium-high confidence. YM seeds include bad plaquettes and incompatible coarse bonds, not just large spin gradients. The super-polylogarithmic monitor and gauge chart costs change the quantitative balance. Balaban's separate aged-region R replacement was not identified as a literal transformation in these interfaces.

MRS93 p.376 supplies another older protection mechanism: asymmetric corridors toward lower frequencies distribute a large-field saving over additional boxes and compensate high/low coupling costs. GK85 p.222 Eq.(6.5) retains bad-region energy and entropy in the inductive bound. These strengthen older ancestry of protected rare-region reserves; the Lean inventory/support geometry and its particular reserve schedule remain distinct.

17. Nonlinear contraction and bounds on irrelevant terms — III.10 (b)

Function and size. Also within III.10. The exact step is reorganized so that regular errors contract and forced terms remain rare while all derivative/support costs fit their budgets. This closes the retained density class beyond a formal perturbation series.

Evidence. YM/RG08/Steam/Routing.lean:14–37, YM/RG08/Steam/Routing.lean:85–91 makes the correction order triangular in coupling degree and field rank. YM/RG/Activities.lean:98–111 states the regular/forced bounds that closure must preserve. N2 Theorem 5.1, pp.41–42, §§5.3–5.6 and Eq.(5.43), p.55, gives the closest exact-step closure architecture. N3 §5, pp.39–55, strengthens contractions with reflection/inversion cancellations. B-RG-I p.258, Eqs.(0.28)–(0.30), and B-RG-II §2 provide older summable irrelevant remainders and localization/exponentiation.

Classification. ADAPTED-NLSM 45%, ADAPTED-BALABAN 25%, STANDARD estimates 15%, candidate-NEW uniform extension 15%; medium confidence. The same representation does not make the gauge estimates automatic: cubic vertices, gauge degeneracy, color and four-dimensional marginality must be handled. The inspected interfaces support architectural attribution; they do not constitute a line-by-line proof of every closure inequality.

GK85 §§6,9 pp.221–222,233–237 and BY90 Theorems A/B pp.360–363 already close local analytic activity classes after marginal extraction and bound regulator losses. MRS93 §§VI–VII pp.368–376 treats gauge background determinants, protected large fields and coupling-flow remainders. Thus closing a gauge RG with a marginal direction is not a defensible broad novelty claim; only the controlled realization of this particular retained transformation remains a candidate, with medium-low novelty confidence.

18. Volume-uniform normalization and cancellation of gauge powers — III.11 (a)

Function and size. Within III.11's 42,556 lines. A Haar-shell lower bound and triangular plaquette upper bound control normalized insertions uniformly in torus size. The key normalization cancels extensive powers of the small coupling before a chessboard root is taken. The unrooted estimate still has an exponential volume cost; the uniform statement is the resulting per-cell/root price.

Evidence. YM/Next/RANK/Determinant.lean:76–111, YM/Next/RANK/Determinant.lean:189–217, YM/Next/RANK/Determinant.lean:240–274 uses exponents \(3 V + 1\) and \(3 V - 3\), leaving a global \(g^{- 4 \dim G}\) factor. YM/Next/RANK/StationMeter.lean:8–14, YM/Next/RANK/StationMeter.lean:28–44 states the per-cell price without an extensive logarithmic loss. YM/Next/RANK/Stamp.lean:131–167 converts the ratio to the root bound. B-Large-II pp.356–359, Eqs.(1.1)–(1.10), has determinant-sensitive local normalization ratios, an older general task. N3 Proposition 7.1 and Lemmas 7.3–7.4, pp.61–65, instead compare through rough-region fillings against a positive reference law.

Classification. STANDARD integration with substantial ADAPTED-BALABAN/ADAPTED-OTHER normalization antecedents; high confidence in the older general task, low confidence in novelty of the particular torus estimate. No specific NLSM adaptation is established for this cancellation proof. The direct source check rules out assigning all of III.11 to N3 merely because both contain integrated comparison estimates.

MRS93 p.368 Eq.(VI.1) combines the gauge/ghost determinant ratio, counterterms and background action; Lemma VI.1 p.369, Eq.(VI.5), bounds the combined large-field dressing factor by one for its stabilizing cutoff. It explicitly explains why a cruder per-volume bound would not suffice. This substantially strengthens ADAPTED-OTHER antecedents for the normalization problem. Seiler82 pp.118–121 instead treats matter determinants in a prescribed regular external gauge field; the four-dimensional statement is explicitly Quasi-Theorem 6.10, not a pure-gauge measure result.

The exact \(3 V + 1\) versus \(3 V - 3\) sweep/root estimate is unmatched in the targeted passages, but remains a low-confidence novelty candidate. The first count is ordinary spanning-tree arithmetic; possible substantive novelty lies in the uniform numerator/lower-bound estimates and their use for composite insertions.

19. Interacting/free rank and finite detector comparison — III.11 (b)

Function and size. Also within III.11. Finite test systems compare interacting composite labels with the free Maxwell quotient, control equation terms and stabilize the field-space rank. Good-scale estimates permit corrected trial systems and coherent choices of coordinates.

Evidence. YM/MergeProof/Powder.lean:19–71 combines equation-count savings with Maxwell rank data. YM/SharpGram/Weak/Compatible.lean:140–159 gives compatible changes of coordinates across degrees. N3 §9, pp.74–82, normalizes spin sources, and §11.3, pp.93–94, compares a specific spin cumulant; neither is the all-composite Maxwell rank argument. B-RG-I §4, pp.281–291, controls gauge Taylor constraints, without the continuum label-identification endpoint displayed here.

Classification. STANDARD finite-dimensional stability 45%, ADAPTED-OTHER comparison strategy 15%, candidate-NEW simultaneous detector/rank construction 40%; medium confidence. The potentially new content is the set of uniform inputs and their coherent use, not the elementary fact that invertibility survives small perturbations. A deeper comparison with constructive composite-operator literature is a priority before claiming originality across the literature.

MRS93 §VIII pp.377–378 supplies an arbitrary-order gauge-fixed elementary-source hierarchy, but p.329 explicitly leaves construction of renormalized physical composite operators outside the paper. It therefore strengthens the old source-analysis context without supplying the Maxwell equation-quotient and simultaneous finite-detector comparison at YM/MergeProof/Powder.lean:19–71. The latter remains a bounded candidate; neither all insertions nor rank stability alone is new.

20. Reflection positivity, chessboard roots and derivative budgets — III.12 (a)

Function and size. Within III.12's 18,944 lines. Reflection positivity yields a seminorm controlling disseminated observables. Gauge-invariant local fields and their finite differences receive volume-uniform bounds suitable for continuum extraction.

Evidence. YM/Cells/Certificate.lean:14–24 packages reflected positivity, product dissemination and the root triangle inequality. YM/Extraction/Center.lean:21–23, YM/Extraction/Center.lean:38–40 states the centered observable budget. FILS78 §4, preprint pp.33–36, Theorems 4.1–4.3, is the direct older source for abstract and torus chessboard estimates. N3 Lemma 7.3, p.62, uses chessboard exponential moments for kinetic rows; N2 §§2,9 uses reflection positivity for a distinct partition-doubling mass-gap argument.

Classification. STANDARD/ADAPTED-OTHER 65%, ADAPTED-NLSM source-estimate affinity 15%, candidate-NEW all-label derivative application 20%; high confidence in the base inequality, medium in the adaptation. The same positivity principle has different downstream uses. Here a root bound and derivative control must not be relabelled a full transfer-gap estimate.

Seiler82 §2a pp.14–16 directly proves lattice-gauge reflection positivity (Theorem 2.1), the quotient Hilbert-space construction and transfer contraction, then states the chessboard inequality (Theorem 2.2). Its displayed chessboard formulation uses local matter functions and refers the gauge-field extension to another source. This directly strengthens the older lattice-gauge setting; the general chessboard comparison also rests on FILS. Osterwalder–Seiler 1978 remains unavailable in full.

21. Fixed-volume convergence and analytic tuning — III.12 (b)

Function and size. Also within III.12. Fixed-level transfer data give Cauchy correlations at a fixed physical period, first on convenient probes and then generally. Analytic coupling families and a common successful recursion constrain tuning without asserting uniqueness of all possible continuum theories.

Evidence. YM/FullVolume/Gables/Vendee.lean:44–61 gives the run and coefficient convergence. YM/FullVolume/Fields/Capital.lean:55–61 states the Cauchy/confluence step. YM/FullVolume/Tuned.lean:58–72 passes from dense probes and equicontinuity to the general limit. YM/FullVolume/Finale/Program.lean:60–77, YM/FullVolume/Finale/Program.lean:133–158 records the successful-recursion and analytic-family scope. N3 Proposition 10.2, pp.84–86, likewise obtains fixed-torus source convergence from coarse-level comparisons. N3 §§12–15, pp.95–120, goes further by removing bare-offset ambiguity with additional comparisons.

Classification. STANDARD limit/analytic arguments 45%, ADAPTED-NLSM construction affinity 30%, specialized extension 25%; medium confidence. Lean's fixed-volume full-mesh convergence does not imply a unique infinite-volume limit. The N3 inclined-block and incommensurable-period mechanism has not been identified as the tuning mechanism here.

MRS93 pp.327–330 already states a fixed-IR ultraviolet Schwinger-function limit and compares tentative/exact coupling schedules. BY90 p.364 proves analyticity under RG iteration and of the pressure. These strengthen the older construction/analyticity context, without supplying the Lean tuned composite full-mesh Cauchy statement. Seiler82 pp.100–101 distinguishes existence of scaling limits from independence of the chosen physical normalization; neither fixed-volume convergence nor analytic tuning alone establishes universality.

22. Extension to all labels and all finite orders — III.13

Function and size. 22,077 lines. One family must serve every finite degree and requested derivative order, with compatible labels, budgets and comparisons. The construction is selected once rather than separately for each later observable.

