Pith. sign in

REVIEW 4 major objections 6 minor 93 references

The 6D phase-space distribution of 5,000 halo tracers carries 2.5-9.9 times more information about the Milky Way-LMC masses and halo shape than standard all-sky velocity moments, and a joint 19-dimensional summary of basis-function coeffici

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 11:03 UTC pith:34VHYLSQ

load-bearing objection First real benchmark of information content in MW–LMC 6D phase space, with solid validation and an honest anisotropy stress test; headline ratios are approximate until CFM calibration is pinned down. the 4 major comments →

arxiv 2607.20006 v1 pith:34VHYLSQ submitted 2026-07-22 astro-ph.GA

LMC-induced Perturbations in the Milky Way Halo II: Bridging Field-level Inference and Summary-level Simulation-Based Inference

Yanjun Sheng , Yuan-Sen Ting , Tomasz R\'o\.za\'nski , Xiang-Xiang Xue , Roland M. Crocker , Jiashu Pan This is my paper
classification astro-ph.GA
keywords Milky Way haloLarge Magellanic Cloudphase-space inferencesimulation-based inferencebasis function expansionMOPED compressionvelocity momentsflow matching
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper asks how much information about the Milky Way and Large Magellanic Cloud is contained in the full six-dimensional phase-space positions of halo stars, compared with the conventional summaries used by astronomers. It shows, at one fiducial simulation, that the raw particle dataset tightens constraints on the MW mass, LMC mass, halo concentration, and flattening by factors of 2.5 to 9.9 compared with all-sky velocity moments. It then shows that a physically interpretable compression of the phase space into multipole basis-function coefficients, further compressed into four summary statistics, recovers part of that gain, and that combining these with velocity moments closes most of the remaining gap while remaining interpretable. The central discovery is an information hierarchy: field-level likelihood > joint summary > BFE+MOPED alone > velocity moments, with concrete factors quantifying each step. This matters because future surveys can decide how much modelling effort is worth investing.

Core claim

The paper trains a conditional flow matching model on 90% of an N-body simulation suite to evaluate the exact likelihood of a 5,000-particle 6D phase-space catalogue at a held-out parameter point, and uses it as a field-level benchmark. Against this benchmark, the 15-component all-sky velocity-moment summary is 2.5-9.9 times less constraining for the four parameters (MW mass, LMC mass, concentration, flattening). A four-dimensional MOPED compression of a 10-channel basis-function expansion (density plus three momentum and six dispersion fields) sits between the benchmarks, and a 19-dimensional vector combining these summaries with velocity moments reaches within a factor of 1.3-2.9 of the fi

What carries the argument

Three objects carry the argument. A conditional flow matching (CFM) generative model, trained on 90% of the simulation suite, provides a tractable likelihood for individual 6D phase-space particles and therefore a field-level information benchmark. A biorthogonal basis function expansion (BFE) projects the density field and the first and second velocity-moment fields onto spherical-harmonic and radial eigenfunctions, separating the reflex dipole, the wake, and the halo-shape quadrupole into interpretable channels. The MOPED algorithm then compresses the 10,800 active BFE coefficients into four linear summary statistics, one per model parameter, preserving the Fisher information of the origin

Load-bearing premise

The neural likelihood model, trained on 90% of the simulation suite and checked at a single held-out parameter point, is accurate enough to serve as the reference for how much information the raw phase-space data contains; if it is miscalibrated at other points, the quoted information ratios and the MOPED summary directions derived from its gradients shift.

What would settle it

Compute the neural-likelihood posterior for, say, five additional held-out simulations spread across the parameter box and check marginal and joint coverage; if the coverage deviates from nominal at any of these points, the field-level benchmark and the 2.5-9.9x information ratios used to rank summaries are not trustworthy. Alternatively, at one off-fiducial point, replace the neural-derived score gradients with explicit N-body finite differences and see whether the MOPED compression and the joint-summary gains survive.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If the field-level benchmark holds, any summary-based analysis that uses only all-sky velocity moments is leaving a factor of 2.5-9.9 in parameter constraining power unused.
  • A joint 19-dimensional summary of BFE+MOPED and velocity moments provides most of the practical gain while remaining interpretable, reaching within a factor of 1.3-2.9 of the information limit.
  • The mutual-information result implies that angular multipole information and radial-bin moments are complementary; combining them is always worthwhile, not a matter of taste.
  • Because the MW-mass and concentration channels are driven by velocity-dispersion monopoles, inference from real, radially biased tracer populations will require careful treatment of anisotropy, or the mass and concentration estimates will shift.
  • The BFE channel decomposition gives a diagnostic route: if a posterior shift is driven by a specific physical channel (e.g., second-moment velocities), that channel can be censored or modelled better.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The single-fiducial calibration leaves open the possibility that the field-level benchmark is optimistic elsewhere; a straightforward extension would test coverage at several additional held-out simulation points before trusting the 2.5-9.9x ratios in survey analyses.
  • If this information hierarchy persists at other parameter points, the practical implication is that future halo surveys should consider collecting enough tracers to reach the joint-summary limit before investing in full field-level inference.
  • The strong anisotropy sensitivity of the MW-mass and concentration channels suggests that adding a tracer-anisotropy nuisance parameter to the summary emulator may be more impactful than refining the density or velocity expansion.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper presents a field-level likelihood benchmark for the constraining power of 6D phase-space tracer data on the MW-LMC parameter set (M_MW, M_LMC, c, q), based on a Conditional Flow Matching (CFM) generative model trained on the HaloDance N-body suite. Against this benchmark, the authors compare all-sky velocity-moment summaries and a BFE+MOPED linear compression, and then combine the two summary sets into a 19-dimensional vector. The headline claims are: (i) the CFM field-level posterior is tighter than a velocity-moment Fisher forecast by factors of 2.5-9.9 (Table 1); (ii) the joint 19D summary vector comes within a factor of 1.3-2.9 of the field-level benchmark (Table 2); and (iii) the BFE+MOPED summaries are physically interpretable and complementary to moments. The summary-level pipeline is validated extensively on 185 held-out parameter grids with P-P, TARP, and anisotropy stress tests.

