Pith. sign in

REVIEW 5 major objections 7 minor 24 references

Unexplored Opportunities for Automatic Differentiation in Astrophysics

T0 review · 5 major / 7 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Nine astrophysical domains—from warp metrics to asteroseismology—have the right mathematical structure for gradient-based discovery, and the proposed GRASP framework supplies the blueprint.

desk verdict A useful but overclaimed survey: the 'unexplored' and 100–1000× speedup claims do not survive contact with the paper's own references, and one core formula is wrong. read the letter →

arxiv 2507.09379 v1 pith:DE5Q65PC submitted 2025-07-12 astro-ph.IM physics.comp-ph

Marc Bara This is my paper · ORCID
classification astro-ph.IMphysics.comp-ph
keywords automaticdifferentiationgradient-basedoptimizationdiscoveryframeworkGRASPastrophysicsinverseproblemsParetofrontierscientificmachinelearning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that automatic differentiation, already successful in gravitational-wave and exoplanet work, has a much larger untouched territory in astrophysics. It identifies nine domains where the forward models are smooth and differentiable in principle, yet gradient-based exploration has not, to the author's knowledge, been applied. For each domain it supplies loss functions, constraints, and the mathematical route to gradients—adjoint methods, the Hellmann-Feynman theorem, Wirtinger calculus, and implicit differentiation—and ties them together under GRASP, a conceptual framework rather than a codebase. If the claim is right, these problems stop being black-box optimization targets and become landscapes that multi-start gradient descent can map systematically, with 100–1000× fewer forward evaluations per update.

What carries the argument

The object that carries the argument is GRASP, 'Gradient-based Reconstruction of Astrophysical Systems & Phenomena,' defined as a conceptual framework and set of design patterns for making astrophysical computations differentiable. The actual mathematical mechanism is reverse-mode automatic differentiation: for a scalar loss $L(\theta)$, one forward pass plus one backward pass yields the exact gradient $\nabla_\theta L$ at a cost comparable to two or three forward solves, regardless of the dimension of $\theta$. Around that core, the paper assembles a reusable toolkit: Wirtinger derivatives $\partial f/\partial z$ and $\partial f/\partial z^*$ for complex-valued radio measurements; adjoint equations for ODE/PDE constraints; the Hellmann-Feynman identity $\partial\omega_{n\ell}^2/\partial\theta_i = \langle\xi_{n\ell}|\partial L/\partial\theta_i|\xi_{n\ell}\rangle$ for eigenvalue problems; implicit differentiation for fixed points like Kepler's equation; and neural or basis-function parameterizations of unknown fields. GRASP's role is to show that all nine domains reduce to these same primitives, so a single set of design principles governs the exploration.

What would settle it

Run the paper's proposed differentiable stellar-pulsation inversion on a benchmark star with synthetic frequencies, starting from many random initializations; if gradient descent repeatedly fails to recover structures within the frequency error bars, or if gradient magnitudes through the eigenvalue solver vary erratically across seeds, the completeness claim for asteroseismology would be refuted.

Watch

Extended reading notes

Core claim

The paper's central claim is that automatic differentiation—exact gradients computed through computational graphs in one forward and one backward pass—can be transplanted into nine astrophysical domains that, to the author's knowledge, have never been explored this way, even though their forward models are built from differentiable operations. Those domains are Alcubierre warp-metric optimization, cosmic-string network evolution, domain-wall dynamics, false-vacuum decay and bubble nucleation, interferometric calibration, black-hole image reconstruction, asteroseismic inversion, radial-velocity exoplanet characterization, and asteroid Yarkovsky thermophysical characterization. For each, the paper states the discovery problem, writes down the loss and constraints, and gives the mathematical route to the gradient—adjoint sensitivity for ODEs and PDEs, the Hellmann-Feynman theorem for eigenfrequencies, Wirtinger calculus for complex visibilities, and implicit differentiation for fixed-point problems such as Kepler's equation. These routes are assembled under GRASP, a unified set of design patterns rather than a single codebase, and the paper asserts that the mathematical foundations are complete, with implementation of the nine applications left as the stated future direction.