Evidence. YM/SharpGram/Assembly/Flowers.lean:75–130, uv_log, constructs the common family while stages increase their orders. YM/Source/Construction.lean:24–62 makes the degree factory, stages and normalized construction explicit. N3 §9.1–9.2, pp.75–80, already extends the density flow to local spin sources and proves fixed-layer convergence. B-N-Large §4 and B-N-Corr p.493 establish that generating functionals and correlations are part of Balaban's broader programme. Neither cited source supplies this exact common gauge-composite hierarchy with the later OPE/stress requirements.

Classification. STANDARD compatibility/diagonal organization 35%, ADAPTED-NLSM source machinery 25%, candidate-NEW all-label extension 40%; medium confidence. The enlarged gauge-composite spaces, growing derivative/degree requirements and simultaneous OPE/stress comparisons are substantive. The order “choose one trajectory, then all spin moments” already occurs in N3 Theorem 1.2, p.6, and §9.1, pp.75–76; that quantifier order alone is not new. This is a leading location for new constructive work whose detailed priority still needs a wider literature comparison.

MRS93 p.329 explicitly excludes renormalized physical composite-operator construction, while its §VIII pp.377–378 does treat arbitrary elementary field insertions. BY90 p.359 Eq.(1.13) already controls all variational derivative orders in an analytic norm. The surviving candidate is therefore the simultaneous gauge-composite quotient/rank/stress/OPE construction, not “all insertions,” “all derivatives,” or one common trajectory in isolation.

23. Iteration and selection of one construction with uniform bounds — III.14

Function and size. 14,239 lines. Local estimates, terminal shooting and increasing stages are assembled into a single sequence of lattice laws. That sequence is fixed before the later extraction order and comparison requests.

Evidence. YM/Independent/ConstructionRun.lean:14–31 fixes the run and its uniform bounds before observable demands. YM/Source/Construction.lean:65–72 preserves the normalization and translations without retuning. N2 §7, pp.61–65, and N3 §6, pp.55–60, provide the closest terminal-coupling trajectory organization. B-RG-I Theorem 2, p.259, expresses the older tuned-coupling objective; B-N-II Theorem 1, p.680, provides a complete iterative framework in a different model.

Classification. ADAPTED-NLSM 35%, ADAPTED-BALABAN 20%, STANDARD assembly 35%, specialized extension 10%; medium confidence. A long formal iteration proof is not a corresponding quantity of new physical ideas. Serving all spin moments with one trajectory is already present in N3; the added obligations here are the larger gauge-composite and derivative comparison systems.

24. Activity costs, local observable derivatives and analytic trace estimates — III.15–III.16

Function and size. III.15: 7,366 lines; III.16: 3,931 lines. Activity-fusion estimates bound hard and soft contributions; local observable estimates control derivatives of individual insertions. Analytic trace-product estimates supply bounds needed later in the continuum construction. These are cost and derivative tools; the density-fusion theorem alone is not an external-source propagation theorem.

Evidence. YM/Progress/DIRECTED_RECIPE/Tender.lean:32–62, YM/Progress/DIRECTED_RECIPE/Tender.lean:124–133 distinguishes hard/soft costs and restores the regular/forced activity budget. YM/Next/SCHEDULE/SlotPay.lean:108–159 bounds a transported local slot; YM/Next/SCHEDULE/TensorCost.lean:52–74, YM/Next/SCHEDULE/TensorCost.lean:85–89 gives the derivative bound for a k-leg atom, including its threshold power. YM/Complex/Gaps.lean:170–176, YM/Complex/Gaps.lean:206–210 bounds trace products using analytic interpolation and a Cauchy estimate in the gap. N2 §5.4, p.49, supplies a close hard/soft reassembly setting; its Lemma 5.3, pp.47–48, bounds inserted factors. N3 Proposition 11.3, pp.90–91, likewise needs analytic growth bounds for reconstruction, without this exact trace estimate. Actual source transfer is a separate interface, YM/Source/Insertion.lean:26–68, YM/Source/Insertion.lean:89–110, discussed in §4.3.

Classification. III.15: STANDARD derivative/fusion estimates 60%, ADAPTED-NLSM reassembly affinity 20%, specialized gauge cost bounds 20%. III.16: STANDARD analytic tools about 90%, specialized packaging 10%; medium-high confidence. Cauchy's inequality, product derivative bounds and positive-operator trace estimates should not be credited as new NLSM or Yang–Mills ideas. The construction-specific work is obtaining the correct dimension- and volume-uniform hypotheses for those tools.

GK85 pp.218–221 supplies an older selective reassembly setting, and BY90 pp.359–363 already tracks derivative norms through Gaussian integration. MRS93 §VIII pp.377–378 gives elementary source identities, with physical composite renormalization explicitly excluded at p.329. These comparisons reinforce the distinction between standard derivative/source operations and the Lean-specific quantitative bounds for all gauge composites.

25. IV.1 — Infinite-volume extraction and Euclidean hierarchy

Size: 7,336 lines in spans of declarations used by the main theorem.

What it does. It turns volume-uniform observable bounds into a common limit table, retains fixed-volume mesh limits, extracts volumes, and passes separated smeared correlations to that table. The distinction between generic point limits and integrated limits is represented explicitly.

Lean evidence. YM/Extraction/Lipschitz.lean:156–160 defines RawLimits with continuity, mesh, volume and integrated clauses. YM/Extraction/Held.lean:135–137, crops, obtains a common extraction from the projected chessboard estimates. YM/Extraction/Limits.lean:57–66 distinguishes uniform fixed-period mesh convergence from volume convergence on compact generic configurations. YM/Extraction/Integrate.lean:115–153, ordinary, applies dominated convergence after a common compact-support bound. The proof explicitly invokes dominated convergence at YM/Extraction/Integrate.lean:152.

Closest antecedents. N3 Proposition 10.2, pp.84–86, first establishes fixed-torus source convergence, then approximates continuous tests and exchanges the cutoff and volume limits using equation (10.8). The abstract compactness, diagonal extraction and dominated-convergence steps are STANDARD. The upstream use of exact renormalized local sources to obtain a common limiting hierarchy is the same general method as N3 §§9–10. The exact infinite-volume mechanism differs: N3 uses uniform exponentially small finite-volume errors from a transfer-gap construction; the Yang–Mills public target permits subsequences.

Classification: STANDARD + ADAPTED-NLSM at the observable-construction interface; medium confidence on immediate lineage, high on the difference in conclusions. A claim that infinite volume itself is a direct transplant of N3 would be too strong.

Rough mechanism split: STANDARD 75%, NLSM source-construction affinity 15%, specialized extension 10%.

26. V.1 — Phase decomposition

Size: 5,718 lines in spans of declarations used by the main theorem.

What it does. Reflection positivity gives a Hilbert space and positive-time contraction semigroup. Compressed whole-word observables on its zero-energy space generate an algebra; proving that algebra commutative permits a compact character-space mixture of phases.

Lean evidence. YM/OS/Dynamics.lean:63–107 constructs the semigroup and transverse unitary action on the reflection Hilbert space. YM/Ground/Permute.lean:73–93, exchanges, interchanges the middle two entries in a four-entry train by using two coordinate directions and a signed rotation. YM/Ground/Permute.lean:95–136, gens, turns that identity into commutation of the compressed operators. YM/Ground/Permute.lean:149–161, adjunction, extends commutation to their closed star algebra. YM/Phase/Mixture.lean:65–86 specifies the compact character space, probability measure, full support and representation of the ground-state functional. YM/Phase/Mixture.lean:113–117, barycenter, gives the correlation mixture identity.

Closest antecedents. The first step is OS73 §4(a–b), p.90: positive form, quotient/completion, self-adjoint contraction semigroup. Gelfand representation and a positive functional represented by a measure are STANDARD functional analysis; they are not distinctively Balaban or NLSM. The inspected N3 proof does not make this phase decomposition: §11.2, p.91, obtains vacuum uniqueness from already-established exponential clustering. Its target has a single canonical periodic state and a positive gap (Theorems 1.1–1.2).

Classification: STANDARD core with candidate-NEW implementation of the train/axis exchange and phase-uniform machinery; medium confidence for novelty relative to these sources, low for priority across the literature. There is no evidence here that the phase-algebra argument is inherited from N3. Before a strong novelty claim, compare the extremal-state and asymptotic-abelian constructions in the older Euclidean/algebraic-QFT literature.

Rough mechanism split: STANDARD 70%, candidate-NEW phase-uniform organization 30%.

27. V.2 — Reconstruction

Size: 6,419 lines in spans of declarations used by the main theorem.

What it does. It builds analytic continuations and boundary distributions from Euclidean data, then realizes fields on a Hilbert space with locality, spectrum and covariance properties. The reconstruction theorems have explicit hypotheses; the assembly discharges them for the chosen hierarchy.

Lean evidence. YM/Hilbert/Kolmogorov.lean:67–92, host, completes a positive kernel and proves density of the generating vectors. YM/Work/C01.lean:99–117 derives a common controlled field-vector family from slab bounds. YM/Integrated.lean:182–212, reconstructionArcs, assembles A16 and C01–C29. ChallengeSpec.lean:135–143 makes analytic continuation, its Euclidean root, and distributional boundary convergence explicit. PublicTheory.lean:270–319 lists the resulting spectrum, unique-vacuum, domain, covariance and locality clauses.

Closest antecedents. OS73 §4 p.90 gives the null quotient, Hilbert completion, contraction semigroup and boundary-value route. Its originally announced sufficiency theorem cannot be cited without the correction: OS75 p.282 identifies OS73 Lemma 8.8 as false. OS75 §III p.285 uses a stronger distribution topology; §IV.1 p.287 Eq.(4.1) gives the useful sufficient linear-growth condition, with one fixed seminorm slope and factorial bounds in field order. Its §IV.2 pp.288–289 separates analytic continuation from the stronger estimates needed for tempered boundary values. N3 Proposition 11.3, pp.90–91, applies the corrected theorem, checking a factorial/linear growth estimate. That NLSM application is not the origin of reconstruction: it is another use of the same older theorem.

