{"id":"7a235221-30f0-4464-ab5e-7a744486624b","arxiv_id":"2607.23973","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Connected-patch orientation correlations ξ±(θ,ν) give a coordinate-independent test of statistical isotropy on the sphere and separate global from local alignment in sheared random fields, while dipole modulation leaves no orientation signal.","lead":"This paper introduces a new statistical tool — correlations between the orientations of connected patches on a sphere — for testing whether a sky map is the same in every direction. Earlier shape-based tests either change their answer when the coordinate system changes or silently wash out real signals; the new method avoids both problems in simulated maps.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Finite-size transport distortion is quantified only for β in isotropic fields; the eigenvector orientation error—which ξ± directly uses—and the large-component weight in anisotropic ensembles are unmeasured, leaving the 'intrinsic, unbiased' claim under-supported.","rationale":"The reader's weakest assumption correctly identifies the finite-size geometric distortion as the load-bearing soft spot. My read sharpens it in two ways. First, the actual observable ξ± is not a function of β alone; it uses the eigenvector orientation angle ψ, and transport-induced rotations of v_μ enter ϵ₊ and ϵₓ at first order. Appendix B quantifies only Δβ, which is insensitive to the orientation error. Second, the claim that pair counting does not weight by size is not dispositive because the |ϵ| weight is shape-based, and in the simulated ensembles the most elongated components are plausibly the large ones. For the global-shear field, the coherent alignment makes components large along the preferred direction, so the distortion could be larger than in the isotropic calibration. This does not mean the paper's conclusions are wrong; it means the central 'intrinsic, unbiased' assertion is not yet supported for the regime where the strongest claims are made. The proposed size-cut and eigenvector-orientation checks are concrete and would settle whether the concern lands. The reader already marks the paper conditional on this and related reproducibility issues; my read does not move the verdict.","tokens_in":28959,"tokens_out":7138,"duration_ms":79657,"concrete_test":"Using the authors' pipeline (or a reimplementation), recompute ξ± for the global- and local-shear ensembles under a size cut that retains only components with A_μ ≤ A_smooth (or, equivalently, removes the top 10% of the total |ϵ| weight). If the ξ₊ decay scale or the antipodal ξ₋ amplitude shifts by more than the ensemble error bars, the large-component transport distortion is contaminating the headline discrimination. Independently, on 10 isotropic realizations, measure the eigenvector error Δψ_μ between the transported and non-transported construction; if the components carrying the largest |ϵ| have Δψ_μ ≳ 5–10°, the orientation signal in ξ± is not faithfully extracted and the 'unbiased' claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section IV defines the estimator by great-arc transporting each component's boundary normals to its geometric center p̄_μ and then diagonalizing W^{0,2}_{1,μ}. The central claim that ξ± is an intrinsic, unbiased test of statistical isotropy therefore assumes this short transport is a negligible distortion for the components that actually drive the pair averages. The paper's only check (Appendix B, Fig. 9) compares the shape parameter β with and without transport on 10 isotropic realizations; it reports Δβ ≲ 1% for most components. That is not the quantity ξ± needs. (i) ξ± enters through the orientation angle ψ via ϵ₊ = −|ϵ| cos 2ψ and ϵₓ = −|ϵ| sin 2ψ; a small transport-induced rotation of the principal eigenvector v_μ changes ϵ₊ and ϵₓ at first order in the angle (derivative −2 sin 2ψ), so a few-degree eigenvector error can be a large fractional error in the correlation. Appendix B never measures the eigenvector direction. (ii) The pair average weights each pair by |ϵ_μ||ϵ_η|; the most elongated components contribute most. There is no demonstration that those components are small; in Gaussian fields large excursion components are typically the most elongated, and in the global-shear ensemble the preferred direction coherently elongates components over large scales. The paper's statement that 'pair counts do not weight by size' misses that the shape weight |ϵ| is itself size-correlated. (iii) The distortion is only tested for isotropic fields. In anisotropic fields, components are larger and more coherently oriented, so the same transport can rotate v_μ non-negligibly; this is precisely the regime in which the ξ₋ 'globally coherent fraction' claim at large separations is made. Thus the assertion that ξ± is intrinsic/unbiased rests on an unquantified approximation for the population that dominates the statistic.