Load-bearing premise

The framework assumes that the forward models in all nine domains are smooth enough for automatic differentiation through their numerical solvers to give stable, accurate gradients, including near the acknowledged singularities such as black-hole horizons, cosmic-string cusps, and stellar convective boundaries.

Editorial extensions

If this is right

  • If the claim is correct, population-based searches in these nine domains can be replaced by gradient descent with exact gradients at roughly two or three forward solves per update, a 100–1000× reduction in per-iteration cost.
  • Multi-start gradient descent and Pareto-frontier exploration become affordable, so fitting one best model gives way to mapping entire families of physically valid solutions.
  • The simplest domains—interferometric calibration and asteroseismic inversion—are the natural first proof-of-concept implementations, with the harder general-relativistic and black-hole cases following once the patterns are validated.
  • In the general-relativistic domains, neural-network parameterizations of unknown fields would let optimization search over spacetime geometries and bounce profiles that human-constructed analytic forms never encode.
  • Jointly differentiating both model parameters and regularization weights, as sketched for black-hole imaging, would make hyperparameter selection part of the optimization instead of manual tuning.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper: the 'unexplored' list is an author-knowledge claim; a systematic audit of recent differentiable-physics work in each domain could re-rank the nine or remove some, without changing the core argument that the remaining domains are ripe.
  • Beyond the paper: the 100–1000× speedup depends on backward passes through stiff or discontinuous solvers costing about the same as one or two forward passes; a targeted benchmark on the asteroid heat-diffusion or stellar-structure ODEs would show whether memory and non-smoothness eat into that factor.
  • Beyond the paper: the same design patterns extend naturally to other smooth-forward-model problems, such as cosmic-ray propagation, radiative-transfer inversions, or magnetohydrodynamic closure fitting, which the paper does not list.
  • Beyond the paper: if multi-start gradient exploration becomes standard, the practical bottleneck shifts from finding solutions to selecting among the many constraint-satisfying solutions the optimizer would produce.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 7 minor

Summary. The paper argues that automatic differentiation (AD) can be applied to nine astrophysical domains (Alcubierre metrics, cosmic strings, domain walls, false vacuum decay, interferometric calibration, black hole image reconstruction, asteroseismology, radial velocity exoplanets, and asteroid thermophysical characterization) that, to the authors' knowledge, remain unexplored by gradient-based methods. It proposes GRASP, a conceptual framework for differentiable astrophysics, and claims that AD yields 100–1000× reductions in computational cost per iteration compared with population-based approaches. For each domain the paper provides a mathematical formulation, algorithms, and numerical considerations, but it contains no implementation, benchmark, or experimental validation.

Significance. If the central claims were substantiated, the paper would serve as a useful roadmap for applying AD to a broad set of astrophysical problems. The authors correctly identify several technique classes (Wirtinger calculus, implicit differentiation of Kepler's equation, adjoint sensitivity) that are valuable in these settings, and the GRASP idea of codifying common differentiable operations could be a useful organizing principle. However, the paper's significance is critically undermined by the absence of evidence for its two empirical claims—that the nine domains are unexplored and that speedups of 100–1000× per iteration follow—and by formal errors in the mathematical foundations that are presented as complete. The paper also fails to engage with prior work that it itself cites as already performing gradient-based optimization in closely related settings, which directly contradicts the novelty assertion.