Classification: predominantly STANDARD/ADAPTED-OTHER (Osterwalder–Schrader), high confidence. The technical adaptation is to a graded gauge-invariant field algebra with all finite-degree labels and to phase-dependent tables, rather than a three-component scalar spin field. The cited passages do not establish equivalence of the graded Lean growth/topology hypotheses with OS75 Eq.(4.1). YM/Work/C01.lean:99–117 supplies a controlled vector family, not that displayed joint Schwartz bound by itself. The remaining issue is a separate theorem/domain comparison, not lack of access. A direct reconstruction may prove boundary estimates by another route; no gap follows from this unresolved correspondence.

Rough mechanism split: STANDARD/ADAPTED-OTHER 95%, specialized implementation 5%.

28. V.3 — Reconstruction for every phase

Size: 4,063 lines in spans of declarations used by the main theorem.

What it does. The same field labels, stress data and rotation action are carried into each character phase. Uniform estimates establish the zero-energy projection and decay after subtracting it, so vacuum uniqueness is obtained phase by phase without asserting a mass gap.

Lean evidence. YM/EveryPhase/Pack.lean:117–132 defines an Almanac containing every chamber, field theory, label identification and attachment to the common data. YM/EveryPhase/Extension.lean:75–80, all_characters, assembles the completed family. YM/EveryPhase/Vacuum.lean:115–144, all_plant, identifies the zero-energy spectral projection with the vacuum rank-one operator on a dense family and extends it by continuity. YM/EveryPhase/Decay.lean:173–192, inverse_square, bounds the residual semigroup matrix element by \(Ms^{- 2}\), uniformly in phase. The proof uses a low-energy spectral-mass bound \(\nu ((0, x]) \leqslant Dx^2\) at YM/EveryPhase/Decay.lean:182–186. That input is stated explicitly as YM/EveryPhase/Light.lean:221–232, light_bound; its proof uses locality at large spatial separation and the spectral annulus estimate.

Closest antecedents. N3 Proposition 11.4, pp.91–93, also uses positive spectral measures and a dense set of field-generated vectors. But there the semigroup matrix elements are exponentially bounded and hence a whole interval of low energies is excluded. Here the low-energy mass estimate implies polynomial decay after subtraction of the full zero-energy projection, and does not exclude arbitrarily small positive energies.

Classification: STANDARD spectral reasoning + candidate-NEW/ADAPTED-OTHER phase-uniform organization; medium confidence. This is a comparison at the level of a general method with N3, with a materially different estimate and endpoint. It supplies counterevidence to the idea that the Yang–Mills continuum construction is simply the NLSM mass-gap proof with another target manifold.

Rough mechanism split: STANDARD spectral tools 65%, candidate-NEW phase-uniform organization 35%.

OS75 p.283, Remark 5, already permits extension to denumerably many spinor fields; going beyond one scalar is not a new reconstruction principle. OS73 p.89 also distinguishes exponential clustering and a mass gap. The potential contribution here is the particular phase-uniform input estimates, notably YM/EveryPhase/Light.lean:221–232, not reconstruction applied to several fields or a family by itself. Confidence in that specific corpus-relative extension remains medium.

29. VI.1 (a) — Ward identities and full rotation covariance

Size: within VI.1's 7,689 lines in spans of declarations used by the main theorem.

What it does. Local stress-shell identities produce translations and rotation generators on finite-degree fields. The proof upgrades collinear identities to all separated configurations and supplies a genuine compact rotation action compatible with the field filtration.

Lean evidence. YM/Euclid/Stress.lean:88–97 defines stress laws on separated configurations; it explicitly says contact terms are not tested. YM/Next/ROTATIONS/Route.lean:114–123 proves the collinear angular sum and a line-sum identity. YM/Ward/Walk.lean:15–26, YM/Ward/Walk.lean:60–84 supplies analytic continuation along segments and walks. YM/Ward/Axis.lean:86–107, YM/Ward/Axis.lean:130–139 connects arranged configurations to collinear ones and then completes the argument by density and continuity. YM/SharpGram/Weak/Compatible.lean:140–159, tune_grades, constructs changes of coordinates compatible across all degrees up to a fixed horizon and preserves units/reality/signed symmetries.

Closest antecedents. Local current insertions generating infinitesimal symmetry are STANDARD Ward/Noether reasoning. N3 §§10.2–10.3, pp.84–87, are an especially useful negative comparator: rotations there follow from comparing tilted regulators and a dense cyclic subgroup, not from constructing a local spacetime stress tensor. A direct NLSM attribution for the Lean rotation argument is unsupported by this source pair.

Classification: STANDARD Ward principle + candidate-NEW implementation beyond the inspected NLSM and Balaban UV outputs; medium confidence. Claiming originality across the literature of the analytic-vector and finite-dimensional representation steps would be unjustified; those are established tools.

Rough mechanism split: STANDARD Ward/representation tools 45%, candidate-NEW constructive implementation 55%.

30. VI.1 (b) — Exterior flux at infinity

Size: crosses units; the exterior-flux proof occupies YM/Next/ROTATIONS/Yard.lean:1–115 (115 lines) and YM/Next/ROTATIONS/Route.lean:1–124 (124 lines), with upstream estimates for coordinate directions.

What it does. It proves that the exterior terms left by large rotation cutoffs vanish, allowing a local shell identity to become a global infinitesimal rotation identity without an assumed exponential clustering estimate.

Lean evidence. YM/Next/ROTATIONS/Yard.lean:49–77, pick_price, constructs reflection Hilbert chambers in every coordinate direction, splits signs, uses exterior_size, and obtains a decay bound with exponent \(- 1\) for the actual stress integrand. The realizations in all coordinate directions and the upstream spectral estimate appear at YM/Next/ROTATIONS/Yard.lean:54, YM/Next/ROTATIONS/Yard.lean:67–70. The input S.Hierarchy in YM/Euclid/Words.lean:140–148 contains positivity, translation invariance, smoothness and bounds, with no clustering or mass-gap clause. At YM/Next/ROTATIONS/Yard.lean:73–77, a factor with budget 100 combines with budget −101 to leave budget −1. YM/Next/ROTATIONS/Yard.lean:105–114, blows, transfers this to the concrete correlation function for collinear insertions. YM/Next/ROTATIONS/Route.lean:49–67, drained and axisTurn, sum the exterior terms and conclude that a constant with a decaying budget is zero.

Closest antecedent. N3 §11.1–11.2, pp.89–93, uses exponential clustering and positive-energy spectral reconstruction, but does not contain this all-axis exterior-flux device. General reflection positivity and spectral Cauchy–Schwarz are standard; the particular cutoff/sign construction is a candidate distinct contribution.

Classification: candidate-NEW, medium confidence within these sources, low on exhaustive priority. The source reading covers the proof through its use of the exterior spectral estimate, but not every upstream moment lemma for coordinate directions. This supports the mechanism comparison, not a complete verification of the flux estimate.

Rough mechanism split: STANDARD spectral inequalities 35%, candidate-NEW exterior-flux organization 65%.

31. VI.2 — Maxwell reference theory and stress normalization

Size: 14,170 lines in spans of declarations used by the main theorem.

What it does. It supplies a free Maxwell hierarchy on the same graded label format, its conserved stress tensor and its shell Ward action. The free reference fixes the interacting stress and rotation structures through short-distance comparison.

Lean evidence. YM/Next/MAXWELL/System.lean:106–114 proves the hierarchy; YM/Next/MAXWELL/System.lean:153–170 defines the rotation action; YM/Next/MAXWELL/System.lean:172–179 defines and proves conservation of the free stress. YM/Next/MAXWELL/Current.lean:184–187 defines the shell flux as the integral against the cutoff gradient. YM/Next/MAXWELL/Force.lean:131–186, jet_force, extends the point-source flux identity to arbitrary jet probes by linearity and differentiation. YM/SharpGram/Weak/Compatible.lean:140–159 is a direct anchor for the coherent interacting/free label comparison used downstream.

Closest antecedents. Wick contractions and the Maxwell stress/Noether identity are textbook free-field calculations. N3 §11.3, pp.93–94, compares renormalized sources to a terminal spin law; it does not construct this Maxwell jet algebra or use it to fix spacetime stress. The two share the general tactic of identifying limiting operators by a controlled comparison model, not the same reference theory or normalization mechanism.

Classification: mostly STANDARD free-field mathematics, with candidate-NEW assembly/normalization in this nonperturbative construction; high confidence for the standard part, medium for the adaptation. The number of Lean lines devoted to Wick jets should not be interpreted as a comparable number of new physics ideas.

Rough mechanism split: STANDARD Maxwell mathematics 80%, specialized normalization 20%.

32. VII.1 — Short-distance freeness and finite detector systems

Size: 8,004 lines in spans of declarations used by the main theorem.

What it does. It identifies the leading short-distance correlations with Maxwell Wick correlations and chooses compatible finite-degree coordinates for the interacting field spaces. Gaussian convergence is used to distinguish labels and fix their symmetries, not merely to approximate the partition function.

Lean evidence. YM/Pert/Rig/Linked.lean:59–69, zerothProfile, identifies the zeroth perturbative coefficient with the smeared free Wick moment. YM/Pert/Warp/Approach.lean:19–32 proves the moment limit by recursively pairing contractions. YM/Pert/Warp/Horizons.lean:65–68 passes to polynomial rows. YM/SharpGram/Weak/Compatible.lean:140–159 gives coherent degree-wise changes of basis.

Closest antecedents. N3 §5.3, pp.42–45, uses ambient Gaussian completion for an exact RG step; its §11.3 compares source moments to a terminal spin law. Those are related Gaussian/comparison methods, but neither cited passage establishes the Maxwell composite-field short-distance statement. Asymptotic freedom and Gaussian leading short-distance behavior are older perturbative expectations: Gross–Wilczek 1973 and Politzer 1973, specified below.

Classification: STANDARD Wick/perturbative principle + ADAPTED-OTHER constructive implementation and candidate-NEW coherent detector use; medium confidence. A comparison of zerothProfile alone is insufficient to certify the full interacting asymptotic statement; the remainder estimates enter from the ultraviolet units.

Rough mechanism split: STANDARD/ADAPTED-OTHER 60%, candidate-NEW coherent detector use 40%.

33. VII.2 — Trace anomaly and the all-radii asymptotic

Size: 20,645 lines in spans of declarations used by the main theorem, including some nonfree exclusions.

