{"id":"a685451a-77d2-451d-b2b8-29c5bd6ecbbc","arxiv_id":"2608.13429","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A fully differentiable neutrino oscillation engine computes exact gradients through layered-Earth geometry and downstream analysis, validated against external codes and finite differences.","lead":"MANGO is a JAX-based engine that computes neutrino oscillation probabilities and exact derivatives of those probabilities with respect to many physical inputs, including detector depth and Earth-shell densities. It is useful because gradients can then flow through the entire analysis chain, from physics to detector response, enabling one-pass sensitivity studies and experiment design.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The advertised singularity-free exactness at degenerate spectra is unsupported: the default divided-difference backend's eigenvalue solve passes through arccos(r)/3, whose derivative diverges as r approaches ±1, and no derivative-level test at exact degeneracy is provided; Section VII.d documents…","rationale":"The reader's weakest_assumption identifies precisely the same load-bearing concern: the exactness claim at degenerate spectra rests on the divided-difference backend, but the paper provides only forward-accuracy evidence there, while the one derivative-level degeneracy test it reports (for the eigh fallback) fails badly. I agree that this is a scoped but genuine gap rather than a collapse of the engine. The forward validation is strong: cross-code agreement at 1e-9 to 1e-5, finite-difference agreement at 1e-10 to 1e-9 for standard parameters, archived code, and the Fisher-matrix derivative cross-check all support the main utility of the package. The issue is that the strongest advertised property, singularity-free exact derivatives at exact level crossings, is not verified at the derivative level for the default 3-flavor path. The manuscript itself flags the arccos derivative divergence and the eigh failure, so the concern is internal to the paper's own evidence rather than an external consensus disagreement. A conditional verdict is the right calibration: the central engine is credible, but the headline exactness claim needs either a derivative-level degeneracy test or an explicit scope limitation.","tokens_in":17452,"tokens_out":3099,"duration_ms":33545,"concrete_test":"Run a derivative-level regression at an exactly degenerate 3-flavor spectrum: set the Hamiltonian so that two eigenvalues coincide (for example, by taking Delta m^2_31 = 0 in normal ordering or by tuning the matter potential to the crossing condition) and compute dP_alpha beta / d theta_13, dP / d Delta m^2_21, and reverse-mode gradients of a scalar loss through the layered-Earth product, comparing against central finite differences with step sizes spanning 1e-3 to 1e-7 and, where possible, against an independent analytic derivative. Require agreement to better than 1e-6 relative, or characterize the divergence rate. If the default backend passes this test, the singularity-free claim is supported; if it fails, the central exactness claim must be scoped to non-degenerate spectra.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest claim in Section IV is that automatic differentiation yields exact derivatives of any output with respect to any input, with Section III item 1 further promising C^infty smooth AD derivatives across exact mass level-crossings. The load-bearing premise is that the default divided-difference backend has a smooth and accurate derivative at exactly degenerate spectra. That premise is not established. The backend obtains eigenvalues via the trigonometric cubic, which 'passes through arccos(r)/3, whose derivative diverges as r approaches ±1'; at exact degeneracy the eigenvalue solve resolves the degenerate pair only to about 5e-7 of the spectral radius. The paper validates AD safety at degeneracy solely by a forward-propagator error of 5e-14 typical against SciPy, plus broad finite-difference agreement away from degeneracy. A forward error of 5e-14 does not bound the derivative error: AD differentiates the implemented approximate function, and near r = ±1 the derivative of that approximation is not the derivative of the true propagator; the diverging arccos derivative can inject large spurious gradient components. The paper itself shows the failure mode in Section VII.d: at an exact 3+1 degeneracy the eigh fallback gives -67.8 versus a finite-difference value of -5.54. For the default 3-flavor path it reports no analogous derivative-level test at exact degeneracy. Thus the C^infty, singularity-free claim is an extrapolation, not a demonstrated property.