Significance. If the quantitative claims survive scrutiny, the paper makes a useful contribution: it supplies an information bound for the MW-LMC problem and demonstrates that a modest number of interpretable summaries can retain much of the field-level information. The methodological structure—field-level benchmark plus summary-level SBI—is timely and the validation effort is above the norm for this literature. The anisotropy stress test is a particularly honest and valuable systematic check. The central numerical claims, however, rest on the CFM likelihood being a well-calibrated conditional density, and the same CFM is also used to build the MOPED summaries. Because no simulation-based calibration of the CFM is reported, the quantitative ratios in Tables 1-2 are not yet fully supported.

major comments (4)
  1. [Section 2.3, Table 1] The field-level benchmark is validated only at a single held-out fiducial point (Fig. 3). No P-P, TARP, or equivalent coverage test is reported for the CFM likelihood itself; Appendix D calibrates only the summary-level MDN emulator. Since the CFM marginal widths are the denominator of the headline 2.5-9.9 ratios and the reference for the 1.3-2.9 gap, an overconfident or biased CFM would directly inflate or shift these numbers. Please calibrate the CFM posterior on a set of held-out simulations (the 10% isotropic hold-out set provides roughly 185 such points) and report coverage diagnostics. If the CFM is miscalibrated, the benchmark should be re-derived or the ratios quoted as upper limits.
  2. [Appendix B, Eq. (B2)] The MOPED score derivatives are obtained by averaging CFM-generated realizations at off-grid parameter points. The same CFM defines the field-level benchmark, so the comparison is partly self-referential: a bias in the CFM's localised, high-order channels can simultaneously bias the benchmark widths and the MOPED projection directions in a way that makes the summary pipeline look closer to (or farther from) the field-level limit than it really is. Please demonstrate robustness, for example by recomputing the MOPED directions using simulation-based derivatives at the nearest LHS grid points, or by quantifying how much the Table 2 ratios change under plausible CFM mis-calibration.
  3. [Section 2.3, grid resolution] The 11-point-per-parameter grid spacing is comparable to the reported marginal widths: for q, Δq=0.006 versus σ_q=0.0084; for M_MW, Δ≈0.005×10^12 versus σ≈0.0066×10^12; for M_LMC, Δ≈0.024×10^11 versus σ≈0.036×10^11. The posterior marginals are also smoothed with a one-bin Gaussian kernel. At 1.3-2 grid spacings, the quoted widths may be significantly affected by the discretisation and smoothing rather than by the likelihood itself. Please recompute the CFM posterior on a finer local grid (or use a continuous interpolation/emulator) and verify that the Table 1 ratios and the subsequent comparisons are stable.
  4. [Section 2.2, Appendix A] The CFM likelihood is repeatedly described as 'exact', but the implemented log-likelihood uses the Hutchinson trace estimator with n_H=8 probe vectors and a fixed-step RK4 solver with 128 steps (Eq. A5-A6). These are controlled approximations, not exact evaluations. Please either soften the terminology or provide convergence checks in N_H and solver steps for the specific benchmark widths reported in Table 1.
minor comments (6)
  1. [Abstract and Section 3.2] The claim that MOPED 'preserves their Fisher information' is stated without the caveat that the implemented version uses a diagonal covariance approximation and finite-sample estimates. The caveat appears later in Appendix B and is handled honestly; please state it in the main text when MOPED is introduced.
  2. [Table 2] Please state explicitly in the table caption that the CFM column comes from a smoothed discrete grid posterior with one-bin Gaussian smoothing, and add the grid spacings so readers can assess the resolution issue directly.
  3. [Section 4.1] The mutual information estimates in Table 4 are reported without uncertainties. Since the complementarity conclusion (and hence the construction of the joint 19D vector) is based on these numbers, a simple bootstrap over validation points or a sensitivity check across training seeds would be useful.
  4. [Appendix D] The q parameter has a marginal P-P KS p-value of 3.1×10^-9, which is flagged and explained as weak leverage in parts of the prior. This is acceptable, but the explanation could be strengthened by showing that the q bias does not affect the parameter-ordering claims (the joint vector still beats moments and approaches CFM).
  5. [Figures 2 and 3] The qualitative overlap in Fig. 2 is convincing but could be quantified with a two-sample test on the 1D marginals or a lightweight distance metric. In Fig. 3, the CFM contours are smoothed grid posteriors while the Fisher forecasts are analytic ellipses; this asymmetry is fine but should be noted in the caption.
  6. [Section 6.3] The anisotropy stress test is a strong feature of the paper. Please also report the posterior widths in the anisotropic case explicitly (the text reports shifts but not the width comparison); this would help readers judge whether the failure is mainly a bias or also a width miscalibration.