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper addresses the ambiguity of defining Minkowski tensors on S^2. It shows that the standard pixel-sum estimator is non-covariant under local rotations, while the parallel-transport estimator, although covariant, systematically isotropizes anisotropic signals. To evade both problems, the authors define Minkowski tensors for individual connected components of excursion sets and construct pairwise orientation correlation functions ξ±(θ,ν) built from scalar quantities. They validate these statistics on simulated isotropic, dipole-modulated, globally sheared, and locally sheared Gaussian fields, and provide an analytic calculation showing that dipole modulation affects the traceless part of the tensor only at second order in the modulation amplitude. The central claim is that ξ± provide an intrinsic, unbiased, scale-dependent test of statistical isotropy on S^2.","tokens_in":29240,"tokens_out":3099,"duration_ms":36636,"significance":"If the central claim holds, this is a useful methodological contribution: it offers a covariant statistic that retains sensitivity to anisotropic alignment, in contrast to the one-point estimators analyzed in the paper. The analytic result for dipole modulation (Eq. 48), the explicit non-covariance demonstration (Eqs. 6–7), and the extensive validation on four ensembles of N_real=400 simulations are concrete strengths. The statistic is also falsifiable: it makes distinct predictions for global versus local shear, and the mirror relation ξ−(θ)≃ξ+(180°−θ) is a sharp, testable consequence of the construction. The paper would be a valuable reference for future CMB and large-scale-structure isotropy analyses, provided the finite-size transport issue is addressed quantitatively.","major_comments":[{"comment":"The practical covariance of ξ± rests on the assumption that great-arc transport of boundary normals to the component center p̄_μ produces negligible distortion. Appendix B quantifies this only for the scalar shape parameter β, in isotropic fields, on 10 realizations. But ξ± enters through the orientation angle ψ via ε_+ = −|ε| cos2ψ and ε_× = −|ε| sin2ψ; an eigenvector error δψ changes ε_+ and ε_× at first order (derivative −2 sin2ψ). Moreover, the pair average weights by |ε_μ||ε_η|, and the most elongated components carry the largest weight, while the paper's statement that 'pair counts do not weight by size' misses the fact that |ε| is size-correlated in Gaussian fields. The finite-size distortion must be measured for the eigenvector direction, for isotropic and anisotropic fields, and the resulting |ε|-weighted bias in ξ± must be shown to be subdominant to the signals in Figs. 6–7. Th","section":"§IV and Appendix B"},{"comment":"The interpretation that 'the decay scale of ξ+ measures the angular coherence of the alignment, and the antipodal amplitude of ξ− measures its globally coherent fraction' is not given a quantitative definition or calibration. For the locally sheared field, the ξ− rise is attributed to a sub-dominant globally coherent component generated by the quadratic construction in Q̃, but no measure of 'coherent fraction' is defined, and the mapping between the ξ− amplitude and that fraction is not established. Without a formal definition or a controlled test (e.g., fields with known coherent fractions), this central interpretation remains heuristic. Please define the quantity and demonstrate the claimed proportionality, or weaken the claim accordingly.","section":"§VI and §VIII"}],"minor_comments":[{"comment":"Appendix B measures Δβ for β only; it would be more informative to also plot the transport-induced rotation of the principal eigenvector, since that is the quantity entering ξ±. The current figure cannot rule out a few-degree rotation that would induce O(10%) fractional changes in ε_+ and ε_×.","section":"Appendix B"},{"comment":"The notation ξ+ and ξ− is introduced immediately after Eq. (14) but the sign convention is not explicitly motivated. The reader would benefit from a sentence stating that these are the standard weak-lensing E/B-mode combinations, with ξ− negative for 45°-rotated patterns, to facilitate comparison with the lensing literature.","section":"§IV"},{"comment":"The dipole and shear amplitudes λ=0.76 are fixed 'to ensure a large ∼40% dipole modulation amplitude and a ∼4% shear contribution'. For the dipole field, λ=0.76 gives a 40% amplitude, but the shear fractional contribution is stated as ≃0.056λ, so λ=0.76 gives ∼4.3%; the wording could be tightened to avoid implying the same parameter controls both effects in the same way.","section":"§V"},{"comment":"The error bars in Fig. 6 are the standard error of the mean over 400 realizations. For a cosmological application, a single sky has cosmic variance; the error bars therefore do not represent the expected scatter of a single realization. Consider showing the realization scatter (e.g., shaded 16/84% bands) in addition to the error on the mean, as done in Fig. 5.","section":"Figure 6"},{"comment":"The marching-triangle algorithm updates W^{0,2}_1 by adding |e_12|(n⊗n) evaluated at one endpoint t_1, and the text notes this neglects variation of n along the arc. This is reasonable for small triangles, but a brief comment on the induced error (order |e_12|^2 times the field curvature) would help quantify the approximation.","section":"Appendix A"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid methodological contribution with a clear and correct core derivation. The main risk is not the covariance argument or the analytic dipole result, but the finite-size transport distortion, which is only tested for β in isotropic fields and not for the eigenvector orientation used by ξ±. This is a fixable issue: the authors should add numerical tests measuring the transport-induced rotation of v_μ, weighted by |ε|, for both isotropic and anisotropic ensembles, and show that ξ± are robust. If such tests confirm the current conclusions, the paper would be publishable; without them, the central 'intrinsic, unbiased' claim is under-supported."