major comments (5)
  1. [Sections 1.4 and 3] The claim that the nine listed domains "remain unexplored" is internally contradicted by the paper's own references. Citation [25] describes "exploratory optimisation of wormhole geometries," which is the same class of problem as the Alcubierre metric optimization in Section 3.1.1, and citations [22]–[24] already report gradient-based astrophysical inference in related settings. The paper must either explain precisely how these works differ from its proposed "systematic exploration of solution spaces" or substantially soften the novelty claim. As written, the central thesis is not supported.
  2. [Section 3.3.1, Eq. (91)] The Hellmann–Feynman formula as written in Eq. (91) is incorrect for the generalized eigenvalue problem Lξ = ω²Mξ with a density-dependent weight M. Differentiating the normalization ⟨ξ|Mξ⟩=1 yields the additional term −ω²⟨ξ|∂M/∂θ_i|ξ⟩. The omitted term is not generally negligible for stellar oscillation problems, so the proposed gradient formula for asteroseismic frequencies is formally wrong and the claim that the mathematical foundations are complete is thereby undermined.
  3. [Sections 3.1.1, 3.1.2, 3.1.4, and 4.5] Singular terms are regularized with ad hoc small constants, for example in Eqs. (18), (28), and (48), but the paper provides no analysis or numerical test showing that ∇_θ L_ε ≈ ∇_θ L or that the regularized gradients are reliable. Section 4.5 itself concedes that "computational singularities" and "extreme parameter regimes" remain unresolved. Without validation, the assertion that AD will provide accurate gradients for these problems is unsupported.
  4. [Sections 1.2 and 4.4] The claimed 100–1000× reduction in computational cost per iteration is an unsupported assertion. It is based on a comparison to "genetic algorithms" and "population methods" that is not representative of state-of-the-art algorithms used in the cited astrophysical domains, and it ignores the memory and compute overhead of reverse-mode AD through ODE/PDE solvers. No baseline experiments or benchmarks are provided, yet this speedup is load-bearing for the paper's "discovery" narrative.
  5. [Section 3.2.1, Eqs. (62)–(63)] The Wirtinger derivatives for the calibration loss have incorrect signs. For L = ∑ |V_obs − g_i g*_j V_model|²/σ², the derivative with respect to g_i (treating g*_i as independent) should be −g*_j V_model (V_obs − g_i g*_j V_model)* / σ²; the plus sign in Eq. (62) would drive the parameters in the wrong direction. This is a concrete error in one of the nine proposed differentiable formulations.
minor comments (7)
  1. [Section 3.1.3] The entire domain wall subsection is duplicated: Eqs. (30)–(36) and Eqs. (37)–(43) are identical. Please remove the duplicate.
  2. [Section 3.1.4] "Forward solution of the ODE (Eq. 2)" should refer to the bounce equation in Eq. (45), not to Eq. (2).
  3. [Section 3.1.2] "The equations of motion derived from (24)" should reference Eq. (22), the Nambu-Goto action, not Eq. (24).
  4. [Section 3.1.1] The algorithm and the numerical-considerations text are inconsistent about the spatial grid: Eq. (16) proposes logarithmic radial spacing, but the algorithm uses a Cartesian uniform grid with r_s = sqrt(x_i^2 + y_j^2 + z_k^2). The paper should clarify the actual discretization and volume weights.
  5. [Section 3.1.1, Eq. (10)] The expression for T^00 assumes a static bubble; the paper should state more prominently that the optimization concerns static snapshots only, since the discovery claim is about spacetime geometries in general.
  6. [Data and Software Availability] The section promises example implementations that "will be made available," but no code or demonstration is provided. For a paper whose central claims are computational, this is a significant limitation.
  7. [Abstract and Section 4.4] The abstract cites a "4× computational speedup" from GRAF but does not state the baseline or methodology behind this number, making the connection to the present paper's claims difficult to evaluate.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper offers proposed differentiable formulations, not predictions derived from fitted inputs, and its sole self-citation is motivational.

full rationale

The paper contains no derivation chain in which an output is equivalent to an input by construction. Its central claims are (i) that nine astrophysical domains are unexplored for gradient-based discovery and (ii) that one can write down differentiable formulations for each. These are programmatic proposals and mathematical specifications, not empirical predictions fitted to data. The only self-citation, reference [4] (GRAF), appears in the abstract and introduction as motivation for extending a discovery-oriented approach to astrophysics; no equation, theorem, or domain-specific gradient formula depends on GRAF for its validity, so the self-citation is not load-bearing. The paper explicitly acknowledges unresolved practical issues in Section 4.5, including singularities and extreme parameter regimes, and states that 'the immediate future work is clear: implement these nine applications'; this is a limitation of the completeness claim, not evidence of circularity. The Hellmann-Feynman formula in Eq. (91) may be incomplete for the generalized eigenproblem, but that is a mathematical correctness issue, not a case of the paper deriving its conclusion from its own assumption. Accordingly, the appropriate circularity score is 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The central claims depend on unverified smoothness, an unsupported literature assertion, and hand-chosen hyperparameters. The mathematical components are standard, but the paper adds no independent evidence that the proposed discovery loops converge or that the nine domains are actually unexplored.