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces MANGO, a JAX-based neutrino oscillation engine that computes forward probabilities for vacuum, constant-density, layered-PREM, arbitrary-profile, and adiabatic solar propagation, and that provides automatic differentiation of every output with respect to all inputs, including oscillation parameters, geometry, Earth-shell densities, and BSM couplings. The central technical claim is that a divided-difference matrix-exponential backend avoids eigenvector-derivative singularities and thus supplies C^infty smooth, exact derivatives even at degenerate spectra. The paper reports reverse-mode Jacobian costs of 2.5-3x the forward pass, validates probabilities against OscProb and NuFast/NuFast-Earth to 1e-9-1e-5, checks derivatives against finite differences, and demonstrates a full differentiation chain through a binned Poisson Fisher matrix and its inverse for Earth tomography, including derivatives of the marginalized uncertainty with respect to detector angular resolution.","tokens_in":1431,"tokens_out":1344,"duration_ms":88797,"significance":"If the claims hold, MANGO would fill a genuine gap: existing oscillation codes do not provide input-agnostic exact derivatives, and CHIC covers only constant-density three-flavor propagation. The paper's strengths are its broad validation program (external codes, analytic limits, finite differences, and cross-backend consistency), its open-source release with reproducible benchmark scripts and a Zenodo archival DOI, and its demonstration that reverse-mode cost is essentially independent of input dimension, which makes 369-parameter sensitivity calculations practical. The end-to-end Fisher-matrix differentiation and the functional solar-tomography kernel are interesting and potentially useful for gradient-based inference and experimental design. However, the marquee singularity-free differentiability claim at exact spectral degeneracies is not supported by the evidence presented, and one validation-table entry appears to conflict with the text; these issues need to be resolved before the central claim can be accepted.","major_comments":[{"comment":"The advertised C^infty, singularity-free differentiability at exact mass level-crossings is not established for the default divided-difference backend. Section III acknowledges that the trigonometric cubic eigenvalue solve passes through arccos(r)/3, whose derivative diverges as r approaches +/-1, and that an exactly degenerate pair is resolved only to about 5e-7 of the spectral radius. A forward-propagator error of 5e-14 does not bound the derivative error, because AD differentiates the implemented approximate function; near r=+/-1 the derivative of that approximation can contain large spurious components. No AD-versus-finite-difference test at an exactly degenerate three-flavor spectrum is reported, and the only degenerate derivative test reported anywhere, Section VII.d for the eigh fallback in a 3+1 model, fails badly (-67.8 versus -5.54). The authors should supply a derivative-level validation at exact degeneracy for the default backend, ideally with a scan approaching degeneracy, or explicitly restrict the singularity-free claim.","section":"Section III, item 1; Section VII.d; Section IX"},{"comment":"The validation table contains a row reading 'AD vs. FD 3+1 sterile partial P/partial Delta m^2_41 at Delta m^2_41 = Delta m^2_31: 12x(see text)' that is incomplete and appears to contradict Section VII.d, which reports an 'unreliable AD derivative of -67.8' at exactly this degeneracy. Either the metric and numerical value should be stated explicitly, or the row should be removed; as written, the table implies that exact-degeneracy derivatives are validated while the text says the opposite.","section":"Table III"},{"comment":"The unconditional statement in Section IV that automatic differentiation 'yields exact derivatives of any output with respect to any input' and the abstract's claim that every computed quantity is differentiable are too broad in light of the square-root cusps of partial P/partial cos theta_z at shell-grazing angles acknowledged in Section VII.e and the degenerate-spectrum failure in Section VII.d. The central differentiability claims should be stated with explicit measure-zero caveats so that 'exact' is understood to mean 'exact away from the enumerated non-differentiable sets'.","section":"Section IV; Section VII.e"}],"minor_comments":[{"comment":"The text mentions 'Nuwaves' as a validation or comparison tool, but no Nuwaves reference appears in the bibliography; a reference should be added or the name removed.","section":"Sections V and IX; Acknowledgements"},{"comment":"The production weights w_k are used in the expression for F_i(r) before they are defined in the following line; reordering the definitions would improve readability.","section":"Section II.B.e, Eq. (10)"},{"comment":"The one-sided-difference convergence