What it does. The centered trace is tied to the physical stress tensor and its short-distance two-point behavior is normalized. An asymptotic first obtained along admissible scaling systems is promoted to every small positive radius.

Lean evidence. YM/TraceExtension/WilsonTrace.lean:48–49 defines the centered trace in the selected phase. YM/Nonperturbative/HeadlineSpec.lean:84–95 explicitly identifies its field operator with stress trace minus its vacuum expectation and states the short-distance law. YM/TraceExtension/Verse/Day.lean:41 defines the intermediate amplitude \(12 (d_2^2 / b_0^2) \dim \mathfrak{g}/ \pi^4\). YM/TraceExtension/Verse/Cape.lean:53–76, border, uses monotonicity of the positive spectral Laplace transform to compare nearby radii. YM/TraceExtension/Verse/Cape.lean:77–88, outside, argues by a bad sequence of radii and the ability to fit a subsequence to a scaling system. The public numerical conclusion is ChallengeSpec.lean:172–174: \(3 \dim \mathfrak{g}/ \pi^4\) after multiplication by \(r^8 \log^2 (1 / r)\).

Closest antecedents. CDJ77 §§II–IV pp.439–445 gives the renormalized normal-product/operator-mixing trace law, formulated in terms of beta functions and anomalous dimensions beyond a one-loop slogan. Its full Eq.(3.19), p.443, retains gauge-fixing, ghost and equation-of-motion operators; the simpler full-trace Eq.(3.21) is stated for on-shell nonzero-momentum matrix elements. Section IV extends the operator relation to arbitrary momentum. The beta-function trace mechanism is therefore a detailed older antecedent. It does not by itself yield a constructed positive Euclidean hierarchy or the all-radii two-point limit. Identifying coefficients requires fixing the coupling inside the curvature, stress sign and normal-product convention; no off-shell equality or numerical convention match is inferred merely from the familiar formula. N3 §§11.3 and 15.2–15.3 concern a fourth cumulant and canonical normalization, not a four-dimensional stress-trace ultraviolet law. Promoting controlled subsequences to a full limit is a standard logical argument; the hard construction-specific input is the uniformity allowing all prescribed radius sequences.

Classification: old trace-anomaly principle + candidate-NEW constructive realization and all-radii control; medium confidence. The correspondence is high-confidence for the older trace mechanism and the difference in stated endpoint, medium for the specific constructive novelty candidate. The one-loop coefficient and its matching to the full gauge-fixed determinant remain unaudited in full. Nor should the cancellation in the final amplitude be treated as a numerical cross-check of that coefficient.

Rough mechanism split: STANDARD/ADAPTED-OTHER anomaly and limit principles 40%, candidate-NEW constructive realization 60%.

34. VII.3 — Nonfreeness

Size: 3,303 lines in spans of declarations used by the main theorem for free detection/positivity; the terminal exclusion proof is largely included in VII.2.

What it does. The free reference has a nondegenerate reflected Gram form, and the trace singularity rules out finite combinations of ordinary free mass-shell Wick contractions. The public claim concerns a specified class of comparison profiles.

Lean evidence. ChallengeSpec.lean:151–166 defines the mass-shell mode and finite Wick-profile class. YM/TraceExtension/Wick/Log.lean:96–108, no_powerlog, proves that a profile with the specified finite power/log expansion cannot have the positive inverse-log asymptotic. YM/TraceExtension/Wick/Refute.lean:75–80 applies this to the Euclidean trace. YM/TraceExtension/Wick/Refute.lean:86–105 transfers the exclusion to smeared polynomial-field descriptions. The module's precise limitation is at YM/TraceExtension/Wick/Refute.lean:4–8: it makes no claim against a continuous-mass generalized free field.

Closest antecedent and difference. N3 Proposition 11.5, pp.93–94, proves non-Gaussian behavior by a nonzero separated connected four-point function. It computes the leading cumulant from the uniform sphere moments \(\mathbb{E} [(U^1)^2] = 1 / 3\), \(\mathbb{E} [(U^1)^4] = 1 / 5\). That is a different witness from the trace inverse logarithm; “nontriviality from a controlled nonfree observable” is only the shared general method.

Classification: candidate-NEW choice and rigorous realization of a nonfree witness; STANDARD asymptotic algebra once the trace law is known. High confidence on difference from N3; medium/low confidence on broad novelty because no exhaustive search for older trace-based nontriviality arguments was made.

Rough mechanism split: STANDARD asymptotic exclusion 50%, candidate-NEW constructive witness 50%.

35. VII.4 (a) — Separated OPE

Size: part of 10,193 lines in spans of declarations used by the main theorem.

What it does. The product of clustered insertions is expanded in finite-degree local fields with a remainder bounded by its next degree, with arbitrary small power slack. One coefficient system works for the mixture and every character phase, independently of the spectator word and truncation horizon. For each fixed spectator word the bounds are uniform in phase and allowed spectator positions; the eventual radius threshold may depend on that word, as quantified in ChallengeSpec.lean:74–76.

Lean evidence. YM/OPE/Orbits/Kernel.lean:4–25 states the intended exact scope and excludes convergence as the horizon tends to infinity. YM/OPE/Orbits/Kernel.lean:62–64 gives the coefficient and algebra-error bounds. YM/OPE/Orbits/Kernel.lean:52–58 passes the same estimates to probability mixtures. YM/OPE/Orbits/Kernel.lean:73–77 constructs one reader before demands by selecting a prescription once. ChallengeSpec.lean:70–80 gives the public coefficient/remainder inequalities, with the quantifiers for spectators and states visible.

Closest antecedents. W69 §II pp.1500–1501 Eq.(2.2) already proposes a weak OPE between fixed states, with finitely many terms at each finite order, powers and logarithms, elementary or composite fields, and extensions to multiple products. Its coefficient functions do not depend on the external states. Thus finite-order organization, composites and common coefficients are older ideas, not candidate innovations of this construction. Zimmermann, Normal products and the short distance expansion in the perturbation theory of renormalizable interactions (1973), pp.570–601, publisher abstract, proves Wilson's expansion in perturbation theory for two scalar fields. These are substantial conceptual antecedents; neither supplies the displayed nonperturbative Yang–Mills hierarchy and its quantitative power-with-slack bounds, uniform over allowed positions and phases for each fixed spectator word. Zimmermann's full paper remains unavailable; only its abstract is used. Wilson's common coefficients alone are not evidence of novelty in the Lean detector prescription. The inspected N3 §§9–11 renormalize the spin sources and their moments, with no corresponding all-composite, all-phase OPE statement in those sections or its main theorem.

Classification: ADAPTED-OTHER for the finite-order weak/composite OPE framework, high confidence; candidate-NEW only for the constructed hierarchy's quantitative estimates and compatibility, medium confidence relative to the corpus and low for worldwide novelty. Do not call the existence of an OPE itself a new idea. Do not claim perturbative coefficient matching, convergence in truncation order, or uniqueness of the prescription.

Rough mechanism split: ADAPTED-OTHER OPE principle and standard expansion tools 60%, candidate-NEW uniform implementation 40%.

36. VII.4 (b) — Contact extensions

Size: within VII.4; about 3,400 lines for contact extensions and about 290 for multiple-contact extensions.

What it does. The separated correlation functions are extended to distributions on collision diagonals using normal Taylor subtractions and sums over collapse scales and forests. The resulting extension agrees with the original hierarchy away from coincidences.

Lean evidence. YM/Contacts/Normal/Subtractions.lean:54–75 defines the normal Taylor polynomial, cutoff, and \(1 - T\) remainder on test functions. YM/Contacts/Normal/Subtractions.lean:76–83 proves commutation of subtractions in orthogonal normal directions. YM/Contacts/Solstice/Result.lean:23–31 records the shell-sum identities and summability over admissible forests. ContactSpec.lean:10–14 is exactly an extension-and-agreement statement for distributions on the full configuration space. YM/Contacts/Solstice/Result.lean:3–6 explicitly distinguishes this from reflection positivity on overlapping supports, contact Ward identities, and real-time time-ordered products.

Closest antecedent. BF00 §5.2, Theorems 5.2–5.3, preprint pp.23–25, gives the same core distribution-extension device. Equation (36) sums dyadic shells; equation (38) projects a test onto its Taylor-vanishing part; equation (39) defines an extension after subtraction. BF00 §6 then treats submanifolds and transverse scaling, making the normal-variable comparison more than a superficial resemblance.

Classification: ADAPTED-OTHER/STANDARD for extension by Taylor subtraction, high confidence; candidate-NEW input bounds and organization for this hierarchy, medium confidence. This is not persuasive evidence of NLSM ancestry: the pertinent older distribution-renormalization literature is a closer match. Existence of an extension is weaker than identification of a unique physical product at coincidence.

Rough mechanism split: STANDARD/ADAPTED-OTHER extension mechanism 80%, specialized input/organization 20%.

OS73 §3 p.89, Remark 1, already distinguishes separated Euclidean data from coincidence values: reconstruction does not determine the latter. This conceptual distinction, read with OS75's corrected setting, reinforces the scope limit; it is not a source for the whole forest-subtraction proof. W69's discussion of composite and time-ordered products likewise does not identify a unique physical contact extension here.

37. VII.5 — Leading-log coupling asymptotics

Size: 232 lines in spans of declarations used by the main theorem, only a small fraction of which is the coupling theorem itself.

What it does. It converts previously established running-coupling control into the asymptotic for the inverse square root of the action coefficient. It does not recompute the one-loop constant in this unit.

Lean evidence. YM/Pert/Asymptotics/RawAxis.lean:13–16 defines \(b_0 = 11 C_A / (48 \pi^2)\) and states the ratio estimate. The proof at YM/Pert/Asymptotics/RawAxis.lean:17–30 reads the Makes.wilson.running estimate and removes the fixed additive offset. At YM/Pert/Asymptotics/RawAxis.lean:35–48 it takes inverse square roots to get the leading-log limit. The module says at YM/Pert/Asymptotics/RawAxis.lean:3–4 that this proxy is not identified with a physical X-scheme coupling beyond the proved comparison.