free parameters (4)
  • Penalty weights lambda1, lambda2, lambda3 for boundary and constraint enforcement = suggested ranges, e.g., 10^2 to 10^4
    Introduced by hand in multiple algorithms such as Alcubierre metric, bubble nucleation, and RV fitting to enforce constraints; no data or convergence analysis is provided.
  • Surface correction parameters a0, a1 = unspecified
    Added in asteroseismic inversion (Eq. 97) to correct near-surface frequency errors; these are fitted coefficients in the proposed scheme.
  • Regularization hyperparameters lambda_k for imaging and inversion = unspecified
    Total variation, entropy, ring prior, and smoothness weights in black hole imaging and inversion are chosen by hand.
  • Regularization epsilon for singular denominators = 10^-4 to 10^-8
    Added to denominators to regularize r approaching zero and cusp singularities; the values are arbitrary and untested.
assumptions (5)
  • standard math Reverse-mode automatic differentiation computes exact gradients at roughly one to two forward-pass cost
    Assumed in Sections 1.2 and 4.4 as the basis for all speedup claims; true in the ideal setting, but cost in practice depends on memory and solver structure.
  • domain assumption The physical forward models are differentiable and their numerical solvers support stable backpropagation
    Assumed for all nine domains, e.g., differentiating through the stellar pulsation eigenvalue problem (Section 3.3.1) and geodesic ray tracing (Section 3.2.2); not demonstrated.
  • ad hoc to paper The nine listed domains are unexplored by gradient-based methods
    Asserted 'to our knowledge' in the abstract and Section 3; the paper cites AD in wormhole optimization [25], pulsar timing [22], FRB dispersion [23], and Zeeman-Doppler imaging [24], weakening the claim.
  • ad hoc to paper Hand-tuned regularizers and stated epsilon values render singular solutions tractable
    Proposed for r approaching zero and cusps in Sections 3.1.1, 3.1.2, and 3.1.4 without convergence or stability analysis.
  • domain assumption Static snapshot approximation is valid for the Alcubierre metric optimization
    Section 3.1.1 sets the bubble velocity time-derivative to zero so the stress-energy expression omits moving-bubble terms; only static configurations are explored.
invented entities (1)
  • GRASP (Gradient-based Reconstruction of Astrophysical Systems and Phenomena)
    purpose: Conceptual framework and design principles for making astrophysical computations differentiable across nine domains
    Proposed in the abstract and Section 4.3 as a framework, but not implemented; no code, API, or numerical demonstration exists in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unexplored Opportunities for Automatic Differentiation in Astrophysics." pith.science (2026). https://pith.science/paper/DE5Q65PC

@misc{pith2026250709379,
  author       = {Pith},
  title        = {Pith review of: Unexplored Opportunities for Automatic Differentiation in Astrophysics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DE5Q65PC}},
  note         = {Machine review of arXiv:2507.09379}
}
read the original abstract

We present a systematic analysis of automatic differentiation (AD) applications in astrophysics, identifying domains where gradient-based optimization could provide significant computational advantages. Building on our previous work with GRAF (Gradient-based Radar Ambiguity Functions), which discovered optimal radar waveforms achieving 4x computational speedup by exploring the trade-off space between conflicting objectives, we extend this discovery-oriented approach to astrophysical parameter spaces. While AD has been successfully implemented in several areas including gravitational wave parameter estimation and exoplanet atmospheric retrieval, we identify nine astrophysical domains where, to our knowledge, gradient-based exploration methods remain unexplored despite favorable mathematical structure. These opportunities range from discovering novel solutions to the Einstein field equations in exotic spacetime configurations to systematically exploring parameter spaces in stellar astrophysics and planetary dynamics. We present the mathematical foundations for implementing AD in each domain and propose GRASP (Gradient-based Reconstruction of Astrophysical Systems & Phenomena), a unified framework for differentiable astrophysical computations that transforms traditional optimization problems into systematic exploration of solution spaces. To our knowledge, this is the first work to systematically delineate unexplored domains in astrophysics suitable for automatic differentiation and to provide a unified, mathematically grounded framework (GRASP) to guide their implementation.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 20 canonical work pages

