pith. sign in

arxiv: 2606.20180 · v1 · pith:WX53TWCCnew · submitted 2026-06-18 · ⚛️ physics.ins-det

Raw-Hit Muon Tomography: A Measurement-Domain Formulation for Cosmic-Ray Muon Imaging

Zhizheng Zhao , Changhao Qin , Rongfeng Zhang , Zibo Qin , Qite Li This is my paper

Pith reviewed 2026-06-26 15:15 UTC · model grok-4.3

classification ⚛️ physics.ins-det
keywords cosmic-ray muon tomographyraw-hit formulationscattering tomographyenergy-loss imagingFermi-Eyges covarianceBethe-Bloch integralGeant4 benchmarkROC-AUC
0
0 comments X

The pith

Raw-Hit Muon Tomography formulates cosmic muon imaging directly on detector hits using a Student-t likelihood from Fermi-Eyges residuals and a Bethe-Bloch line integral for momentum loss.

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

The paper introduces Raw-Hit Muon Tomography (RHMT) to keep the inverse problem in the raw measurement domain instead of first reducing hits to per-muon scattering summaries. RHMT-S removes the unknown straight track from the hits, models the residuals with Fermi-Eyges covariance, and marginalizes the unknown scattering scale to obtain a blank-calibrated Student-t likelihood. RHMT-E separately fits hits in a six-plane magnetic spectrometer to extract each muon's log momentum loss and treats it as a Bethe-Bloch integral of electron-density contrast. In a controlled Geant4 benchmark the combined approach raises mean ROC-AUC from 0.81 for angular-scattering baselines to 0.84-0.86 and supplies independent contrast for materials such as aluminium where scattering is weak.

Core claim

Raw-Hit Muon Tomography (RHMT) formulates the inverse problem in the measurement domain by projecting out the unknown straight track from detector hits and evaluating the residual contrast with a Fermi-Eyges covariance; marginalizing the unknown scattering scale produces a blank-calibrated Student-t-type likelihood for RHMT-S, while RHMT-E separately estimates log momentum loss in a six-plane spectrometer as a line integral of electron density contrast. In controlled Geant4 benchmarks this yields mean ROC-AUC values of 0.84-0.86 compared with 0.81 for angular-scattering reconstruction baselines, and supplies independent contrast for materials where scattering is weak.

What carries the argument

Raw-Hit Muon Tomography (RHMT) operating directly on detector hits, with RHMT-S using Fermi-Eyges covariance after track projection and marginalization to a Student-t likelihood, and RHMT-E using six-plane spectrometer fits to Bethe-Bloch line integrals of momentum loss.

If this is right

  • Material discrimination improves when the full set of hit coordinates is retained rather than collapsed to a single scattering angle per muon.
  • Energy-loss contrast becomes available as an orthogonal channel for low-Z materials where scattering contrast is weak.
  • The Student-t likelihood after marginalization supplies automatic calibration against empty regions without separate background runs.
  • Performance gains appear in the sparse-hit regime typical of four- to six-plane cosmic-muon setups.

Where Pith is reading between the lines

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

  • The formulation could be combined with existing track-fitting pipelines to add a second contrast map without extra hardware.
  • Real-world deployment would require checking whether the marginalization step remains stable when hit resolutions vary across detector planes.
  • The same measurement-domain approach might extend to other charged-particle imaging modalities that record only a few plane crossings.

Load-bearing premise

The Geant4 simulation faithfully reproduces real detector hit resolutions, material interactions, and particle responses so that the reported ROC-AUC gains translate to physical data.

What would settle it

A side-by-side comparison of ROC-AUC values obtained from the same material targets imaged with a physical detector versus the Geant4 benchmark values of 0.84-0.86.

Figures

Figures reproduced from arXiv: 2606.20180 by Changhao Qin, Qite Li, Rongfeng Zhang, Zhizheng Zhao, Zibo Qin.