Circularity Check

0 steps flagged

No significant circularity: the CFM benchmark is a held-out predictive test, the summary comparisons are measured quantities, and the shared CFM/MOPED score derivatives are a calibration caveat rather than a definitional reduction.

full rationale

Walking the derivation chain, I find no step in which a claimed prediction is identical by construction to a fitted input. The field-level benchmark is produced by a Conditional Flow Matching model trained on a 90% split of the HaloDance simulations with the fiducial simulation held out (Sec. 2.2, Figs. 2–3); the truth lying inside the 68% contour is a genuine out-of-sample test, not a re-statement of training data. The velocity-moment Fisher forecast is computed from the phase-space data and its covariance (Eqs. 1–3), independent of the CFM posterior. The BFE+MOPED summaries are built from the BFE coefficients of the same HaloDance particles, and the MOPED score derivatives are indeed drawn from the trained CFM (Appendix B), so the benchmark and the MOPED directions share a fitted model. This is a real shared-model systematic: if the CFM likelihood is miscalibrated, both the benchmark widths and the MOPED directions could be biased in correlated ways, affecting the quantitative 1.3–2.9 gap. However, the MOPED Fisher constraints are measured quantities (Sec. 3.3) rather than the CFM posterior itself, and MOPED's information-preservation property is a design identity, not a predicted discovery. The 185-grid summary-level validation and the anisotropy stress test provide independent checks of the qualitative ordering. Self-citations to the HaloDance suite are normal provenance and do not carry the argument through an unverified uniqueness claim. The limitations noted in the paper—single-fiducial CFM check, coarse 11-point grid, and the q P-P deviation—are calibration and precision concerns, not circularity.

Axiom & Free-Parameter Ledger

4 free parameters · 7 axioms · 0 invented entities

The central information-content claims rest on two load-bearing sets of assumptions: (i) the CFM emulator is a faithful likelihood (free network parameters, curriculum weights, no calibration test), and (ii) the summary-level Fisher/MI framework is valid (Gaussianity, diagonal covariance, chosen n_max). The HaloDance simulation suite itself (NFW+disk+bulge, first-infall LMC, beta=0) is an input from prior work and is not independently verified here. No invented physical entities appear.

free parameters (4)
  • CFM neural network weights = ~8.5e6 parameters
    Trained on 90% of isotropic HaloDance simulations; the field-level likelihood benchmark and the MOPED score derivatives depend entirely on this fit (Appendix A).
  • BFE+MOPED radial truncation n_max = 30
    Selected by minimizing the fiducial Fisher posterior volume over n_max=5-40 (Fig 7); constraints in Tables 1-2 are quoted at this chosen value.
  • CFM curriculum weighting w_i = unspecified schedule
    The loss weights that concentrate training near the fiducial cosmology are described qualitatively; they affect CFM fidelity and hence the benchmark.
  • CFM posterior grid smoothing kernel = 1 bin Gaussian
    CFM marginal widths in Table 1 are measured after smoothing the grid posterior by a one-bin Gaussian kernel; at sigma_q ~ 1.4 bins this choice affects the quoted width.
axioms (7)
  • domain assumption CFM likelihood is a faithful approximation to the HaloDance phase-space distribution at the held-out fiducial and across parameter space.
    Invoked in Sections 2.2-2.3 (benchmark) and Appendix B (MOPED scores). Only a one-point visual/coverage check is provided (Figs 2-3); no simulation-based calibration.
  • domain assumption Gaussian likelihood for the 15 velocity moments and the 19D joint summaries.
    Eq. 1 and Eq. 12; chi-square and skewness checks in Section 5 support it for the joint summaries, but the residual q coverage deviation (Appendix D, p_KS=3e-9) shows it is not exact.
  • domain assumption Biorthogonal BFE with NFW r_s=16 kpc, l_max=5, n_max=40 spans the halo fields in 30-120 kpc.
    Section 3.1; reconstruction tests show residuals at shot-noise level, but the basis choice sets the feature space from which information is measured.
  • ad hoc to paper MOPED with diagonal covariance preserves Fisher information.
    Section 3.2/Appendix B: authors state the diagonal approximation is 'conservative rather than lossless' and that including off-diagonal terms changes the Fisher volume by 0.15-0.51. The ideal MOPED guarantee does not hold in the implemented version.
  • domain assumption Dark matter particles as equal-mass tracers with no selection function; beta=0 fiducial.
    Sections 2.1 and 6.3; anisotropy stress test shows strong bias when this is violated (Delta c = +1.24, Delta M_MW = -0.59).
  • domain assumption Uniform prior over the HaloDance parameter box.
    Used for the CFM grid posterior and dynesty sampling (Sections 2.3 and 5).
  • standard math Fisher/Cramer-Rao bound applies to the summary forecasts.
    Eq. 1 and Section 2.3; standard statistical result used to convert summary Fisher matrices into marginal sigma constraints.

pith-pipeline@v1.3.0-alltime-deepseek · 29126 in / 18282 out tokens · 180848 ms · 2026-08-01T11:03:20.197077+00:00 · methodology

0 comments
read the original abstract

The gravitational interaction between the Milky Way (MW) and the Large Magellanic Cloud (LMC) drives the outer halo into dynamical disequilibrium, imprinting the masses and structural parameters of both galaxies onto the 6D phase-space distribution of halo tracers. This signal has been characterised with summary statistics ranging from low-order velocity moments to basis function expansions, yet how much information these summaries discard, and whether they are complementary, remains unclear. We address these questions by comparing a field-level likelihood benchmark with physically interpretable summaries for constraining $(M_{\mathrm{MW}}, M_{\mathrm{LMC}}, c, q)$, where $c$ and $q$ are the MW halo concentration and flattening. A Conditional Flow Matching (CFM) model trained on the HaloDance $N$-body suite provides an exact likelihood at a held-out fiducial point; for 5,000 tracers in $30$--$120$~kpc it tightens marginal constraints by factors of $2.5$--$9.9$ over an all-sky velocity-moment forecast. We then expand the halo density and velocity fields in a multipole basis-function expansion (BFE) and compress the coefficients with the Massive Optimised Parameter Estimation and Data compression (MOPED) algorithm into four parameter-sensitive summaries that preserve their Fisher information. A variational mutual-information analysis shows that the BFE+MOPED summaries and the velocity moments are complementary, so we combine them into a joint $19$-dimensional vector as our primary inference pipeline: it tightens the marginal constraints by up to $15$ per cent over BFE+MOPED alone and by $30$--$71$ per cent over velocity moments alone, reaching within a factor of $1.3$--$2.9$ of the field-level benchmark. We thus establish a physically interpretable summary-level route to MW--LMC inference alongside the field-level benchmark that bounds its information content.