sequence (8e-4, 3e-6, 3e-8, 7e-10) is informative, but the step sizes h=1e-4 through 1e-7 should be shown alongside the values in the caption rather than only in prose after the table.","section":"Table IV caption"},{"comment":"The sentence claiming that the divided-difference backend 'guarantees exact, smooth derivatives across degenerate points' repeats the unsupported central claim; even if the authors choose to retain the claim, it needs a reference to the new derivative-level validation requested above.","section":"Section VII.d"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"MANGO is a real contribution and I'd send it to a serious referee, but the marquee claim needs work. What's genuinely new is not AD itself—CHIC already gives analytic derivatives—but the JAX-native application to layered-Earth geometry, per-shell densities, detector depth, continuous solar profiles, and the inverse Fisher matrix. That is a useful gap, and the paper fills it convincingly. The validation is unusually thorough for software of this type: forward probabilities against OscProb and NuFast at 1e-9 to 1e-5, AD against finite differences at 1e-10 to 1e-6 across standard, geometric, matter, and BSM parameter classes, internal cross-backend consistency, and archived code with CI tests. The end-to-end Fisher-matrix derivative through detector response and inverse covariance, checked against an explicit scan, is a nice demonstration that the pipeline works beyond toy level.\n\nThe soft spot is exactly where the reader put it. The paper promises C^infinity smooth, singularity-free AD derivatives across exact mass level-crossings. That is not established. Section III admits the default backend's trigonometric cubic passes through arccos(r)/3, whose derivative diverges as r approaches ±1, and that the eigenvalue solve resolves degenerate pairs only to ~5e-7 of the spectral radius. The AD safety argument is then based on a forward propagator error of ~1e-14. That does not bound the derivative error: AD differentiates the implemented approximate function, and near r=±1 that derivative is not the true derivative. Section VII.d documents exactly this failure mode for the eigh fallback (AD -67.8 vs FD -5.54), and no analogous derivative-level test is given for the default 3-flavor path at exact degeneracy. So the singularity-free claim is an extrapolation, not a demonstrated property. It's a scoped problem—away from exact degeneracy the AD/FD agreement is excellent, and realistic propagation avoids exact degeneracies—but the paper should either provide the missing test or temper the claim. Also minor: the summary names Nuwaves as a validation target, but the paper gives no citation or validation row for it.\n\nFor a reader doing sensitivity studies, HMC, or design optimization with neutrino oscillations, this is a practical tool worth engaging with. The author clearly knows the literature and the limitations section is honest. I'd recommend peer review, with the expectation that the degeneracy claim gets fixed or softened before publication.","headline":"A genuinely useful, well-validated JAX oscillation engine whose one load-bearing marketing claim—singularity-free exact derivatives at degenerate spectra—is not actually demonstrated.","tokens_in":18241,"tokens_out":1792,"would_cite":true,"duration_ms":21316,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"MANGO makes neutrino oscillation probabilities exactly differentiable with respect to every input—geometry, Earth-shell densities, and BSM couplings—at a constant 2.5–3x forward cost, and carries derivatives through the inverse Fisher…","keywords":["neutrino oscillations","automatic differentiation","matter effects","PREM Earth model","Fisher information","beyond-standard-model searches","experimental design","computational methods"],"falsifier":"Set up an exactly degenerate spectrum in the default divided-difference backend (for example, $\\Delta m^2_{41}=\\Delta m^2_{31}$ for a 3+1 model, or an exactly degenerate three-flavor point) and compare the automatic-differentiation derivative against central finite differences refined over several step sizes; if the error does not converge to the finite-difference truncation level, the smoothness claim is falsified. The paper already reports a concrete failure for its eigenvector fallback at such a point ($-67.8$ versus $-5.54$), so the same test applied to the production backend would settle the question.","tokens_in":1904,"feed_emoji":"⚛️","tokens_out":2117,"duration_ms":93373,"temperature":0.7,"pith_summary":"Neutrino oscillation probabilities are routine to compute, but their exact derivatives are not part of any standard engine. This paper presents MANGO, a neutrino oscillation engine built from automatic-differentiation primitives, and claims that