  1. [22]

    G. E. Freedman, Low-Frequency Gravitational Wave Searches and Data Analysis with Hamiltonian Sampling , Ph.D. Dissertation, University of Wisconsin-Milwaukee, 2025

    work page 2025

  2. [25]

    Singh, J

    G. Singh, J. Kaur, G. Singh, H. Singh, J. Gill, R. Singh, and M. Singh, “Artificial Intelligence & Quantum Computing for Causal Loop and Chronology Violation Detection: A Deep Learning Approach to Predicting Traversable Wormhole Stability and Temporal Paradoxes,” International Journal of Innovative Scientific Re- search, vol. 3, no. 2, pp. 1–13, 2025. 25

    work page 2025

  3. [24]

    Bayesian Uncertainty Quantification in High-dimensional Stellar Magnetic Field Models,

    Jennifer R. Andersson, Oleg Kochukhov, Zheng Zhao, and Jens Sj¨olund, “Bayesian Uncertainty Quantification in High-dimensional Stellar Magnetic Field Models,”ICML 2023 Workshop on Machine Learning for Astrophysics, PMLR 202, 2023

    work page 2023

  4. [1]

    PyTorch: An Imperative Style, High-Performance Deep Learning Library

    Paszke, A., et al., “PyTorch: An Imperative Style, High-Performance Deep Learning Library”, Advances in Neural Information Processing Systems, vol. 32, pp. 8024-8035, 2019

    work page 2019

  5. [2]

    JAX: composable transformations of Python+NumPy programs

    Bradbury, J., et al., “JAX: composable transformations of Python+NumPy programs”, 2018. [Online]. Available: http://github.com/google/jax

    work page 2018

  6. [3]

    TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems

    Abadi, M., et al., “TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems”, 2015. Software available from tensorflow.org

    work page 2015

  7. [4]

    Differentiable Radar Ambiguity Functions: Mathematical Formulation and Computational Implementation

    M. Bara, “Differentiable Radar Ambiguity Functions: Mathematical Formulation and Computational Implemen- tation”, arXiv preprint arXiv:2506.22935, 2025

    work page Pith review arXiv 2025

  8. [5]

    Searching For Black Holes Using Auto Differentiation

    W. Doyle, “Searching For Black Holes Using Auto Differentiation”, Emerging Minds Journal for Student Re- search, vol. 1, pp. 17-38, 2023. 24 Unexplored Opportunities for Automatic Differentiation in Astrophysics A PREPRINT

    work page 2023

Show all 24 references
  1. [6]

    ripple: Differentiable and Hardware-Accelerated Waveforms for Gravitational Wave Data Analysis

    T. D. P. Edwards, K. W. K. Wong, K. Jani, T. Wouters, S. M. Sch ¨afer, and J. Gair, “ripple: Differentiable and Hardware-Accelerated Waveforms for Gravitational Wave Data Analysis”, arXiv preprint arXiv:2302.05329 , 2023

    2023 arXiv

  2. [7]

    Differentiable Strong Lensing: Uniting Gravity and Neural Nets through Differentiable Probabilistic Programming

    M. Chianese, A. Coogan, P. Hofma, S. Otten, and C. Weniger, “Differentiable Strong Lensing: Uniting Gravity and Neural Nets through Differentiable Probabilistic Programming”, Monthly Notices of the Royal Astronomical Society, vol. 496, no. 1, pp. 381-393, 2020

    2020

  3. [8]

    JAX-COSMO: An End-to-End Differentiable and GPU Accelerated Cosmology Li- brary

    J.-E. Campagne, et al., “JAX-COSMO: An End-to-End Differentiable and GPU Accelerated Cosmology Li- brary”, arXiv preprint arXiv:2302.05163, 2023

    2023 arXiv

  4. [9]

    Au- todifferentiable Spectrum Model for High-dispersion Characterization of Exoplanets and Brown Dwarfs