Figures

Figures reproduced from arXiv: 2607.20006 by Jiashu Pan, Roland M. Crocker, Tomasz R\'o\.za\'nski, Xiang-Xiang Xue, Yanjun Sheng, Yuan-Sen Ting.

Figure 1
Figure 1. Figure 1: Schematic of the conditional flow matching setup. Gaussian base particles x0 are connected to HaloDance phase-space particles x1 = (r, v) through interpolated states x𝑡 . The network learns the conditional velocity field u𝑡 (x𝑡 | 𝜽), forward integration generates mock halo tracers, while backward integration maps observed 6D catalogues to the base distribution for likelihood evaluation [PITH_FULL_IMAGE:fi… view at source ↗
Figure 2
Figure 2. Figure 2: Pairwise comparison of the six Galactocentric phase-space coordinates for CFM-generated particles (blue) and particles from a held-out fiducial simulation (grey), both restricted to 30–120 kpc. The fiducial point is (𝑀MW, 𝑀LMC, 𝑐, 𝑞) = (0.7 × 1012 M⊙, 1.5 × 1011 M⊙, 9.415, 1.0) with 𝛽(𝑟 ) = 0. The diagonal panels show the one-dimensional marginals of 𝑥, 𝑦, 𝑧 (kpc) and 𝑣𝑥, 𝑣𝑦, 𝑣𝑧 (km s−1 ), and the lower tr… view at source ↗
Figure 3
Figure 3. Figure 3: CFM-based full-phase-space likelihood benchmark for the held-out fiducial simulation at 𝜽★ with 𝛽(𝑟 ) = 0. The corner plot overlays two contour sets in the same parameter planes: the CFM full-phase-space posterior from the likelihood grid for 𝑁 = 5,000 particles in 30–120 kpc (blue, filled), and the velocity-moment Fisher forecast (red, Gaussian contours). Both sets show the 68 and 95 per cent credible reg… view at source ↗
Figure 4
Figure 4. Figure 4: Spherically averaged density profile of particles drawn from the fiducial simulation within 30–120 kpc. Open circles show the particle density estimate, and the solid blue line shows the BFE reconstruction with 𝑛max = 40 and ℓmax = 5. The BFE traces the particle data across the full radial range, confirming that the adopted radial basis captures the scales relevant for inference. centric coordinates, the s… view at source ↗
Figure 5
Figure 5. Figure 5: BFE reconstruction of the fiducial density and velocity fields in the 60–90 kpc shell. Rows show the fractional density contrast 𝛿𝜌 = 𝜌/⟨𝜌⟩sky − 1 and the Galactocentric velocities 𝑣𝑟 , 𝑣𝑏, and 𝑣ℓ . Columns show the particle field, the BFE reconstruction with 𝑛max = 40 and ℓmax = 5, and the residual. This figure checks that the BFE basis represents the large-scale density asymmetry, radial-velocity reflex … view at source ↗
Figure 6
Figure 6. Figure 6: Response matrix of the four whitened BFE+MOPED summaries to the model parameters, 𝑅𝛼𝛽 = 𝜕⟨𝑠˜𝛼 ⟩/𝜕𝜃𝛽, evaluated at the fiducial point and shown in units of the response per 1𝜎 shift of each parameter. Rows are the summaries 𝑠𝑀MW , 𝑠𝑀LMC , 𝑠𝑐, and 𝑠𝑞, and columns are the model parameters (𝑀MW, 𝑀LMC, 𝑐, 𝑞). The overall sign of each summary is conventional, so only relative signs within a row are meaningful. T… view at source ↗
Figure 7
Figure 7. Figure 7: Radial truncation diagnostic for the BFE+MOPED compression. The left panel shows the four marginal constraints, each normalised by the mean over the plateau values 𝑛max = 30, 35, and 40, so that 𝑀MW, 𝑀LMC, 𝑐, and 𝑞 appear together. The right panel shows the joint Fisher volume normalised the same way, with the number of active BFE features on the secondary axis. The joint volume reaches its minimum at 𝑛max… view at source ↗
Figure 8
Figure 8. Figure 8: Per-particle MOPED contribution maps from the density and first-moment velocity BFE channels in the 60–90 kpc shell. Columns correspond to the four scalar BFE+MOPED summaries, and rows correspond to 𝜌, 𝑣𝑟 , 𝑣𝑏, and 𝑣ℓ . These maps are not maps of the fields themselves; they show where particles in each channel increase or decrease each summary. The sign is conventional, while the angular pattern identifies… view at source ↗
Figure 9
Figure 9. Figure 9: Per-particle MOPED contribution maps from the diagonal second-moment velocity channels in the 60–90 kpc shell. These maps show where 𝑣 2 𝑟 , 𝑣 2 𝑏 , and 𝑣 2 ℓ contribute to each scalar summary. The dominant response is in 𝑠𝑀MW and 𝑠𝑐, whose nearly monopolar patterns reflect the MW mass–concentration degeneracy in tracer dispersions [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: places all four cases in the same parameter planes and 68 and 95 per cent convention as [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: True-vs-predicted accuracy of the MDN-style Gaussian summary-likelihood emulator for the primary joint-summary pipeline on the isotropic 185-grid validation set. The top row shows posterior medians against the true parameters, and the bottom row shows median-minus-true residuals. 0.6 0.8 1.0 1.2 1.4 1.6 1.8 Predicted MMW (10 12 M ) 8 10 12 14 16 18 20 MLMC (10 10 M ) 6 8 10 12 14 c 0.6 0.8 1.0 1.2 1.4 q 0… view at source ↗
Figure 12
Figure 12. Figure 12: True-vs-predicted accuracy for the anisotropic stress test of the isotropic-trained BFE+MOPED-only pipeline, using the radially varying 𝛽(𝑟 ) HaloDance set. Relative to the corresponding isotropic BFE+MOPED validation, the medians shift toward higher 𝑐 and lower 𝑀MW, while the shifts in 𝑀LMC and 𝑞 are smaller. MNRAS 000, 1–21 (2026) [PITH_FULL_IMAGE:figures/full_fig_p016_12.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