Closest antecedents. Gross–Wilczek, PRL 30 (1973), 1343–1346, and Politzer, PRL 30 (1973), 1346–1349, are the older asymptotic-freedom source. N3 equation (1.4), p.6, and §12.1, p.95, likewise encode a linear inverse-coupling trajectory with a model-specific coefficient. The same general marginal-flow mechanism is present, but \(11 C_A\) is Yang–Mills data, not the sigma-model \(1 / (2 \pi)\) coefficient.

Classification: STANDARD consequence of an earlier nonstandard construction, high confidence. The provenance of the coefficient calculation belongs to the actual one-loop/matching units, not to this short endpoint theorem.

Rough mechanism split: STANDARD endpoint implication about 100%; the upstream coefficient and flow estimates are counted in III.7.

38. VII.6 — Flexible scaling and arbitrary radius sequences

Size: 4,205 lines in spans of declarations used by the main theorem.

What it does. It produces valid scaling stations along subsequences of an arbitrary given sequence of radii tending to zero. These stations retain the finite-degree observable controls needed to derive a full real-radius trace asymptotic.

Lean evidence. YM/TraceExtension/Prom/Zoom/Boats.lean:49–62, boatsSky, fits a sufficiently small prescribed radius to a valid station. YM/TraceExtension/Prom/Zoom/Boats.lean:75–87, wakesSky, recursively obtains a strictly monotone subsequence and stations of increasing degree. YM/TraceExtension/Verse/Cape.lean:77–88 uses that flexibility to obtain the all-radii conclusion.

Closest antecedent. N3 §15.2–15.3, pp.118–120, removes subsequence ambiguity by comparing bounded-offset trajectories and canonically normalizing them. The common logical purpose is to upgrade control of selected scales to an unrestricted limit, but the mechanisms differ: coupling-offset matching versus fitting arbitrary radii to stopping stations.

Classification: STANDARD subsequence logic + candidate-NEW technical extension of the construction estimates; medium confidence. This should not be counted as an independent new RG principle merely because its formal implementation duplicates many estimates.

Rough mechanism split: STANDARD subsequence/limit reasoning 70%, specialized uniform-scale estimates 30%.

39. VIII.1 — Common-witness assembly

Size: 12,056 lines in spans of declarations used by the main theorem.

What it does. It selects the construction before extraction demands, then places the limit, reconstructed phases, OPE, contact extension and trace statements on one common hierarchy.

Lean evidence. Construction.lean:15–31 quantifies the family first, the extraction order next, and the reader/contact/scaling witnesses afterward. Construction.lean:32–40 assembles the internal theorem into public properties. Solution.lean:4–10 specializes the fixed penalty and extracts the public volume refinement. ChallengeSpec.lean:3–4 emphasizes that all correlation terms are evaluations of one hierarchy.

Closest antecedent. N3 completion of Theorem 1.2, pp.94–95, explicitly requires the same field normalizations for construction, gap, and non-Gaussian correlations. Requiring one object to satisfy all conclusions is STANDARD mathematical logic and constructive-QFT discipline. The Lean implementation makes an unusually large quantity of compatibility and choice-order bookkeeping explicit.

Classification: STANDARD assembly/formal infrastructure, high confidence; it should not inflate the share attributed to mathematical novelty.

Rough mechanism split: STANDARD assembly/formal compatibility 95%, specialized organization 5%.

4. Cross-cutting comparisons

4.1 Where the ultraviolet construction ends

Balaban's cited pure-gauge sequence ends in uniform effective-density stability under the small-running-coupling assumption, B-Large-II pp.355–356. The later correlation analysis is not present there. That is a limitation of this particular gauge endpoint, not a statement about all his work. B-N-Corr and B-Corr explicitly study correlations in other models.

N2 joins exact weak-coupling blocking to preliminary mixing and a rectangular partition-function defect, §§2–3,8–9. The terminal finite-box input is transported to fine cutoffs, and repeated positivity/doubling gives a full gap. N3 builds the spin field on top of this massive route, §§8–11, then removes the bounded-offset ambiguity, §§12–15. Its continuum endpoint is canonical for the specified normalization and state.

The Lean instead proceeds from a common all-label construction, YM/SharpGram/Assembly/Flowers.lean:75–130, to fixed-volume full-mesh convergence and volume extraction. It then builds phases, reconstructs each phase internally and identifies stress/rotation/OPE data on the common hierarchy. The internal Construction.lean:15–31 result includes contacts and reconstruction of every phase; Solution.lean:4–10 exports a selected reconstructed theory and the common phase table through ChallengeSpec.lean:24–42, ChallengeSpec.lean:117–143. The full internal package should not be confused with the exact list of conclusions literally exported by JointExistence. Construction.lean:15–31 makes this order visible in the quantifiers. YM/EveryPhase/Decay.lean:173–192 provides polynomial residual decay, not a positive spectral threshold. Thus the NLSM resemblance is strongest before the infrared endpoint, and weakens where the two theories are identified. The absence of a gap assertion is not a proof of gaplessness.

MRS93 is a substantial direct four-dimensional gauge UV comparator. Its pp.325–327,352–355 uses regularized axial gauge, local background-dependent gauges and boson/ghost propagators, stabilizing cutoffs and symmetry-restoring counterterms for SU(2) in the trivial topological sector. Section VI pp.368–374 controls large-field determinant dressing; §§VII–VIII pp.374–378 treats coupling-flow remainders and Slavnov identities.

Its stated ultraviolet limit consists of gauge-fixed elementary Schwinger functions with a fixed infrared cutoff, p.327. That page expressly leaves a fully detailed proof unwritten and OS positivity as additional work. Page 329 explicitly excludes construction there of the renormalized physical composite operators; arbitrary elementary insertions in §VIII are not the same task. These limits are stronger evidence than a search failing to find an observable theorem. Compare the common composite/phase/OPE family at Construction.lean:15–31 and the volume-refinement target at ChallengeSpec.lean:177–188.

Seiler82 §5a pp.99–103 distinguishes a scaling-limit programme from its realization: the displayed pure YM loop limit is two dimensional, and the stability discussion is three dimensional. His external-field matter determinants in §6b pp.118–121 do not constitute a four-dimensional gauge-field construction. Neither older source should be described as lacking a continuum discussion; neither supplies the stated Lean endpoint.

4.2 Reflection positivity and chessboard estimates

FILS78 §4, preprint pp.33–36, already proves abstract and torus chessboard inequalities. Seiler82 §2a pp.14–16 supplies direct lattice-gauge reflection positivity and the transfer construction, Theorem 2.1. Theorem 2.2 states a chessboard bound for local matter functions and refers the gauge-field extension onward; FILS supplies the general dissemination principle. Osterwalder–Seiler 1978 remains unavailable in full: its abstract gives the reflection/transfer and strong-coupling setting, not a continuum weak-coupling theorem. The Lean's basic dissemination and seminorm statement, YM/Cells/Certificate.lean:14–24, belongs to that older tradition. Its gauge-law proof is consumed directly from Slice.reflected_positive.

N2 uses positivity for rectangular doubling and a spectral-gap conclusion, §§2,9. N3 uses both kinetic-row exponential moments, Lemma 7.3, p.62, and later massive reconstruction. The Lean uses chessboard roots to control composite observables and finite differences, YM/Extraction/Center.lean:21–40, and spectral positivity to manage phases and exterior flux. The principle is shared; the required estimates and endpoint are different. The word “reflection positivity” alone provides no evidence that the mass-gap argument was transplanted.

4.3 Composite insertions and observables

B-RG-I's gauge identities control Taylor coefficients of effective actions, §§4–5. B-Corr and B-N-Corr show that correlation analysis belongs to the wider Balaban programme. It would be wrong to infer an NLSM origin merely because the Lean has observables. MRS93 §VIII pp.377–378 treats arbitrary elementary gauge-fixed field insertions and Slavnov identities; p.329 explicitly leaves renormalized physical composite-operator construction outside the paper. This is a direct older source-analysis comparison, with a sharply different composite endpoint.

N3 §9.1–9.2, pp.75–80, carries independent spin sources and normalizes their linear response through the flow. The Lean shares exact source passage and retained supports: YM/Source/Insertion.lean:26–68 defines the inserted coefficient and YM/Source/Insertion.lean:89–110 requires its exact push along the same unmarked history. These are finite-degree/nilpotent coefficient constructions, not literally N3's analytic multivariate spin-source neighborhood; their realization is part of the common-family construction in YM/SharpGram/Assembly/Flowers.lean:75–130. Its observable target is wider: all finite-degree gauge-invariant composite labels, equation quotients, derivatives, stress and OPE channels. YM/Labels/Polynomial.lean:60–112, YM/MergeProof/Powder.lean:19–71 and YM/OPE/Orbits/Kernel.lean:62–77 show these added obligations. The source-propagation strategy is shared; the common all-label closure and identification are not supplied by the cited spin-source sections.

W69 §II pp.1500–1501 Eq.(2.2) already includes weak finite-order expansions, composite fields, multiple products and coefficients independent of external states. Z73 remains an abstract-only perturbative comparator. The potential contribution is the nonperturbative quantitative remainder for this constructed hierarchy, with bounds uniform in phase and allowed positions for each fixed word, together with the common composite-field identifications. Common coefficients and finite-order organization alone are not new ideas. The eventual radius threshold can depend on the spectator word; this is not one bound uniform over all such words. Likewise collision extension uses old Taylor-subtraction mathematics, BF00 §5.2, Eqs.(36)–(39). It does not automatically add contact Ward identities or physical time-ordered products; YM/Contacts/Solstice/Result.lean:3–6 explicitly limits the claim.

4.4 Running coupling, tuning and normalization

Balaban isolates the gauge-action marginal direction by symmetry and Taylor analysis, B-RG-I p.258 and §5. N2 preserves a fixed-reference geometric partition while shooting the actual coupling, Eq.(5.1), p.38, and §7. The Lean combines these two features, YM/RG/History.lean:63–85 and YM/RG/Activities.lean:70–75. Its exact reset must restore the whole density class, YM/AnalyticBridge/VERIFIED_RESET_STEP/Evaluation.lean:71–95. That restoration is more closely formulated in N2 Eqs.(5.40)–(5.42) than in the broad phrase “renormalize the coupling”.

