Evidence Paper Review

 Quick Explanation Copied

 Long Explanation

Core Contribution and Supporting Evidence

The paper solves a concrete computational problem in arithmetic geometry: given a superelliptic curve YK : y^n = f(x) over a local field K, compute the oK-lattice M of integral differential forms — a quantity needed for the Néron model of the Jacobian and for numerically testing the Birch and Swinnerton-Dyer conjecture Motivation via Néron models and BSD verification.

The key conceptual move is stated in the abstract and Proposition 1.1/Corollary 1.2: instead of computing a regular model (the prior state of the art, e.g. Donnelly's Magma implementation or Dokchitser's scripts, both with strong restrictions), it suffices to build a model with only rational singularities, for which the canonical sheaf pushforward is desingularization-invariant. Theorem 1.5 proves that if n is invertible in oK and the branch divisor on a regular P1-model is normal crossing, the normalization Y has only tame quotient (hence rational, by Boutot) singularities Theorem 1.5: rational singularities via Kummer covers.

Algorithms 3.10/3.13 construct the required regular P1-model using MacLane inductive valuations and shortest N-paths, extending Rüth's thesis framework without assuming an algebraically closed residue field — a genuine technical refinement over prior work Valuation-based model construction without algebraically closed residue field.

Validation: Two Worked Examples

The only quantitative validation is two examples over Q5 sharing the branch polynomial f = (x²−5)³−5⁵: the genus-2 curve yields basis B = (5, x), the genus-4 curve yields B = (5², 5x, x²−5, y). The per-valuation data from Tables 1–2 show the internal machinery (orders of vanishing v(dx) and w(η)) is nontrivial and varies across the valuation tree:

These observed bases are direct, checkable outputs; correctness in principle follows from Theorem 4.14, but no independent cross-check against an alternative implementation is reported Algorithm 4.12 and correctness theorem.

Critical Assessment and Blind Spots

Strengths: complete proofs, reproducible public code (regularmodels and MCLF repositories on GitHub), careful bookkeeping of non-integral order-of-vanishing conventions, and honest positioning against Magma/Dokchitser alternatives.

Limitations the paper states or implies: scope restricted to superelliptic curves with n invertible in oK (Assumption 1.4) — tame covers only, so wild cases (p | n) remain open; only two examples, both over Q5 with the same f; the constructed model is not regular, and the authors defer regular-model computation to future work. Blind spots BGPT adds: no benchmarking of runtime or scaling behavior versus existing implementations; no adversarial examples (e.g. highly degenerate f, large residue-field extensions where the footnote-7 issue — multiple extensions per valuation — could blow up computation); no numerical cross-validation of the lattice via discriminant/conductor invariants. The proof of Proposition 1.1 also leans on Boutot's result outside its stated hypotheses (footnote 1 concedes schemes over a field of characteristic zero), which the authors argue goes through unchanged — plausible but not fully re-derived, a residual epistemic gap.

What would change the assessment: a counterexample where the normalization of a normal-crossing model fails rationality, or a case where the computed basis fails v(ω) ≥ 0 for some vertical valuation, would falsify the core correctness claim. Confidence in the theory is high; confidence in practical coverage breadth is moderate pending more examples.

 BGPT Paper Review

Study Novelty

The rational-singularities-instead-of-regular-models strategy and the extension of MacLane valuation trees to non-algebraically-closed residue fields are genuinely new for superelliptic curves, though the ingredients (Boutot, MacLane, Kummer covers) are classical; closely related results appear in Obus–Srinivasan and Muselli.

Scientific Quality

Complete proofs, referee-reviewed in Math. Comp., public implementation, honest treatment of conventions and edge cases. Deductions: reliance on Boutot outside stated hypotheses (footnote 1), only two examples, no benchmarks or cross-validation against alternative implementations.

Study Generality

Scope is limited to superelliptic curves with n invertible in oK (tame covers); the valuation framework generalizes in principle (Remark 1.6 notes plane quartics with V4-monodromy), but the algorithm as presented is class-specific.

Study Usefulness

Directly useful for BSD verification pipelines and reduction-type computations, covering cases not handled by Magma regular-model implementations; usefulness limited by niche scope until integrated or extended.

Study Reproducibility

Full public Sage/MCLF code (github.com/swewers/regularmodels) with a Jupyter notebook reproducing the examples; however, no independent replication or timing data is reported in the paper.

Explanatory Depth

Deep local analysis (adjunction, codifferent, Kummer quotient singularities, reduced bases) with clear mechanistic reasoning, though some proofs cite prior results rather than rederive them, and motivation for design choices like η = dx/y^(n-1) is brief.