93 extracted references · 7 canonical work pages · 3 internal anchors

  1. [1]

    The LMC Corona Favors a First Passage

    The LMC Corona Favors a First Passage. arXiv e-prints , keywords =. doi:10.48550/arXiv.2510.03395 , archivePrefix =. 2510.03395 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2510.03395

  2. [2]

    , keywords =

    The Galaxy in Context: Structural, Kinematic, and Integrated Properties. , keywords =. doi:10.1146/annurev-astro-081915-023441 , archivePrefix =. 1602.07702 , primaryClass =

    Pith/arXiv arXiv

  3. [3]

    , keywords =

    Measuring the Mass of the Large Magellanic Cloud with Stellar Streams Observed by S ^ 5. , keywords =. doi:10.3847/1538-4357/ac2e93 , archivePrefix =. 2107.13004 , primaryClass =

    Pith/arXiv arXiv

  4. [4]

    , keywords =

    A determination of the Large Magellanic Cloud dark matter subhalo mass using the Milky Way halo stars in its gravitational wake. , keywords =. doi:10.1051/0004-6361/202347450 , archivePrefix =. 2309.12989 , primaryClass =

    Pith/arXiv arXiv

  5. [5]

    Third-epoch Magellanic Cloud Proper Motions. II. The Large Magellanic Cloud Rotation Field in Three Dimensions. , keywords =. doi:10.1088/0004-637X/781/2/121 , archivePrefix =. 1305.4641 , primaryClass =

    Pith/arXiv arXiv

  6. [6]

    , keywords =

    The Mass of the Large Magellanic Cloud from the Three-dimensional Kinematics of Its Globular Clusters. , keywords =. doi:10.3847/1538-4357/ad1f58 , archivePrefix =. 2401.14458 , primaryClass =

    Pith/arXiv arXiv

  7. [7]

    , keywords =

    The total mass of the Large Magellanic Cloud from its perturbation on the Orphan stream. , keywords =. doi:10.1093/mnras/stz1371 , archivePrefix =. 1812.08192 , primaryClass =

    Pith/arXiv arXiv

  8. [8]

    , keywords =

    Tango for three: Sagittarius, LMC, and the Milky Way. , keywords =. doi:10.1093/mnras/staa3673 , archivePrefix =. 2009.10726 , primaryClass =

    Pith/arXiv arXiv 2009

  9. [9]

    , keywords =

    S ^ 5 : Probing the Milky Way and Magellanic Clouds potentials with the 6D map of the Orphan-Chenab stream. , keywords =. doi:10.1093/mnras/stad551 , archivePrefix =. 2211.04495 , primaryClass =

    Pith/arXiv arXiv

  10. [10]

    , keywords =

    Implications of the Milky Way Travel Velocity for Dynamical Mass Estimates of the Local Group. , keywords =. doi:10.3847/1538-4357/aca01f , archivePrefix =. 2204.07173 , primaryClass =

    Pith/arXiv arXiv

  11. [11]

    , keywords =

    The Distant Milky Way Halo from the Southern Hemisphere: Characterization of the LMC-induced Dynamical Friction Wake. , keywords =. doi:10.3847/1538-4357/adbf08 , archivePrefix =. 2410.00114 , primaryClass =

    Pith/arXiv arXiv

  12. [12]

    , keywords =

    And Yet it Moves: The Dangers of Artificially Fixing the Milky Way Center of Mass in the Presence of a Massive Large Magellanic Cloud. , keywords =. doi:10.1088/0004-637X/802/2/128 , archivePrefix =. 1408.4128 , primaryClass =

    Pith/arXiv arXiv

  13. [13]

    , keywords =

    The influence of Sagittarius and the Large Magellanic Cloud on the stellar disc of the Milky Way Galaxy. , keywords =. doi:10.1093/mnras/sty1574 , archivePrefix =. 1710.02538 , primaryClass =

    Pith/arXiv arXiv

  14. [14]

    , keywords =

    Response of the Milky Way's disc to the Large Magellanic Cloud in a first infall scenario. , keywords =. doi:10.1093/mnras/stx2146 , archivePrefix =. 1608.04743 , primaryClass =

    Pith/arXiv arXiv

  15. [15]

    , keywords =

    Hunting for the Dark Matter Wake Induced by the Large Magellanic Cloud. , keywords =. doi:10.3847/1538-4357/ab32eb , archivePrefix =. 1902.05089 , primaryClass =

    Pith/arXiv arXiv 1902

  16. [16]

    , keywords =

    Reflex motion in the Milky Way stellar halo resulting from the Large Magellanic Cloud infall. , keywords =. doi:10.1093/mnrasl/slaa029 , archivePrefix =. 2001.09142 , primaryClass =

    Pith/arXiv arXiv 2001

  17. [17]

    , keywords =

    Detection of the LMC-induced sloshing of the Galactic halo. , keywords =. doi:10.1093/mnras/stab1828 , archivePrefix =. 2010.13789 , primaryClass =

    Pith/arXiv arXiv 2010

  18. [18]

    Galaxies , keywords =

    The Effect of the LMC on the Milky Way System. Galaxies , keywords =. doi:10.3390/galaxies11020059 , archivePrefix =. 2304.09136 , primaryClass =

    Pith/arXiv arXiv

  19. [19]