Figure 1
Figure 1. Figure 1: Measurement-domain view of RHMT. (a) A muon is recorded only through its crossings of detector planes. In the spectrometer geometry shown here, six RPC planes bracket two dipole magnets and the sample. (b) RHMT-S uses the hit residuals left after removing the straight track to estimate radiation-length contrast λ = 1/X0. (c) RHMT-E uses the magnetic bend to estimate the log momentum loss ℓ = ln(pin/pout), … view at source ↗
Figure 2
Figure 2. Figure 2: Benchmark geometry. Each scene contains one U-shaped block in a 400 × 400 × 470 mm background. The block has a 150 mm square footprint, a 60 mm slot cut 105 mm from the top centre, and 45 mm arms; it is 235 mm thick and centred at half-depth. The six scenes pair three object materials (lead, water, aluminium) with two backgrounds (SiO2, concrete). 4.3 Protocol and Metric We vary exposure over full, 50k, an… view at source ↗
Figure 3
Figure 3. Figure 3: Representative reconstructions at full exposure, σpos=1 mm. Rows pair each tier’s best scattering baseline with RHMT: ASR vs. RHMT-S (tracker, top) and ASR+momentum vs. RHMT-E (spectrometer, bottom). Columns are the six scenes; each panel shows the column image minus the background median (red = excess, blue = deficit; dashed outline = truth). 8 [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
read the original abstract

Cosmic-ray muon tomography records only a few detector-plane crossings per particle, while material information enters through stochastic scattering and energy loss along the path. Most pipelines first compress these hits to a per-muon scattering summary and assign a nominal momentum, moving the inverse problem away from the raw measurements. We introduce Raw-Hit Muon Tomography (RHMT), a measurement-domain formulation built directly on detector hits. RHMT-S projects out the unknown straight track and evaluates the remaining hit contrast with a Fermi--Eyges covariance; marginalizing the unknown scattering scale gives a blank-calibrated Student-$t$-type likelihood. RHMT-E fits the hits in a six-plane magnetic spectrometer to estimate each muon's log momentum loss and models it as a Bethe--Bloch line integral of the electron-density-related contrast $\rho Z/A$. In a controlled Geant4 benchmark, RHMT-S improves the mean ROC-AUC over four-plane scattering baselines ($0.84$--$0.86$ versus $0.81$ for ASR), and RHMT-E provides a separate energy-loss contrast for aluminium, where scattering contrast is weak.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The paper introduces Raw-Hit Muon Tomography (RHMT), a measurement-domain formulation for cosmic-ray muon imaging that operates directly on detector hits rather than compressed per-muon summaries. RHMT-S projects out the unknown straight track, applies a Fermi-Eyges covariance to the residual hit contrast, and marginalizes an unknown scattering scale to obtain a Student-t-type likelihood. RHMT-E fits hits in a six-plane magnetic spectrometer to estimate log momentum loss modeled via a Bethe-Bloch line integral of the electron-density contrast ho Z/A. In a controlled Geant4 benchmark the manuscript reports that RHMT-S improves mean ROC-AUC to 0.84-0.86 versus 0.81 for an ASR four-plane scattering baseline, while RHMT-E supplies additional contrast for aluminum where scattering contrast is weak.

Significance. If the benchmark results are robust, the measurement-domain construction with explicit marginalization of the scattering scale offers a calibrated likelihood that avoids early compression of the data and supplies an independent energy-loss channel. This is a methodological strength that could improve discrimination in low-plane-count or low-contrast regimes. The use of standard Fermi-Eyges and Bethe-Bloch models together with the reported numerical gains over an established baseline constitutes a concrete, falsifiable advance.

major comments (1)
  1. [Geant4 benchmark] Geant4 benchmark section: the reported mean ROC-AUC improvement (0.84-0.86 versus 0.81) is presented without error bars, the number of simulated events, exclusion criteria, or a statistical test of the difference; these omissions are load-bearing for the central empirical claim.
minor comments (2)
  1. [Abstract] Abstract: the description of the six-plane spectrometer for RHMT-E would benefit from an explicit statement of the field strength and plane spacing to allow immediate comparison with existing spectrometers.
  2. [Methods] Notation: the symbol for the electron-density contrast ho Z/A is introduced in the abstract but its precise definition and units should be restated at first use in the methods to avoid ambiguity.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive assessment of the RHMT formulation and for identifying the need for additional statistical rigor in the Geant4 benchmark. We address the single major comment below.

read point-by-point responses

  1. Referee: [Geant4 benchmark] Geant4 benchmark section: the reported mean ROC-AUC improvement (0.84-0.86 versus 0.81) is presented without error bars, the number of simulated events, exclusion criteria, or a statistical test of the difference; these omissions are load-bearing for the central empirical claim.