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real contribution, not a fake controversy. The two-pixel determinant example (Eqs. 6-7) genuinely proves the pixel-sum MT estimator is non-covariant, and the geometric-rotation argument—parallel transport to a common point averages out preferred directions—correctly identifies why the covariant estimator is biased toward isotropy. That alone is worth knowing.\n\nWhat's new: the connected-component correlation functions ξ±(θ,ν), lifted from weak lensing, are a sensible way to get covariance without transport-induced isotropization. The analytic result that dipole modulation enters the traceless MT only at second order in λ (Eq. 48) is a real derivation, not a fit. The simulation campaign is clean: four ensembles, 400 realizations each, and the qualitative distinction between global and local shear is visually and statistically clear. No fitted constants are used; λ, the band limit, θ_G, and n are inputs. The paper also flags its own limitations—frame dependence, look-elsewhere, path dependence, finite-size distortion—which is more than most methods papers do.\n\nThe soft spot is exactly where the stress-test lands. Appendix B quantifies the finite-size transport distortion only for the shape parameter β, on 10 isotropic realizations. But ξ± depends on the orientation angle ψ, not β. A small transport-induced rotation of the principal eigenvector changes ϵ+ and ϵ× at first order, so a 1% shift in β tells you little about the error in the correlation. Further, the pair average weights each pair by |ϵ||ϵ|; the most elongated components dominate, and nothing shows those components are small. And the check is only on isotropic fields—the regime where the global-shear ξ− at large separations is claimed is exactly where anisotropic transport effects could be worst. So 'intrinsic, unbiased' is stronger than what is demonstrated; 'practically covariant up to a quantified-by-β-only distortion' is the defensible claim. This is a real but fixable gap—add an eigenvector-error test on anisotropic ensembles.\n\nAlso: no code or data are shipped yet, which makes independent verification a reimplementation project. And the 'ξ+ decay measures angular coherence' framing is calibrated on two designer shear fields; in the local case the decay scale is partly inherited from the ℓ=4,5 band cuts. Fine as qualitative diagnostics, less fine as a universal measurement claim.\n\nVerdict: send it to a serious referee. It deserves publication after targeted revisions. Anyone working on CMB morphology or sphere-based MTs should read it; I'd bring it to reading group and would cite the covariance/isotropization analysis.","headline":"A serious, useful methods paper with a clean covariance critique and a promising new statistic; the 'unbiased' claim is one appendix short of being fully earned.","tokens_in":29938,"tokens_out":3314,"would_cite":true,"duration_ms":33597,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that the only unbiased, covariant way to test statistical isotropy with Minkowski Tensors on the sphere is to correlate the orientations of individual connected components, and that the resulting ξ± correlation functions s","keywords":["Minkowski tensors","statistical isotropy","connected components","random fields on the sphere","parallel transport","dipole modulation","cosmic microwave background","morphological statistics"],"falsifier":"Compute ξ± for a strongly sheared field twice: once with per-component tensors evaluated after transport to the component centre, and once with the same tensors evaluated with no transport in a fixed frame. If for the largest components the transport-induced change in the ellipticity β is comparable to the difference between the global-shear and local-shear ξ± curves at large separations, the central claim fails; equivalently, a simulation in which the anisotropic signal is dominated by components with areas much larger than the smoothing area would settle it.","tokens_in":28688,"feed_emoji":"🧭","tokens_out":5894,"duration_ms":56948,"temperature":0.7,"pith_summary":"The paper argues that every existing way of extracting Minkowski Tensors from fields on the two-sphere is unsuited to testing statistical isotropy: the direct pixel sum is not a tensor, so its eigenvalue ratio can change with an arbitrary coordinate choice, while the covariant parallel-transport version systematically rotates vectors toward isotropy and washes out the signal. To escape this, the authors construct Minkowski Tensors for each connected component of an excursion set separately, then build pairwise correlation functions ξ± of the components' orientations relative to the great arc joining them. Because only scalars are averaged, the statistic is intrinsically covariant, and the paper shows on simulated fields that it separates globally coherent shear from locally coherent shear, with the decay scale of ξ+ measuring the angular coherence of alignment and the antipodal amplitude of ξ− measuring the globally coherent fraction. The paper also shows analytically that dipole modulation—the anisotropy pattern behind the CMB hemispherical power asymmetry—leaves no detectable orientation signal, because its effect on the traceless part of the tensor is second order in the modulation amplitude.","feed_headline":"Two new correlations expose sky anisotropy from object shapes","feed_subtitle":"Shape-pair correlations are covariant, separate global from local shear, and stay blind to dipole modulation.","key_machinery":"The central object is the connected-component correlation function pair ξ±(θ,ν), built from per-component Minkowski