    The Open Journal of Astrophysics , keywords =

    Effect of the Large Magellanic Cloud on the kinematics of Milky Way satellites and virial mass estimate. The Open Journal of Astrophysics , keywords =. doi:10.33232/001c.120316 , archivePrefix =. 2405.06017 , primaryClass =

    Pith/arXiv arXiv

  20. [20]

    , keywords =

    Uncovering the first-infall history of the LMC through its dynamical impact in the Milky Way halo. , keywords =. doi:10.1093/mnras/stae2259 , archivePrefix =. 2404.08975 , primaryClass =

    Pith/arXiv arXiv

  21. [21]

    HaloDance simulation suite and observational forecasts

    LMC-induced perturbations in the Milky Way halo ─ I. HaloDance simulation suite and observational forecasts. , keywords =. doi:10.1093/mnras/staf1780 , archivePrefix =. 2507.03663 , primaryClass =

    Pith/arXiv arXiv

  22. [22]

    The Milky Way - Large Magellanic Cloud Interaction with Simulation Based Inference

    The Milky Way - Large Magellanic Cloud Interaction with Simulation Based Inference. arXiv e-prints , keywords =. doi:10.48550/arXiv.2510.04735 , archivePrefix =. 2510.04735 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2510.04735

  23. [23]

    , keywords =

    A simulation-based inference approach to the dynamics of the MW─LMC system - validation. , keywords =. doi:10.1093/mnras/staf2111 , archivePrefix =. 2507.10667 , primaryClass =

    arXiv

  24. [24]

    , keywords =

    Quantifying the Milky Way, LMC and their interaction using all-sky kinematics of outer halo stars. , keywords =. doi:10.1093/mnras/stag926 , archivePrefix =. 2603.08788 , primaryClass =

    Pith/arXiv arXiv

  25. [25]

    Shaping the Milky Way. II. The dark matter halo's response to the LMC's passage in a cosmological context. arXiv e-prints , keywords =. doi:10.48550/arXiv.2511.02031 , archivePrefix =. 2511.02031 , primaryClass =

    doi:10.48550/arxiv.2511.02031

  26. [26]

    , keywords =

    The effect of dynamical friction on the orbit of the Magellanic Clouds. , keywords =. doi:10.1086/154049 , adsurl =

    work page doi:10.1086/154049

  27. [27]

    , keywords =

    The Magellanic Stream and the Galaxy with a Massive Halo. , keywords =. doi:10.1093/pasj/32.4.581 , adsurl =

    doi:10.1093/pasj/32.4.581

  28. [28]

    , keywords =

    On the proper motion of the Magellanic Clouds and the halo mass of our galaxy. , keywords =. doi:10.1093/mnras/198.3.707 , adsurl =

    work page doi:10.1093/mnras/198.3.707

  29. [29]

    , keywords =

    The Proper Motion of the Large Magellanic Cloud Using HST. , keywords =. doi:10.1086/498972 , archivePrefix =. astro-ph/0508457 , primaryClass =

    Pith/arXiv arXiv

  30. [30]

    Third-epoch Magellanic Cloud Proper Motions. I. Hubble Space Telescope/WFC3 Data and Orbit Implications. , keywords =. doi:10.1088/0004-637X/764/2/161 , archivePrefix =. 1301.0832 , primaryClass =

    Pith/arXiv arXiv

  31. [31]

    , keywords =

    Are the Magellanic Clouds on Their First Passage about the Milky Way?. , keywords =. doi:10.1086/521385 , archivePrefix =. astro-ph/0703196 , primaryClass =

    Pith/arXiv arXiv

  32. [32]

    , keywords =

    The frequency of dwarf galaxy multiples at low redshift in SDSS versus cosmological expectations. , keywords =. doi:10.1093/mnras/sty2041 , archivePrefix =. 1807.06673 , primaryClass =

    Pith/arXiv arXiv

  33. [33]

    , keywords =

    Dear Magellanic Clouds, welcome back!. , keywords =. doi:10.1093/mnras/stad2612 , archivePrefix =. 2306.04837 , primaryClass =

    Pith/arXiv arXiv

  34. [34]

    The Milky Way Tomography with Subaru Hyper Suprime-Cam: Implications for the past orbit of the Large Magellanic Cloud

    The Milky Way Tomography with Subaru Hyper Suprime-Cam: Implications for the past orbit of the Large Magellanic Cloud. arXiv e-prints , keywords =. doi:10.48550/arXiv.2604.10281 , archivePrefix =. 2604.10281 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2604.10281

  35. [35]

    Nature Astronomy , keywords =

    Detection of the Milky Way reflex motion due to the Large Magellanic Cloud infall. Nature Astronomy , keywords =. doi:10.1038/s41550-020-01254-3 , archivePrefix =. 2011.10581 , primaryClass =

    Pith/arXiv arXiv 2011

  36. [36]

    Dynamical Friction. II. The Rate of Escape of Stars from Clusters and the Evidence for the Operation of Dynamical Friction. , year = 1943, month = mar, volume =. doi:10.1086/144518 , adsurl =

    doi:10.1086/144518 1943

  37. [37]

    , keywords =

    Revisiting Dynamical Friction: The Role of Global Modes and Local Wakes. , keywords =. doi:10.3847/1538-4357/ac0627 , archivePrefix =. 2007.13763 , primaryClass =

    Pith/arXiv arXiv 2007

  38. [38]

    , keywords =

    Structure, Kinematics, and Observability of the Large Magellanic Cloud's Dynamical Friction Wake in Cold versus Fuzzy Dark Matter. , keywords =. doi:10.3847/1538-4357/ace533 , archivePrefix =. 2307.00053 , primaryClass =

    Pith/arXiv arXiv

  39. [39]