    Authors: We agree that these details are necessary to substantiate the central empirical claim. The manuscript as submitted reports only the mean ROC-AUC values. In the revised version we will (i) report error bars on the ROC-AUC means (standard error across independent simulation runs or bootstrap resampling), (ii) state the total number of simulated muon events and the number retained after any quality cuts, (iii) describe the exclusion criteria applied to tracks, and (iv) include a statistical test (e.g., paired Wilcoxon signed-rank test on per-run AUCs) together with the resulting p-value for the observed improvement over the ASR baseline. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper introduces RHMT-S and RHMT-E as new measurement-domain likelihood constructions (Fermi-Eyges covariance after track projection with marginal Student-t; Bethe-Bloch line integral on fitted log-momentum loss). Performance is reported as empirical ROC-AUC gains inside a controlled Geant4 benchmark (0.84-0.86 vs 0.81 baseline). No step reduces by definition to a fitted input, no self-citation chain is load-bearing, and no ansatz or uniqueness claim is smuggled in. The benchmark numbers are external to the internal equations and do not collapse to them by construction.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The method rests on standard domain models (Fermi-Eyges multiple-scattering covariance and Bethe-Bloch energy-loss formula) plus marginalization over an unknown per-muon scattering scale; no new particles or forces are introduced.

free parameters (1)
  • unknown scattering scale
    Marginalized in the Student-t-type likelihood of RHMT-S; its distribution is not specified in the abstract.
axioms (2)
  • domain assumption Fermi-Eyges covariance accurately describes hit contrast after projecting out the straight track
    Invoked to evaluate remaining hit contrast in RHMT-S.
  • domain assumption Bethe-Bloch formula gives the line integral of electron-density contrast from log momentum loss
    Used in RHMT-E to model energy-loss contrast.

pith-pipeline@v0.9.1-grok · 5743 in / 1402 out tokens · 27906 ms · 2026-06-26T15:15:28.406805+00:00 · methodology

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

43 extracted references · 1 linked inside Pith

  1. [1]

    Agostinelli et al

    S. Agostinelli et al. GEANT4—a simulation toolkit.Nuclear Instruments and Methods in Physics Research A, 506(3):250–303, 2003

    2003

  2. [2]

    Gradient-descent-based reconstruc- tion for muon tomography based on automatic differentiation in PyTorch.arXiv preprint arXiv:2511.05226, 2025

    Jean-Marco Alameddine, Felix Sattler, Maurice Stephan, and Sarah Barnes. Gradient-descent-based reconstruc- tion for muon tomography based on automatic differentiation in PyTorch.arXiv preprint arXiv:2511.05226, 2025

    arXiv 2025

  3. [3]

    Alvarez, Jared A

    Luis W. Alvarez, Jared A. Anderson, F. El Bedwei, James Burkhard, Ahmed Fakhry, Adib Girgis, Amr Goneid, Fikhry Hassan, Dennis Iverson, Gerald Lynch, Zenab Miligy, Ali Hilmy Moussa, Mohammed Sharkawi, and Lauren Yazolino. Search for hidden chambers in the pyramids.Science, 167(3919):832–839, 1970

    1970

  4. [4]

    Atmospheric muons as an imaging tool

    Lorenzo Bonechi, Raffaello D’Alessandro, and Andrea Giammanco. Atmospheric muons as an imaging tool. Reviews in Physics, 5:100038, 2020

    2020

  5. [5]

    Bonomi, P

    G. Bonomi, P. Checchia, M. D’Errico, D. Pagano, and G. Saracino. Applications of cosmic-ray muons.Progress in Particle and Nuclear Physics, 112:103768, 2020

    2020

  6. [6]