Tensors W^{0,2}_{1,μ}. Each component's tensor is evaluated at its geometric centre after short great-arc transport of boundary normals; its eigenvalues give βμ=Λ2/Λ1 and the principal eigenvector vμ. With ψμ the angle between vμ and the geodesic tangent to the partner component, the spin-2 quantities ϵ+ = −|ϵ|cos2ψ and ϵ× = −|ϵ|sin2ψ (with |ϵ|=(1−β)/(1+β)) are scalars, so their pair averages ξ± are coordinate-independent. The work these objects do is to replace the ambiguous tensor integral on the curved sphere with comparisons of intrinsic shapes at separated points, avoiding","core_discovery":"The paper's central claim is that statistical isotropy on S² can be tested without choosing a frame by turning the Minkowski Tensor from a global integral into a per-object shape measurement. For each connected component of the excursion set at threshold ν, one extracts the principal-axis eigenvector and ellipticity from the component's own boundary, then defines scalar spin-2 quantities ϵ± relative to the geodesic to a partner component; averaging pairs at separation θ gives ξ±(θ,ν), which are coordinate-independent by construction. The paper shows that this statistic is unbiased for isotropic fields—fluctuations scatter symmetrically, unlike the bounded ratio α—and that in sheared fields t","pith_inferences":["The same construction should work for other tensor-valued morphological descriptors: any per-object shape statistic reduced to scalar angles before averaging inherits the covariance property, so ξ±-style estimators could be built for polarization maps or weak-lensing shear fields.","Because ξ± is blind to dipole modulation, a null result from orientation statistics does not bound all anisotropy: it must be paired with band-resolved scalar statistics such as W1 to cover different symmetry breakings of SO(3).","The mirror relation ξ−(θ)≈ξ+(180°−θ) seen for global shear suggests a compact diagnostic: for any coherent alignment with antipodal symmetry, the signal migrates from ξ+ to ξ− as separation grows, and a 45°-rotated 'cross' shear would flip the sign of ξ−—a prediction the paper states but does not yet simulate.","One can calibrate the statistic before cosmology use by applying it to simulated maps with a planted finite-range alignment and checking that the inferred ξ+ decay scale matches the planted coherence length."],"forward_implications":["The one-point eigenvalue ratio α, in either its non-covariant or covariant form, should not be used alone to claim or exclude statistical isotropy on the sphere; the paper demonstrates that a frame rotation can null a real shear signal.","For shear-type anisotropies, the decay scale of ξ+ gives a direct measurement of the angular coherence length of structure alignment, and the large-separation amplitude of ξ− gives the fraction of alignment that is globally coherent, separating models that a single global statistic cannot tell apart.","Dipole modulation of the type studied in CMB power-asymmetry analyses produces no significant ξ± or α signal; the dominant effect appears instead in the trace W1 measured in latitude bands, with a variance modulation linear in the amplitude λ.","Because ξ± are not bounded quantities, finite-area noise scatters them symmetrically about zero for isotropic fields, removing the noise bias that makes ⟨α⟩<1 even in the isotropic case.","The statistic is ready to be applied to CMB temperature maps and two-dimensional galaxy projections as a scale-resolved isotropy test."],"fun_headline_variants":["Shape-pair correlations test sphere isotropy","Covariant correlations expose sky shear patterns","Minkowski tensor correlations reveal scale-dependent alignments","Object-shape statistics probe cosmic anisotropy","Component correlations detect global and local shear"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that short great-arc transport of each component's boundary normals to its geometric centre distorts component shapes negligibly; the paper only quantifies this on ten isotropic realisations, and if large, elongated components in anisotropic fields—precisely those most weighted in ξ±—are significantly isotropized, the claimed discrimination between global and local alignment would blur.","fun_headline_variants_meta":{"raw":{"variants":["Shape-pair correlations test sphere isotropy","Covariant correlations expose sky shear patterns","Minkowski tensor correlations reveal scale-dependent alignments","Object-shape statistics probe cosmic anisotropy","Component correlations detect global and local shear"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000176,"raw_usage":{"total_tokens":1151,"prompt_tokens":794,"completion_tokens":357,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":291}},"tokens_in":538,"tokens_out":357,"duration_ms":4317,"temperature":1.0,"reasoning_tokens":291,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T23:22:23.831129+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute ξ± for a strongly sheared field twice: once with per-component tensors evaluated after transport to the component centre, and once with the same tensors evaluated with no transport in a fixed frame. If for the largest components the transport-induced change in the ellipticity β is comparable to the difference between the global-shear and local-shear ξ± curves at large separations, the central claim fails; equivalently, a simulation in which the anisotropic signal is dominated by components with areas much larger than the smoothing area would settle it.","supporting_citations":[],"review_version":1}