    , keywords =

    All-sky dynamical response of the Galactic halo to the Large Magellanic Cloud. , keywords =. doi:10.1038/s41586-021-03385-7 , archivePrefix =. 2104.09515 , primaryClass =

    Pith/arXiv arXiv

  40. [40]

    , keywords =

    The radial variation of the LMC-induced reflex motion of the Milky Way disc observed in the stellar halo. , keywords =. doi:10.1093/mnras/stae1363 , archivePrefix =. 2402.10841 , primaryClass =

    Pith/arXiv arXiv

  41. [41]

    , keywords =

    All-sky Kinematics of the Distant Halo: The Reflex Response to the LMC. , keywords =. doi:10.3847/1538-4357/addab6 , archivePrefix =. 2406.01676 , primaryClass =

    Pith/arXiv arXiv

  42. [42]

    , keywords =

    Exploring the interaction between the MW and LMC with a large sample of blue horizontal branch stars from the DESI survey. , keywords =. doi:10.1093/mnras/staf1219 , archivePrefix =. 2410.09149 , primaryClass =

    Pith/arXiv arXiv

  43. [43]

    arXiv e-prints , keywords =

    The velocity field of our Milky Way outer stellar halo based on DESI DR2. arXiv e-prints , keywords =. doi:10.48550/arXiv.2604.01628 , archivePrefix =. 2604.01628 , primaryClass =

    doi:10.48550/arxiv.2604.01628

  44. [44]

    , keywords =

    The M31─M33 Interaction: Impact on M31's Center-of-mass Motion and Satellite Orbits. , keywords =. doi:10.3847/1538-4357/adc992 , archivePrefix =. 2504.02805 , primaryClass =

    Pith/arXiv arXiv

  45. [45]

    , keywords =

    The Pisces Plume and the Magellanic wake. , keywords =. doi:10.1093/mnrasl/slz101 , archivePrefix =. 1904.07909 , primaryClass =

    Pith/arXiv arXiv 1904

  46. [46]

    , keywords =

    Quantifying the Impact of the Large Magellanic Cloud on the Structure of the Milky Way's Dark Matter Halo Using Basis Function Expansions. , keywords =. doi:10.3847/1538-4357/ac0b44 , archivePrefix =. 2010.00816 , primaryClass =

    Pith/arXiv arXiv 2010

  47. [47]

    arXiv e-prints , keywords =

    Modeling the recent interactions between the Magellanic Clouds and Milky Way. arXiv e-prints , keywords =. doi:10.48550/arXiv.2602.05021 , archivePrefix =. 2602.05021 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2602.05021

  48. [48]

    , keywords =

    The effect of the deforming dark matter haloes of the Milky Way and the Large Magellanic Cloud on the Orphan-Chenab stream. , keywords =. doi:10.1093/mnras/stac3108 , archivePrefix =. 2205.01688 , primaryClass =

    Pith/arXiv arXiv

  49. [49]

    , keywords =

    Mapping the Distorted Dark Matter Distribution of the LMC─SMC System Prior to Milky Way Infall with Basis Function Expansions. , keywords =. doi:10.3847/1538-4357/ae4a9d , archivePrefix =. 2601.00946 , primaryClass =

    Pith/arXiv arXiv

  50. [50]

    , keywords =

    Constraining the Milky Way Halo Kinematics via Its Linear Response to the Large Magellanic Cloud. , keywords =. doi:10.3847/1538-4357/ac7139 , archivePrefix =. 2201.05589 , primaryClass =

    Pith/arXiv arXiv

  51. [51]

    , keywords =

    Quantifying the Stellar Halo's Response to the LMC's Infall with Spherical Harmonics. , keywords =. doi:10.3847/1538-4357/ab9b88 , archivePrefix =. 2006.08621 , primaryClass =

    Pith/arXiv arXiv 2006

  52. [52]

    Euclid preparation. I. The Euclid Wide Survey. , keywords =. doi:10.1051/0004-6361/202141938 , archivePrefix =. 2108.01201 , primaryClass =

    Pith/arXiv arXiv

  53. [53]

    Journal of Astronomical Telescopes, Instruments, and Systems , keywords =

    Astrometry with the Wide-Field Infrared Space Telescope. Journal of Astronomical Telescopes, Instruments, and Systems , keywords =. doi:10.1117/1.JATIS.5.4.044005 , archivePrefix =. 1712.05420 , primaryClass =

    arXiv

  54. [54]

    Science China Physics, Mechanics, and Astronomy , keywords =

    Introduction to the Chinese Space Station Survey Telescope (CSST). Science China Physics, Mechanics, and Astronomy , keywords =. doi:10.1007/s11433-025-2809-0 , archivePrefix =. 2507.04618 , primaryClass =

    arXiv

  55. [55]

    arXiv e-prints , keywords =

    The DESI Experiment Part I: Science,Targeting, and Survey Design. arXiv e-prints , keywords =. doi:10.48550/arXiv.1611.00036 , archivePrefix =. 1611.00036 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1611.00036

  56. [56]

    Research in Astronomy and Astrophysics , year = 2012, month = sep, volume =

    The Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST). Research in Astronomy and Astrophysics , year = 2012, month = sep, volume =. doi:10.1088/1674-4527/12/9/003 , adsurl =

    doi:10.1088/1674-4527/12/9/003 2012

  57. [57]

    The Messenger , keywords =

    4MOST: Project overview and information for the First Call for Proposals. The Messenger , keywords =. doi:10.18727/0722-6691/5117 , archivePrefix =. 1903.02464 , primaryClass =

    Pith/arXiv arXiv 1903

  58. [58]

    , keywords =

    The wide-field, multiplexed, spectroscopic facility WEAVE: Survey design, overview, and simulated implementation. , keywords =. doi:10.1093/mnras/stad557 , archivePrefix =. 2212.03981 , primaryClass =