every output is exactly differentiable with respect to every input: the six mixing parameters, trajectory geometry such as zenith angle and detector depth, each of the 369 density, electron-fraction, and shell-radius parameters of a layered PREM Earth, and beyond-standard-model couplings. The practical payoff is that reverse-mode differentiation evaluates all 369 Earth sensitivities in a single pass at only 2.5–3 times the forward cost, and gradients continue downstream through flux weighting, binning, detector response, and even the inverse Fisher matrix. The paper demonstrates this on a stylized Earth-tomography analysis in which the marginalized core-density uncertainty is itself differentiated with respect to detector angular resolution, a design metric that analytic probability formulas cannot produce.","feed_headline":"Neutrino gradients: all 369 Earth-shell densities at 3x cost","feed_subtitle":"A differentiable oscillation engine also differentiates through the inverse Fisher matrix for experiment design.","key_machinery":"The load-bearing object is a divided-difference evaluation of the matrix exponential: instead of diagonalizing with eigenvectors, whose derivatives contain $1/(\\lambda_i-\\lambda_j)$ singularities, the propagator $\\exp(-iHL)$ is expanded via Newton divided differences of the exponential over the eigenvalues of $H$, which is exact in the confluent limit. For three flavors the eigenvalues come from a closed-form trigonometric cubic, and the whole graph is built from automatic-differentiation primitives so forward and reverse modes traverse the same code. The chord-geometry construction maps zenith angle, production height, and detector depth to the segment-length vector in closed form, which is what makes derivatives with respect to geometry and individual Earth shells well defined.","core_discovery":"The central claim is that a composable, JAX-native engine can make neutrino oscillation pipelines differentiable end to end. For a piecewise-constant layered Earth, propagation is an ordered product of segment propagators $\\exp(-iH_k L_k)$, with segment lengths determined by a closed-form chord construction, so all probability derivatives arise from exact automatic differentiation rather than hand-derived formulas. The paper verifies three-flavor and layered-Earth probabilities against external benchmarks to $10^{-9}$–$10^{-5}$ and validates every differentiation path against finite differences, with relative deviations typically $10^{-10}$–$10^{-9}$ and worst case $7\\times10^{-7}$. In the worked example, one forward-mode sweep builds the full Fisher matrix; one more nested automatic-differentiation pass through its inverse yields a design sensitivity of $\\partial\\sigma(\\ln\\rho_{\\mathrm{core}})/\\partial\\sigma_{\\cos\\theta_z}=1.186$ and a curvature of $-3.64$, reproducing an explicit scan to 0.1%.","pith_inferences":["If the degeneracy-smoothness claim survives a derivative-level test, gradient-based sampling (HMC, variational inference) over hundreds of Earth and nuisance parameters becomes practical for global oscillation fits—a route the paper gestures at but does not itself execute.","The same nested differentiation through an inverse Fisher matrix generalizes to optimizing bin edges, detector energy resolution, and exposure allocation for any oscillation measurement, not just core tomography.","Extending the divided-difference construction to arbitrary $N$ flavors would remove the paper's remaining eigenvector fallback, whose derivative fails at exact degeneracy ($-67.8$ versus $-5.54$); until then, $3+N$ sterile fits at level crossings need care.","The square-root cusps at shell-grazing zenith angles imply gradient-based optimizers should clip or smooth near critical $\\cos\\theta_z$; the paper exposes a function for those critical angles, so the practical impact can be tested directly."],"forward_implications":["At 2.5–3x forward cost, a single reverse pass returns gradients for all six standard oscillation parameters, 42 geometry parameters, and 369 layered-Earth matter parameters, where central finite differencing would need 738 forward evaluations.","Analysis objectives—expected counts, likelihoods, and Fisher information—remain differentiable past the probability stage, since flux weighting, binning, and detector migration are themselves automatic-differentiation functions.","Beyond-standard-model sensitivities (NSI $\\varepsilon$, sterile angles, Lindblad decoherence $\\gamma$, and non-unitary $\\alpha$) are obtained exactly at the Standard-Model point, where limit-setting fits linearize, without manual derivation.","Differentiating through the inverse Fisher matrix gives experimental-design derivatives such as $\\partial\\sigma/\\partial\\sigma_{\\cos\\theta_z}=1.186$, with second curvature $-3.64$, enabling detector-design optimization in a single pass.","Validation matches external codes to $10^{-9}$–$10^{-5}$ in probability and finite differences to $10^{-10}$–$10^{-5}$ in derivatives, so the engine can serve as a drop-in forward and sensitivity calculator."],"supporting_citations":[{"why":"Supplies the automatic-differentiation, jit, vmap, and forward/reverse-mode primitives on which every derivative claim in the paper rests.","marker":"[16]"},{"why":"The NuFast-LBL constant-density algorithm that MANGO transcribes into automatic-differentiation code as its fast baseline and cross-check.","marker":"[5]"},{"why":"The closed-form three-flavor constant-density derivative engine whose reach MANGO extends to arbitrary profiles and geometries.","marker":"[15]"},{"why":"The PREM shell model that defines the layered Earth and its 369 differentiable density, electron-fraction, and radius parameters.","marker":"[22]"},{"why":"External reference engine used to benchmark constant-density and identical-PREM-path probabilities.","marker":"[2]"},{"why":"Layered-Earth reference implementation used in the converged PREM cross-code comparison.","marker":"[25]"},{"why":"Reference eigensolver used to measure propagator error at exactly degenerate spectra, the stress point for derivative smoothness.","marker":"[28]"}],"fun_headline_variants":["Autodiff neutrino engine: 369 Earth-shell densities at 3x cost","Differentiable oscillation engine: gradients through Earth's layers","MANGO: autodiff for oscillations, gradients flow to detector design","Neutrino gradients at 3x cost: differentiate through Earth's shell densities","All Earth-shell densities differentiable: 3x cost, full Fisher sensitivity"],"cache_read_input_tokens":20224,"weakest_assumption_plain":"The entire exact-gradient program depends on the eigenvalue solve returning a true derivative when two neutrino masses become exactly equal; the paper verifies the forward probability there but does not report a derivative-level test for its default backend, and its fallback backend demonstrably gives the wrong derivative at an exact degeneracy.","fun_headline_variants_meta":{"raw":{"variants":["Autodiff neutrino engine: 369 Earth-shell densities at 3x cost","Differentiable oscillation engine: gradients through Earth's layers","MANGO: autodiff for oscillations, gradients flow to detector design","Neutrino gradients at 3x cost: differentiate through Earth's shell densities","All Earth-shell densities differentiable: 3x cost, full Fisher sensitivity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000588,"raw_usage":{"total_tokens":2817,"prompt_tokens":1061,"completion_tokens":1756,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":677,"completion_tokens_details":{"reasoning_tokens":1660}},"tokens_in":677,"tokens_out":1756,"duration_ms":12902,"temperature":1.0,"reasoning_tokens":1660,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:14:18.529447+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Set up an exactly degenerate spectrum in the default divided-difference backend (for example, $\\Delta m^2_{41}=\\Delta m^2_{31}$ for a 3+1 model, or an exactly degenerate three-flavor point) and compare the automatic-differentiation derivative against central finite differences refined over several step sizes; if the error does not converge to the finite-difference truncation level, the smoothness claim is falsified. The paper already reports a concrete failure for its eigenvector fallback at such a point ($-67.8$ versus $-5.54$), so the same test applied to the production backend would settle the question.","supporting_citations":[{"cited_title":"com/jax-ml/jax","cited_arxiv_id":null,"evidence_quote":"Supplies the automatic-differentiation, jit, vmap, and forward/reverse-mode primitives on which every derivative claim in the paper rests."},{"cited_title":"Fast and Accurate Algorithm for Calculating Long-Baseline Neutrino Oscillation Probabilities with Matter Effects: NuFast","cited_arxiv_id":"2405.02400","evidence_quote":"The NuFast-LBL constant-density algorithm that MANGO transcribes into automatic-differentiation code as its fast baseline and cross-check."},{"cited_title":"Differentiable Simulation of a Liquid Argon Time Projection Chamber","cited_arxiv_id":"2309.04639","evidence_quote":"The closed-form three-flavor constant-density derivative engine whose reach MANGO extends to arbitrary profiles and geometries."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"External reference engine used to benchmark constant-density and identical-PREM-path probabilities."},{"cited_title":"Methods17261","cited_arxiv_id":null,"evidence_quote":"Reference eigensolver used to measure propagator error at exactly degenerate spectra, the stress point for derivative smoothness."}],"review_version":1}