    H. Kawahara, Y . Kawashima, K. Masuda, I. J. M. Crossfield, E. Pannier, and D. C. van den Bekerom, “Au- todifferentiable Spectrum Model for High-dispersion Characterization of Exoplanets and Brown Dwarfs”, The Astrophysical Journal Supplement Series, vol. 258, no. 2, p. 31, 2022

    2022

  5. [10]

    exoplanet- dev/jaxoplanet: Astronomical time series analysis with JAX (v0.0.2)

    S. Hattori, L. Garcia, C. Murray, J. Dong, S. Dholakia, D. Degen, and D. Foreman-Mackey, “exoplanet- dev/jaxoplanet: Astronomical time series analysis with JAX (v0.0.2)”, Zenodo, 2024

    2024

  6. [11]

    The warp drive: hyper-fast travel within general relativity

    M. Alcubierre, “The warp drive: hyper-fast travel within general relativity”, Classical and Quantum Gravity , vol. 11, no. 5, pp. L73-L77, 1994

    1994

  7. [13]

    A review of gravitational waves from cosmic domain walls

    K. Saikawa, “A review of gravitational waves from cosmic domain walls”, arXiv preprint arXiv:1703.02576 , 2017

    2017 arXiv

  8. [14]

    Fate of the false vacuum. I. Semiclassical theory

    S. Coleman, “Fate of the false vacuum. I. Semiclassical theory”, Physical Review D, vol. 15, no. 10, pp. 2929- 2936, 1977

    1977

  9. [15]

    A. R. Thompson, J. M. Moran, and G. W. Swenson Jr., Interferometry and Synthesis in Radio Astronomy, 3rd ed. Springer, 2017

    2017

  10. [16]

    First M87 Event Horizon Telescope Results. I. The Shadow of the Supermassive Black Hole

    The Event Horizon Telescope Collaboration, “First M87 Event Horizon Telescope Results. I. The Shadow of the Supermassive Black Hole”, The Astrophysical Journal Letters, vol. 875, p. L1, 2019

    2019

  11. [17]

    Aerts, J

    C. Aerts, J. Christensen-Dalsgaard, and D. W. Kurtz, Asteroseismology. Springer, 2010

    2010

  12. [18]

    A Jupiter-mass companion to a solar-type star

    M. Mayor and D. Queloz, “A Jupiter-mass companion to a solar-type star”, Nature, vol. 378, pp. 355-359, 1995

    1995

  13. [19]

    The Yarkovsky and YORP Effects: Implications for Asteroid Dynamics

    W. F. Bottke Jr., D. V okrouhlick ´y, D. P. Rubincam, and D. Nesvorn ´y, “The Yarkovsky and YORP Effects: Implications for Asteroid Dynamics”, Annual Review of Earth and Planetary Sciences , vol. 34, pp. 157-191, 2006

    2006

  14. [20]

    Vilenkin and E

    A. Vilenkin and E. P. S. Shellard, Cosmic Strings and Other Topological Defects. Cambridge University Press, 2000

    2000

  15. [21]

    Cosmic Mi- crowave Background constraints for global strings and global monopoles,

    A. Lopez-Eiguren, J. Urrestilla, M. Hindmarsh, J. C. Bueno Sanchez, D. Daverio, and M. Kunz, “Cosmic Mi- crowave Background constraints for global strings and global monopoles,” Journal of Cosmology and Astropar- ticle Physics, vol. 2017, no. 07, p. 026, 2017

    2017

  16. [23]

    Comprehensive Bayesian analysis of FRB-like bursts from SGR 1935+2154 observed by CHIME/FRB,

    Utkarsh Giri, Bridget C. Andersen, Pragya Chawla, Alice P. Curtin, Emmanuel Fonseca, Victoria M. Kaspi, Hsiu-Hsien Lin, Kiyoshi W. Masui, Ketan R. Sand, Paul Scholz, Thomas C. Abbott, Fengqiu Adam Dong, B. M. Gaensler, Calvin Leung, Daniele Michilli, Mohit Bhardwaj, Moritz M ¨...

    1935 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.