The physical universal coefficient comes from gauge vector/ghost algebra, not the sigma-model beta function. Abbott82 §3.3, p.48, Eqs.(3.3.1)–(3.3.5), supplies a directly read older calculation of the coefficient. The Lean's momentum-shell calculation and bounded matching are at YM/Next/STATIONARY/OneLoop/MomentumIntegral.lean:326–408 and YM/Next/STATIONARY/Level/Transition.lean:153–178. The latter explicitly transfers the auxiliary coefficient to the stationary one; the matrices must not be conflated.

Finally, the internal coefficient is a coordinate in the action representation. YM/RG/History.lean:25–33 and YM/Pert/Asymptotics/RawAxis.lean:3–4 explicitly withhold identification with an independently defined physical coupling scheme beyond the proved comparison. N3's canonical normalization by susceptibility/second-moment length, Theorem 1.1, pp.5–6, is a different statement. The Lean's successful-recursion tuning uniqueness is not a claim that every possible regulator gives the same continuum theory.

GK85 §9 pp.233–237, especially Eq.(9.36), supplies an older controlled local-marginal extraction with inverse-coupling increments; §12 p.246 tunes the scalar critical mass by nested intervals. BY90 pp.360–363 Eqs.(1.17)–(1.31) extracts constant/quadratic terms and shifts the covariance exactly. MRS93 p.330 Eqs.(II.11)–(II.12) compares tentative and exact gauge-coupling schedules; p.376 explains perturbative orders and remainder control. These strengthen the older RG attribution without identifying the Lean gauge coefficient. GK's positive scalar coupling is infrared asymptotically free, BY's quadratic gradient matrix has a different role, and neither can be substituted numerically for the gauge running coordinate.

4.5 Order of constants and schedule growth

A correct order of choices is common constructive analysis: D-II Theorem 3.1, preprint p.29, and B-N-II Theorem 1, p.680, are explicit examples. N2 p.55, Eq.(5.43), chooses leading uniform constants before block size, later coefficient caps and logarithmic powers, then a large coupling threshold. The Lean similarly distinguishes pre-block choices from later waits, YM/RG/History.lean:43–61 and YM/Next/FLUCTUATIONS/Grants.lean:32–36.

What changes is the demand for one sequence serving unboundedly many finite orders. YM/SharpGram/Assembly/Flowers.lean:77–130 increases order requests and selects the family once. YM/FullVolume/Sluggish.lean:13–19, YM/FullVolume/Sluggish.lean:37–38 distinguishes eventual mesh statements uniform in all periods from a merely fixed-period bound. This quantified extension is meaningful. The existence of many parameters, however, is not evidence for many new ideas, nor proof that the dependency order has been verified throughout.

GK85 p.222 explicitly orders the block factor, locality scale, norm constants and final small-coupling threshold; its numerical reference envelope is already in p.211 Eq.(4.1) and p.221 Eq.(6.2). BY90 Theorems A/B pp.360–363 orders norm choices and controls cumulative regulator losses, with an all-derivative analytic norm at p.359 Eq.(1.13). MRS93 p.330 also separates tentative from exact running parameters. These direct antecedents rule out novelty of reference schedules, safe constant order or all-derivative control in isolation. The narrower Lean question is compatibility of growing stages with all gauge-composite, rank, stress and OPE demands.

4.6 Correlations, full mesh limits and volume limits

N3 Proposition 10.2, pp.84–86, uses its massive estimates to compare tori and pass tests to the limit. Its §15.3, pp.119–120, removes the real bare-coupling ambiguity. The Lean has a full mesh limit at each suitable fixed physical volume, PublicWilson.lean:83–90, followed by a subsequential volume limit, ChallengeSpec.lean:184–188. These quantifiers must remain distinct. The theorem does not thereby supply arbitrary simultaneous mesh/volume limits or uniqueness across all bare-coupling trajectories.

Dominated convergence and diagonal compactness are standard. YM/Extraction/Integrate.lean:115–153 explicitly uses a dominated-convergence theorem. The substantial constructive input is uniform control of the renormalized observables before that theorem is applied. A large number of Lean lines devoted to passing limits is not evidence that the limiting procedure itself originated in 2026.

MRS93 p.327 supplies an older fixed-IR UV limit with a different field/gauge scope and stated proof-detail limitations. Seiler82 pp.100–101 separates scaling-limit existence, independence of physical normalization and Euclidean invariance, warning that the independence claim need not hold. The Lean quantifiers at PublicWilson.lean:83–90 and ChallengeSpec.lean:184–188 should therefore be compared individually, rather than treated as one generic “continuum limit.”

4.7 Rotation restoration, phases and nonfreeness

N3 §§10.2–10.3, pp.84–87, compares inclined blockings and generates a dense rotation subgroup. Lean uses local stress identities, exterior flux, analytic continuation and finite-dimensional label comparison. YM/Ward/Walk.lean:15–26, YM/Ward/Walk.lean:60–84 supplies the analytic continuation along segments/walks; YM/Ward/Axis.lean:86–107, YM/Ward/Axis.lean:130–139 reaches collinear configurations and completes the density argument. The common goal of Euclidean covariance does not make these the same idea.

N3 §11.2 obtains its unique vacuum through exponential clustering. Lean decomposes the zero-energy algebra into characters, YM/Phase/Mixture.lean:65–86, then reconstructs each phase. The character space need not contain more than one inequivalent phase. The same mathematical object satisfying all later conclusions is a standard requirement, made explicit by the final assembly.

N3 §11.3, pp.93–94, detects non-Gaussianity by a nonzero fourth spin cumulant. Lean's trace inverse logarithm rules out the specified finite mass-shell Wick profiles, YM/TraceExtension/Wick/Refute.lean:75–105. Those are different witnesses, and neither should be described as a complete classification of all conceivable generalized free models. The continuous-mass generalized-free case is expressly outside the exclusion at YM/TraceExtension/Wick/Refute.lean:4–8.

5. What appears new, and what is routine

5.1 Ranked candidates for substantive new work

This ranks bounded candidates for substantive constructive work, not established priority. The strongest evidence concerns the precise estimates and their compatibility; the organizing physical principles are older. Generic gauge synchronization, analytic stationary branches and background regularity have direct older antecedents.

  1. One renormalized gauge-composite hierarchy with simultaneous identifications. YM/SharpGram/Assembly/Flowers.lean:75–130, YM/MergeProof/Powder.lean:19–71 and Construction.lean:15–31 combine finite composite labels, Maxwell quotient/rank, stress, phases and OPE on one family. MRS93 p.329 expressly excludes the physical composite-operator construction, although §VIII pp.377–378 includes arbitrary elementary field insertions; B-Large-II pp.355–356 leaves gauge observables for later work; N3 §§9–11 has the narrower spin-field hierarchy and different massive endpoint. Confidence: medium for this specific corpus-relative extension, low for worldwide novelty. Neither “one family” nor “all insertions” alone distinguishes it; broader constructive composite-field literature remains incompletely compared.

  2. Stress-driven restoration of rotations with exterior-flux control without an assumed gap. YM/Next/ROTATIONS/Yard.lean:49–77, YM/Next/ROTATIONS/Yard.lean:105–114, YM/Next/ROTATIONS/Route.lean:49–67 and YM/Ward/Walk.lean:60–84 give the route. N3 §§10–11 uses inclined regulators and massive estimates; MRS93 p.327 leaves the full OS setting outside its fixed-IR endpoint; Seiler82 p.101 presents Euclidean invariance as a requirement, not this construction. Confidence: medium within the corpus, low for priority over wider Euclidean/algebraic-QFT literature. The full exterior-flux argument still requires separate verification.

  3. A constructive all-radii trace two-point law and its specified nonfree exclusion. YM/TraceExtension/Verse/Cape.lean:53–88, ChallengeSpec.lean:168–175 and YM/TraceExtension/Wick/Refute.lean:75–105 are the precise targets. CDJ77 pp.442–445, Eqs.(3.19)–(3.21), gives a detailed older beta-function/operator-mixing trace law, including the distinction between the full relation and its on-shell simplification. N3 §11.3 uses a spin cumulant instead. Confidence: medium for the specific constructive realization, low for novelty of its physical prediction or of using a trace anomaly to detect interaction. The coefficient/stress matching remains a separate technical obligation.

  4. Quantitative finite-order OPE bounds on the common hierarchy, uniform over phases and allowed spectator positions for each fixed word. YM/OPE/Orbits/Kernel.lean:52–77 and ChallengeSpec.lean:70–80 give the common prescription and power-with-slack bounds. W69 §II pp.1500–1501 Eq.(2.2) already has finite-order weak/composite expansions and state-independent coefficients. Z73's full text remains unavailable; its abstract gives a perturbative scalar comparison. MRS93 p.329 does not supply the physical composite construction. Confidence: medium for the quantitative corpus-relative extension, low for worldwide novelty. Common coefficients, composites and finite-order truncation alone are not new; neither uniqueness nor convergence in OPE order is asserted.

  5. Controlled realization of the particular tagged, clipped/masked, compulsory-inventory gauge density transformation. YM/RG/Block.lean:39–79, YM/RG/Activities.lean:42–129, YM/SectorNext/GAUSSIAN_SECTOR_CHANGE/Windows.lean:65–74 and YM/Next/STATIONARY/Level/Transition.lean:153–178 locate the combination. N2 §5 pp.38–55 is the closest complete formula comparison; GK85 p.212 Eq.(4.8) and pp.218–221 already supplies compulsory covering and exact reassembly; B-Var pp.279,295–296 and B-RG-I §§4–5 supplies the gauge background/marginal programme; MRS93 §§VI–VIII pp.368–378 supplies another four-dimensional gauge stability/flow construction. Confidence: medium-low for novelty of the exact combination. A different representation is not by itself a new estimate, and the full stationary determinant bridge has not been verified. The broad idea of closing a gauge RG with controlled marginal flow is an older antecedent.