    Borozdin, Gary E

    Konstantin N. Borozdin, Gary E. Hogan, Christopher Morris, William C. Priedhorsky, Alexander Saun- ders, Larry J. Schultz, and Margaret E. Teasdale. Radiographic imaging with cosmic-ray muons.Nature, 422(6929):277–277, 2003

    2003

  7. [7]

    A. P. Dempster, N. M. Laird, and D. B. Rubin. Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society: Series B (Methodological), 39(1):1–22, 1977

    1977

  8. [8]

    Multiple scattering with energy loss.Physical Review, 74(10):1534–1535, 1948

    Leonard Eyges. Multiple scattering with energy loss.Physical Review, 74(10):1534–1535, 1948

    1948

  9. [9]

    A. Sh. Georgadze. Rapid cargo verification with cosmic ray muon scattering and absorption tomography.Journal of Instrumentation, 19(10):P10033, 2024. 9 Raw-Hit Muon Tomography

    2024

  10. [10]

    Differentiable uncalibrated imaging

    Sidharth Gupta, Konik Kothari, Valentin Debarnot, and Ivan Dokmani ´c. Differentiable uncalibrated imaging. IEEE Transactions on Computational Imaging, 10:1–16, 2024

    2024

  11. [11]

    Hagmann, D

    C. Hagmann, D. Lange, and D. Wright. Cosmic-ray shower generator (CRY) for Monte Carlo transport codes. In2007 IEEE Nuclear Science Symposium Conference Record, volume 2, pages 1143–1146, 2007

    2007

  12. [12]

    Highland

    Virgil L. Highland. Some practical remarks on multiple scattering.Nuclear Instruments and Methods, 129(2):497–499, 1975

    1975

  13. [13]

    Peter J. Huber. Robust estimation of a location parameter.The Annals of Mathematical Statistics, 35(1):73–101, 1964

    1964

  14. [14]

    Jonkmans, V

    G. Jonkmans, V . N. P. Anghel, C. Jewett, and M. Thompson. Nuclear waste imaging and spent fuel verification by muon tomography.Annals of Nuclear Energy, 53:267–273, 2013

    2013

  15. [15]

    Kellman, Emrah Bostan, Nicole Repina, and Laura Waller

    Michael R. Kellman, Emrah Bostan, Nicole Repina, and Laura Waller. Physics-based learned design: Optimized coded-illumination for quantitative phase imaging.IEEE Transactions on Computational Imaging, 5(3):344– 353, 2019

    2019

  16. [16]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In3rd International Confer- ence on Learning Representations (ICLR), San Diego, CA, USA, 2015. arXiv:1412.6980

    Pith/arXiv arXiv 2015

  17. [17]

    Lange, Roderick J

    Kenneth L. Lange, Roderick J. A. Little, and Jeremy M. G. Taylor. Robust statistical modeling using thet distribution.Journal of the American Statistical Association, 84(408):881–896, 1989

    1989

  18. [18]

    Image reconstruction techniques in muography: A review of algorithms and physical principles.Journal of Applied Physics, 138:041101, 2025

    Siyuan Luo, Chuntian Feng, Guoqiang Zeng, Shengyang Feng, Mao Shen, Xuankai Huang, Lizhi Wang, Shuhao Zhao, Xuecheng Du, Song Feng, Min Xiao, Zhiyi Liu, and Xiaodong Wang. Image reconstruction techniques in muography: A review of algorithms and physical principles.Journal of Applied Physics, 138:041101, 2025

    2025

  19. [19]

    G. R. Lynch and O. I. Dahl. Approximations to multiple Coulomb scattering.Nuclear Instruments and Methods in Physics Research B, 58:6–10, 1991

    1991

  20. [20]

    Moli `ere

    G. Moli `ere. Theorie der Streuung schneller geladener Teilchen II: Mehrfach- und Vielfachstreuung.Zeitschrift f¨ur Naturforschung A, 3(2):78–97, 1948

    1948

  21. [21]

    Discovery of a big void in Khufu’s Pyramid by observation of cosmic-ray muons

    Kunihiro Morishima, Mitsuaki Kuno, Akira Nishio, Nobuko Kitagawa, Yuta Manabe, Masaki Moto, Fumihiko Takasaki, Hirofumi Fujii, Kotaro Satoh, Hideyo Kodama, Kohei Hayashi, Shigeru Odaka, S ´ebastien Procureur, David Atti´e, Simon Bouteille, Denis Calvet, Christopher Filosa, Patrick Magnier, Irakli Mandjavidze, Marc Ri- allot, Benoit Marini, Pierre Gable, Y...