    Pith/arXiv arXiv

  59. [59]

    , keywords =

    Mapping the Stellar Halo with the H3 Spectroscopic Survey. , keywords =. doi:10.3847/1538-4357/ab38b8 , archivePrefix =. 1907.07684 , primaryClass =

    Pith/arXiv arXiv 1907

  60. [60]

    Ground-based and Airborne Instrumentation for Astronomy VI , year = 2016, editor =

    Prime Focus Spectrograph (PFS) for the Subaru telescope: overview, recent progress, and future perspectives. Ground-based and Airborne Instrumentation for Astronomy VI , year = 2016, editor =. doi:10.1117/12.2232103 , archivePrefix =. 1608.01075 , primaryClass =

    Pith/arXiv arXiv 2016

  61. [61]

    arXiv e-prints , keywords =

    Deep Learning in Astrophysics. arXiv e-prints , keywords =. doi:10.48550/arXiv.2510.10713 , archivePrefix =. 2510.10713 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2510.10713

  62. [62]

    arXiv e-prints , keywords =

    Flow Matching for Generative Modeling. arXiv e-prints , keywords =. doi:10.48550/arXiv.2210.02747 , archivePrefix =. 2210.02747 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2210.02747

  63. [63]

    arXiv e-prints , keywords =

    Flow Matching Guide and Code. arXiv e-prints , keywords =. doi:10.48550/arXiv.2412.06264 , archivePrefix =. 2412.06264 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2412.06264

  64. [64]

    , keywords =

    The Gravitational Field of Flat Galaxies. , keywords =. doi:10.1007/BF00643095 , adsurl =

    work page doi:10.1007/bf00643095

  65. [65]

    , keywords =

    An Adaptive Algorithm for N-Body Field Expansions. , keywords =. doi:10.1086/300669 , archivePrefix =. astro-ph/9805357 , primaryClass =

    Pith/arXiv arXiv

  66. [66]

    The Journal of Open Source Software , keywords =

    EXP: a Python/C++ package for basis function expansion methods in galactic dynamics. The Journal of Open Source Software , keywords =. doi:10.21105/joss.07302 , archivePrefix =. 2505.06403 , primaryClass =

    Pith/arXiv arXiv

  67. [67]

    , keywords =

    Massive lossless data compression and multiple parameter estimation from galaxy spectra. , keywords =. doi:10.1046/j.1365-8711.2000.03692.x , archivePrefix =. astro-ph/9911102 , primaryClass =

    arXiv 2000

  68. [68]

    , keywords =

    A Universal Density Profile from Hierarchical Clustering. , keywords =. doi:10.1086/304888 , archivePrefix =. astro-ph/9611107 , primaryClass =

    Pith/arXiv arXiv

  69. [69]

    , keywords =

    Three-Dimensional Models for the Distribution of Mass in Galaxies. , keywords =. doi:10.1093/pasj/27.4.533 , adsurl =

    doi:10.1093/pasj/27.4.533

  70. [70]

    , keywords =

    An Analytical Model for Spherical Galaxies and Bulges. , keywords =. doi:10.1086/168845 , adsurl =

    doi:10.1086/168845

  71. [71]

    , keywords =

    An iterative method for the construction of N-body galaxy models in collisionless equilibrium. , keywords =. doi:10.1093/mnras/stu1421 , archivePrefix =. 1402.1623 , primaryClass =

    Pith/arXiv arXiv

  72. [72]

    , keywords =

    Simulating cosmic structure formation with the GADGET-4 code. , keywords =. doi:10.1093/mnras/stab1855 , archivePrefix =. 2010.03567 , primaryClass =

    Pith/arXiv arXiv 2010

  73. [73]

    arXiv e-prints , keywords =

    Building Normalizing Flows with Stochastic Interpolants. arXiv e-prints , keywords =. doi:10.48550/arXiv.2209.15571 , archivePrefix =. 2209.15571 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2209.15571

  74. [74]

    , keywords =

    KRATOS: A large suite of N-body simulations to interpret the stellar kinematics of LMC-like discs. , keywords =. doi:10.1051/0004-6361/202349058 , archivePrefix =. 2404.04061 , primaryClass =

    Pith/arXiv arXiv

  75. [75]

    , keywords =

    Milky Way-est: Cosmological Zoom-in Simulations with Large Magellanic Cloud and Gaia─Sausage─Enceladus Analogs. , keywords =. doi:10.3847/1538-4357/ad554c , archivePrefix =. 2404.08043 , primaryClass =

    Pith/arXiv arXiv

  76. [76]

    arXiv e-prints , keywords =

    Variational Inference with Normalizing Flows. arXiv e-prints , keywords =. doi:10.48550/arXiv.1505.05770 , archivePrefix =. 1505.05770 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1505.05770

  77. [77]

    arXiv e-prints , keywords =

    Neural Spline Flows. arXiv e-prints , keywords =. doi:10.48550/arXiv.1906.04032 , archivePrefix =. 1906.04032 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1906.04032 1906

  78. [78]

    arXiv e-prints , keywords =

    Deep Unsupervised Learning using Nonequilibrium Thermodynamics. arXiv e-prints , keywords =. doi:10.48550/arXiv.1503.03585 , archivePrefix =. 1503.03585 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1503.03585

  79. [79]

    arXiv e-prints , keywords =

    Denoising Diffusion Probabilistic Models. arXiv e-prints , keywords =. doi:10.48550/arXiv.2006.11239 , archivePrefix =. 2006.11239 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2006.11239 2006

  80. [80]

    arXiv e-prints , keywords =

    Score-Based Generative Modeling through Stochastic Differential Equations. arXiv e-prints , keywords =. doi:10.48550/arXiv.2011.13456 , archivePrefix =. 2011.13456 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2011.13456 2011

Showing first 80 references.