  6. The particular finite-torus normalization estimate and its use with composite detectors. YM/Next/RANK/Determinant.lean:240–274 compares exponents \(3 V + 1\) and \(3 V - 3\); YM/Next/RANK/Stamp.lean:131–167 converts the residual global power into a cell-root cost; YM/MergeProof/Powder.lean:48–71 separately supplies detector/rank data. MRS93 p.368 Eq.(VI.1), Lemma VI.1 p.369 and B-Large-II pp.356–359 are substantial normalization antecedents. Seiler82 pp.118–121 concerns external-field matter determinants, not this pure-gauge count. N3 §§7,9 uses different rough-region and spin-source comparisons. Confidence: low for novelty of the torus estimate, medium for the specific combined detector application relative to the sources read. The first exponent is ordinary spanning-tree counting; the exact sweep and uniform hypotheses remain unmatched, not proved historically new.

No item is established as a new physical principle. Potential novelty lies in rigorous quantitative control and compatibility in the stated setting. Source-level formalization makes dependencies and quantifiers explicit; code length does not measure mathematical originality.

5.2 Routine or established ingredients that should receive ordinary credit

Ingredient Lean example Established antecedent and access
Wilson lattice law and Haar invariance YM/Wilson/Law.lean:55–99 B-RG-I p.252 Eq.(0.2), directly read; Wilson 1974 historical reference
Gaussian conditioning and Schur complements YM/PREC/TreeMatch.lean:76–103 B-Prop-I pp.26–29, directly read
Log determinants and inverse expansions YM/AnalyticBridge/RELATIVE_LOGDET_TRACE/Sector.lean:54–71 B-UV3 p.272 Eq.(63), B-Large-II pp.356–358; MRS93 pp.368–369 adds background-dependent stability; N2 p.44 gives the closer retained defect format
Connected graph/tree and cluster bounds YM/Push/ABSOLUTE_SUMS/Clusters.lean:224–243 Bry §3, pp.138–143; Cam §3, pp.525–528
Marginal subtraction by symmetry YM/RG/Activities.lean:70–75 B-RG-I §5, pp.292–297; BBS19 §5.2, preprint p.72, for a different model's localization operator
Gauge one-loop coefficient YM/Next/STATIONARY/OneLoop/MomentumIntegral.lean:326–338 Abbott82 §3.3 p.48, Eqs.(3.3.1)–(3.3.5), directly read
Chessboard dissemination YM/Cells/Certificate.lean:14–24 FILS78 §4, preprint pp.33–36; Seiler82 pp.14–16 for gauge RP and the displayed matter chessboard version
Reflection-positive Hilbert reconstruction YM/OS/Dynamics.lean:63–107 OS73 §4 p.90 for the mechanism; corrected OS75 §IV.1 pp.287–288, Eq.(4.1), directly read in part; exact Lean hypothesis equivalence unresolved
Wick reference calculations YM/Next/MAXWELL/System.lean:106–179 Standard free-field theory; not attributed to a particular recent paper
Dominated convergence YM/Extraction/Integrate.lean:115–153 Standard integration theory; theorem explicitly invoked in Lean
Taylor subtraction at diagonals YM/Contacts/Normal/Subtractions.lean:54–83 BF00 §5.2, Eqs.(36)–(39), and §6.4, directly read
Stability of finite-dimensional rank YM/Next/RANK/Matrices.lean:63–86 Standard linear algebra; the comparison estimates, not this implication, carry the constructive content

Other directly matched older devices are compulsory bad-region covering (GK85 p.212 Eq.(4.8) versus YM/RG/Activities.lean:113–129), analytic stationary gauge backgrounds (B-Var Proposition 6 pp.295–296 versus YM/Work/F04.lean:62–86), and exact local marginal extraction/reblocking (BY90 pp.360–366 versus YM/AnalyticBridge/VERIFIED_RESET_STEP/Evaluation.lean:71–95). Their gauge-specific estimates still require work; their general mechanisms are established antecedents.

6. Open questions, limits and priorities for expert checking

6.1 Coverage and source reliability

The source reading is selective: it does not cover every line of Lean or every proof in the literature. The counts of declarations used by the main theorem rest on existing dependency information and source intervals, and do not independently establish the theorem’s correctness. At the principal citations, definitions and proof bodies support the comparisons; comments alone are not treated as theorems. The comparisons refer to the pinned Lean source and dated primary papers; they do not depend on revisions to secondary expositions.

Full texts of all six OpenAI preprints and the 24 works in §7.3 were available. The reading covered introductions, theorem statements, overall structure and the technical passages cited here, with particular attention to N2 §5 and N3's comparison and continuum sections. This is selective full-text reading, not a claim that every paper or proof was read in full. Human rewrites and assessments are not evidence for the provenance claims, and the papers' own claims of originality are not treated as established historical facts.

The additional sources, including the five 1985 Bałaban gauge papers, MRS93, GK85, BY90, Seiler82, OS73/OS75, CDJ77 and W69, were read in the parts specified in §7.4. Full-text availability does not mean that every page or proof was read. Seiler's geometric appendix was available but not substantively used. OSe78 and Z73 remain available only as previously accessed abstracts. The background references in §7.5 were not directly read. None of these access levels amounts to independent human endorsement of the comparisons.

6.2 Missing or incompletely accessed literature

The five 1985 Bałaban gauge papers, MRS93, GK85, BY90, Seiler82, OS73/OS75, CDJ77 and W69 are available in full text. Relevant sections were read in part, with exact extents in §7. Availability is not a claim to have checked every proof. The old access-based uncertainty about whether Bałaban treated analytic stationary backgrounds and regular gauge transfer is resolved; exact theorem equivalence with the Lean implementation is not.

Still unavailable in full: Osterwalder–Seiler, Annals of Physics 110 (1978), 440–471, and Zimmermann, Annals of Physics 77 (1973), 570–601. Their previously accessed abstracts remain the evidence limit. Seiler's exposition does not silently upgrade access to the Osterwalder–Seiler original.

Reading remains incomplete for the broader Gawędzki–Kupiainen corpus beyond GK85, Brydges–Dimock–Hurd, Feldman–Magnen–Rivasseau–Sénéor, modern rigorous composite/OPE constructions, and older algebraic-QFT phase decompositions. Seiler's cited gauge-fixing and gauge-chessboard extensions also remain incompletely followed. The targeted determinant passages do not establish a predecessor or priority for the exact torus sweep.

MRS93 pp.326–327 explicitly presents an account short of a fully detailed convergence proof; B-Var p.308 omits detailed derivative calculations; Seiler82 p.121 explicitly labels its four-dimensional external-field determinant statement a Quasi-Theorem. Those qualifications are part of the sources' scope. They neither erase the antecedent ideas nor certify the Lean estimates. Polyakov, Hasenfratz, Zinn-Justin and Lüscher remain literature leads unless a specific result is directly matched.

6.3 Highest-value unresolved technical comparisons

  1. The stationary determinant bridge. Follow every premise from the blocked joint Hessian through pole_reduced to the auxiliary vector/ghost calculation, YM/Next/STATIONARY/Level/Transition.lean:153–178. The cited interfaces and coefficient calculation support the comparison; not every estimate supplying that bridge has been examined.

  2. The complete nonlinear large-field step. Compare Balaban's aged-region R ratios, B-Large-I Eqs.(0.2)–(0.6), against every actual Lean seed reorganization, not only the final activity interface. A related replacement hidden upstream would narrow the stated organizational difference.

  3. Gauge normalization priority. Compare the specific torus 3V+1 versus 3V−3 sweep in YM/Next/RANK/Determinant.lean:240–274 with further finite-volume gauge literature and Seiler's references. MRS93 §VI already supplies substantive background determinant stability; Seiler82 §6b treats a different external-field matter problem. No exact match was found in the targeted passages; worldwide novelty is not established.

  4. All-label uniformity and quantifier order. Check that the intended estimates survive the growing orders in YM/SharpGram/Assembly/Flowers.lean:75–130 and the pre-block order of constants. Source similarity does not settle that mathematical obligation.

  5. Exterior flux and phase theory. Read all upstream spectral estimates used by YM/Next/ROTATIONS/Yard.lean:49–77, and compare the train-commutation phase argument with older Euclidean extremal-state methods.

  6. Reconstruction and contacts. Compare the exact graded Lean domains and joint distributional boundary estimates with OS75 §III p.285 and §IV.1 p.287 Eq.(4.1). The conceptual mechanism is settled; theorem equivalence is not. Keep the separate contact extension within its stated scope.

  7. Trace normalization and physical interpretation. Check the stress normalization and field/coupling conventions against CDJ77 pp.442–445 Eqs.(3.19)–(3.21), including gauge/ghost/equation-of-motion terms and the on-shell simplification. Matching the internal running coordinate and renormalized composite normalization to a physical scheme remains unresolved.

  8. Causal provenance. Draft histories, author testimony and generation records would be needed to decide which sources actually influenced the authors. Mathematical matching alone cannot answer that historical question.

These limits do not erase the strongest positive comparisons: the matched retained-density formulas remain direct evidence. They do prevent upgrading “closest inspected antecedent” to a proved history of invention or copying.

7. Bibliography and source access

7.1 Lean source and orientation

The mathematical comparisons use the released Lean source at commit 3456e5e75cf8d4cd558955a5c7638fba2d68f464. The line links throughout this page identify the relevant definitions and proofs. The comparisons and size counts concern the source text and do not independently establish the theorem’s correctness.

For orientation, see statement.md, summary.md, the architecture and the units. These expositions are not primary evidence for mathematical antecedents.

7.2 Six OpenAI preprints

All six are preprints credited to OpenAI, not journal articles. Their full texts were available and read selectively for their structure and the mechanisms cited above. Links point to their dated folders in OpenAI's math preprints; the content comparisons rest on the dated texts, not on subsequent changes to that collection.

7.3 Balaban and related works: 24 sources

Full texts of these 24 works were available; introductions, theorem statements and selected relevant technical sections were read. This is not a claim of complete proof-by-proof reading. The descriptions identify each work's subject so that gauge, spin and scalar results remain distinct.