    2017

  22. [22]

    C. L. Morris, C. C. Alexander, J. D. Bacon, K. N. Borozdin, D. J. Clark, R. Chartrand, C. J. Espinoza, A. M. Fraser, M. C. Galassi, J. A. Green, J. S. Gonzales, J. J. Gomez, N. W. Hengartner, G. E. Hogan, A. V . Klimenko, M. F. Makela, P. McGaughey, J. J. Medina, F. E. Pazuchanics, W. C. Priedhorsky, J. C. Ramsey, A. Saunders, R. C. Schirato, L. J. Schult...

    2008

  23. [23]

    A new method for structural diagnostics with muon tomography and deep learning.Journal of Instrumentation, 20(6):P06034, 2025

    Lorenzo Pezzotti, Davide Cifarelli, Daniele Corradetti, Jos ´e Paulo Costa, Giorgio Gabrielli, Lorenzo Galante, Antonio Gallerati, Ivan Gnesi, Andrea Jouve, and Alessio Marrani. A new method for structural diagnostics with muon tomography and deep learning.Journal of Instrumentation, 20(6):P06034, 2025

    2025

  24. [24]

    Priedhorsky, Konstantin N

    William C. Priedhorsky, Konstantin N. Borozdin, Gary E. Hogan, Christopher Morris, Alexander Saunders, Larry J. Schultz, and Margaret E. Teasdale. Detection of high-zobjects using multiple scattering of cosmic ray muons.Review of Scientific Instruments, 74(10):4294–4297, 2003

    2003

  25. [25]

    Muon imaging: Principles, technologies and applications.Nuclear Instruments and Meth- ods in Physics Research A, 878:169–179, 2018

    S ´ebastien Procureur. Muon imaging: Principles, technologies and applications.Nuclear Instruments and Meth- ods in Physics Research A, 878:169–179, 2018

    2018

  26. [26]

    Millimeter-resolution cosmic-ray imaging via projection-shifted muon transmission tomography

    Zibo Qin, Rongfeng Zhang, Pei Yu, Cheng-en Liu, Liangwen Chen, Feng Zhang, Zaihong Yang, Qite Li, and Qiang Li. Millimeter-resolution cosmic-ray imaging via projection-shifted muon transmission tomography. arXiv preprint arXiv:2512.19747, 2025

    arXiv 2025

  27. [27]

    Rudin, Stanley Osher, and Emad Fatemi

    Leonid I. Rudin, Stanley Osher, and Emad Fatemi. Nonlinear total variation based noise removal algorithms. Physica D: Nonlinear Phenomena, 60(1-4):259–268, 1992

    1992

  28. [28]

    Santonico and R

    R. Santonico and R. Cardarelli. Development of resistive plate counters.Nuclear Instruments and Methods, 187:377–380, 1981. 10 Raw-Hit Muon Tomography

    1981

  29. [29]

    Schultz, Gary S

    Larry J. Schultz, Gary S. Blanpied, Konstantin N. Borozdin, Andrew M. Fraser, Nicolas W. Hengartner, Alexei V . Klimenko, Christopher L. Morris, Chris Orum, and Michael J. Sossong. Statistical reconstruction for cosmic ray muon tomography.IEEE Transactions on Image Processing, 16(8):1985–1993, 2007

    1985

  30. [30]

    Schultz, Konstantin N

    Larry J. Schultz, Konstantin N. Borozdin, John J. Gomez, Gary E. Hogan, J. A. McGill, Christopher L. Morris, William C. Priedhorsky, Alexander Saunders, and Margaret E. Teasdale. Image reconstruction and materialz discrimination via cosmic ray muon radiography.Nuclear Instruments and Methods in Physics Research A, 519(3):687–694, 2004

    2004

  31. [31]