  1. B-Green. Tadeusz Bałaban, “Regularity and Decay of Lattice Green's Functions,” Communications in Mathematical Physics 89 (1983), 571–597. Gauge-covariant difference operators; §2 generalized random walks.

  2. B-Prop-I. T. Bałaban, “Propagators and Renormalization Transformations for Lattice Gauge Theories. I,” Communications in Mathematical Physics 95 (1984), 17–40. Gaussian vector gauge theory, block averages and gauge change.

  3. B-Prop-II. T. Bałaban, “Propagators and Renormalization Transformations for Lattice Gauge Theories. II,” Communications in Mathematical Physics 96 (1984), 223–250. Restrictions on several scales and irregular domains.

  4. B-RG-I. Tadeusz Bałaban, “Renormalization Group Approach to Lattice Gauge Field Theories. I. Generation of Effective Actions in a Small Field Approximation and a Coupling Constant Renormalization in Four Dimensions,” Communications in Mathematical Physics 109 (1987), 249–301. Small-field effective actions, Ward identities, marginal extraction.

  5. B-RG-II. Tadeusz Bałaban, “Renormalization Group Approach to Lattice Gauge Field Theories. II. Cluster Expansions,” Communications in Mathematical Physics 116 (1988), 1–22. Localization and exponentiated connected expansion.

  6. B-Conv. T. Bałaban, “Convergent Renormalization Expansions for Lattice Gauge Theories,” Communications in Mathematical Physics 119 (1988), 243–285. Full density with large-field domains; superrenormalizable closure.

  7. B-Large-I. Tadeusz Bałaban, “Large Field Renormalization. I. The Basic Step of the R Operation,” Communications in Mathematical Physics 122 (1989), 175–202. Repeated renormalization of large-field densities.

  8. B-Large-II. Tadeusz Bałaban, “Large Field Renormalization. II. Localization, Exponentiation, and Bounds for the R Operation,” Communications in Mathematical Physics 122 (1989), 355–392. Conditional four-dimensional ultraviolet stability.

  9. B-Higgs. Tadeusz Bałaban, “(Higgs)₂,₃ Quantum Fields in a Finite Volume. I. A Lower Bound,” Communications in Mathematical Physics 85 (1982), 603–626. Scalar/vector model in dimensions two and three, not pure four-dimensional YM. The original first-page header itself says 603–636; the 24-page PDF ends at 626. The publisher record confirms 603–626.

  10. B-N-I. Tadeusz Bałaban, “A Low Temperature Expansion for Classical N-Vector Models. I. A Renormalization Group Flow,” Communications in Mathematical Physics 167 (1995), 103–154. Low-temperature spin-wave programme in \(d \geqslant 3\).

  11. B-N-Var. Tadeusz Bałaban, “The Variational Problems for Classical N-Vector Models,” Communications in Mathematical Physics 175 (1996), 607–642. Background minimizers, analyticity and regularity.

  12. B-N-Loc-I. Tadeusz Bałaban, “Localization Expansions. I. Functions of the ‘Background’ Configurations,” Communications in Mathematical Physics 182 (1996), 33–82. Localization of nonlocal minimizers and covariance functions.

  13. B-N-II. Tadeusz Bałaban, “A Low Temperature Expansion for Classical N-Vector Models. II. Renormalization Group Equations,” Communications in Mathematical Physics 182 (1996), 675–721. N-vector RG equations, not a general gauge survey.

  14. B-N-III. Tadeusz Bałaban, “A Low Temperature Expansion for Classical N-Vector Models III. A Complete Inductive Description, Fluctuation Integrals,” Communications in Mathematical Physics 196 (1998), 485–521.

  15. B-N-Loc-II. Tadeusz Bałaban, “Renormalization and Localization Expansions. II. Expectation Values of the ‘Fluctuation’ Measures,” Communications in Mathematical Physics 198 (1998), 1–45.

  16. B-N-Large. Tadeusz Bałaban, “The Large Field Renormalization Operation for Classical N-Vector Models,” Communications in Mathematical Physics 198 (1998), 493–534.

  17. B-N-Corr. Tadeusz Bałaban and Michael O'Carroll, “Low Temperature Properties for Correlation Functions in Classical N-Vector Spin Models,” Communications in Mathematical Physics 199 (1999), 493–520. Magnetization and Goldstone asymptotics in \(d \geqslant 3\).

  18. B-Corr. Tadeusz Bałaban and Michael O'Carroll, “A Simple Method for Correlation Functions via the Effective Actions in the Renormalization Group Framework,” Annals of Physics 260 (1997), 1–8.

    The worked model is the massive two-dimensional Gross–Neveu model, not NLSM or YM.

  19. B-Spin. Tadeusz Bałaban, “A Low Temperature Expansion and ‘Spin Wave Picture’ for Classical N-Vector Models,” in V. Rivasseau (ed.), Constructive Physics: Results in Field Theory, Statistical Mechanics and Condensed Matter Physics, Lecture Notes in Physics 446 (Springer, 1995), 201–218.

    Publisher metadata distinguishes the 1995 volume from 2005 online posting.

  20. D-I. J. Dimock, “The Renormalization Group According to Balaban. I. Small Fields,” Reviews in Mathematical Physics 25 (2013), 1330010; arXiv:1108.1335. Cited preprint version: October 28, 2018, 52 pages.

  21. D-II. J. Dimock, “The Renormalization Group According to Balaban. II. Large Fields,” Journal of Mathematical Physics 54 (2013), 092301; arXiv:1212.5562. Cited preprint version: August 25, 2018, 94 pages.

  22. D-III. J. Dimock, “The Renormalization Group According to Balaban. III. Convergence,” Annales Henri Poincaré 15 (2014), 2133–2175; arXiv:1304.0705. Cited preprint version: March 29, 2018, 37 pages. Publisher record gives issue 2014, online 2013. All three Dimock papers treat scalar \(\varphi^4\) in dimension three; they do not supply non-Abelian gauge arguments.

  23. Bry. David C. Brydges, “A Short Course on Cluster Expansions,” in K. Osterwalder and R. Stora (eds.), Critical Phenomena, Random Systems, Gauge Theories, Les Houches Session XLIII (1984), Part I, North-Holland/Elsevier (1986), 129–183. Published in 1986 from the 1984 lectures.

  24. Cam. Camillo Cammarota, “Decay of Correlations for Infinite Range Interactions in Unbounded Spin Systems,” Communications in Mathematical Physics 85 (1982), 517–528. Polymer convergence and decay; not authored by Bałaban.

7.4 Other literature actually used

The following sources supplement the 24 works in §7.3. All are read in part; no complete proof verification is implied.

7.5 Background references and unread originals

Seiler82 and the five 1985 Bałaban gauge papers are directly read sources in §7.4. Osterwalder–Seiler 1978 and Zimmermann 1973 remain unavailable in full, with abstract-only entries there. Names listed in §6 as further literature comparisons are not additional sources silently used here. Direct reading, partial reading, abstract-only access and background knowledge must remain distinguishable.

Source keys

Full publication details and reading limits are in §7. NLSM means nonlinear sigma model; RG means renormalization group; UV means ultraviolet.

Key Source
N1 OpenAI, Exponential decay in two-dimensional classical O(n) models (2026)
N2 OpenAI, Sharp mass bounds for the two-dimensional O(4) model (2026)
N3 OpenAI, The canonical massive continuum limit of the two-dimensional O(3) model (2026)
N4 OpenAI, An Isolated Particle Pole for the Two-Dimensional O(3) Spin Field (2026)
N5 OpenAI, Exact mass asymptotics for the two-dimensional O(4) lattice model (2026)
N6 OpenAI, The Critical Spin Field of the Planar XY Model (2026)
B-Average; B-Gauge; B-Back; B-UV3; B-Var Bałaban, averaging, regular configurations/gauge fixing, background propagators, three-dimensional stability and variational backgrounds (1985)
B-Green; B-Prop-I, B-Prop-II Bałaban, lattice Green's functions and gauge propagators (1983–1984)
B-RG-I, B-RG-II Bałaban, Renormalization Group Approach to Lattice Gauge Field Theories (1987–1988)
B-Conv; B-Large-I, B-Large-II Bałaban, convergent gauge expansions and large-field renormalization (1988–1989)
B-Higgs Bałaban, finite-volume Higgs lower bound (1982)
B-N-I, B-N-II, B-N-III; B-N-Var; B-N-Loc-I, B-N-Loc-II; B-N-Large; B-Spin Bałaban, N-vector RG, variational, localization and large-field works (1995–1998)
B-N-Corr; B-Corr Bałaban and O'Carroll, N-vector correlations (1999) and effective-action correlation methods (1997)
D-I, D-II, D-III Dimock, The Renormalization Group According to Balaban (2013–2014)
Bry; Cam Brydges, cluster expansions (1986); Cammarota, polymer convergence and decay (1982)
Abbott82; GW73; P73 Abbott, background-field method (1982); Gross–Wilczek and Politzer, asymptotic freedom (1973)
FILS78; OSe78 Fröhlich–Israel–Lieb–Simon, reflection positivity; Osterwalder–Seiler, lattice gauge theory (1978)
GK85 Gawędzki–Kupiainen, massless lattice φ⁴₄ and infrared asymptotic freedom (1985)
Seiler82 Seiler, constructive gauge theory and statistical mechanics (1982), read in part
BBS19; BY90 Bauerschmidt–Brydges–Slade, RG method (2019); Brydges–Yau, gradient perturbations (1990)
Fed87; MRS93 Federbush, lattice–continuum duality (1987); Magnen–Rivasseau–Sénéor, YM4 with an infrared cutoff (1993)
OS73, OS75 Osterwalder–Schrader, Euclidean reconstruction (1973, 1975)
CDJ77 Collins–Duncan–Joglekar, trace and dilatation anomalies (1977)
W69; Z73; BF00 Wilson, operator products (1969); Zimmermann, short-distance expansion (1973); Brunetti–Fredenhagen, distribution extension (2000)
Wilson74; Glimm–Jaffe87 Background references: lattice gauge theory and functional-integral QFT