    Stapleton, J

    M. Stapleton, J. Burns, S. Quillin, and C. Steer. Angle statistics reconstruction: a robust reconstruction algorithm for muon scattering tomography.Journal of Instrumentation, 9(11):P11019, 2014

    2014

  32. [32]

    Giles C. Strong, Maxime Lagrange, Aitor Orio, Anna Bordignon, Florian Bury, Tommaso Dorigo, Andrea Gi- ammanco, Mariam Heikal, Jan Kieseler, Max Lamparth, Pablo Mart´ınez Ru´ız del ´Arbol, Federico Nardi, Pietro Vischia, and Haitham Zaraket. TomOpt: differential optimisation for task- and constraint-aware design of par- ticle detectors in the context of m...

    2024

  33. [33]

    Hiroyuki K. M. Tanaka, Toshiyuki Nakano, Satoru Takahashi, Jun Yoshida, Hiroshi Ohshima, Tokumitsu Maekawa, Hiroshi Watanabe, and Kimio Niwa. Imaging the conduit size of the dome with cosmic-ray muons: The structure beneath Showa-Shinzan Lava Dome, Japan.Geophysical Research Letters, 34(22):L22311, 2007

    2007

  34. [34]

    Reshma Ughade and Stylianos Chatzidakis.µTRec: A muon trajectory reconstruction algorithm for enhanced scattering tomography.Journal of Applied Physics, 138(6):064909, 2025

    2025

  35. [35]

    Non-intrusive monitoring of sealed microreactor cores using physics-informed muon scattering tomography with momentum measurements.arXiv preprint arXiv:2603.05712, 2026

    Reshma Ughade and Stylianos Chatzidakis. Non-intrusive monitoring of sealed microreactor cores using physics-informed muon scattering tomography with momentum measurements.arXiv preprint arXiv:2603.05712, 2026

    arXiv 2026

  36. [36]

    U-Net based image enhancement for short-time muon scattering tomography.arXiv preprint arXiv:2602.07060, 2026

    Haochen Wang, Pei Yu, Liangwen Chen, Weibo He, Yu Zhang, Yuhong Yu, Xueheng Zhang, Lei Yang, and Zhiyu Sun. U-Net based image enhancement for short-time muon scattering tomography.arXiv preprint arXiv:2602.07060, 2026

    arXiv 2026

  37. [37]

    On scale mixtures of normal distributions.Biometrika, 74(3):646–648, 1987

    Mike West. On scale mixtures of normal distributions.Biometrika, 74(3):646–648, 1987

    1987

  38. [38]

    R. L. Workman et al. Review of Particle Physics.Progress of Theoretical and Experimental Physics, 2022(8):083C01, 2022

    2022

  39. [39]

    Zimmermann, Christoph Kolbitsch, Patrick Schuenke, and Andreas Kofler

    Felix F. Zimmermann, Christoph Kolbitsch, Patrick Schuenke, and Andreas Kofler. PINQI: An end-to-end physics-informed approach to learned quantitative MRI reconstruction.IEEE Transactions on Computational Imaging, 10:628–639, 2024. 11 Raw-Hit Muon Tomography APPENDIX A Scattering-channel derivation This appendix derives the scattering observable used in S...

    2024

  40. [40]

    A, B) to get the contrastw i (RHMT-S) or the log-loss residualr i (RHMT-E) and the chord

    fit each muon’s hits (Apps. A, B) to get the contrastw i (RHMT-S) or the log-loss residualr i (RHMT-E) and the chord

  41. [41]

    sample the field trilinearly along the chord at the quadrature nodes, giving a per-muon linear operator

  42. [42]

    (4) or (6)] by Adam [16] gradient descent

    minimise the penalised objective [Eq. (4) or (6)] by Adam [16] gradient descent

  43. [43]

    mask voxels crossed by fewer thanmin covchords and column-project to the scoring grid. Every nuisance width is fixed beforehand on an object-free blank by the same pipeline:(gref , ν)maximise (11) on the blank contrasts, andµ ℓ(x, y, p), σℓ(p)are the per-momentum-bin Huber location and scale (Lemma 1). Nothing is fit on the object. D Simulation and benchm...