arxiv: 2604.13503 · v1 · submitted 2026-04-15 · ✦ hep-ex · physics.comp-ph

Recognition: unknown

Enhancing Event Reconstruction in Hyper-Kamiokande with Machine Learning: A ResNet Implementation

Andrew Atta, Eric Thrane, Kimihiro Okumura, Nick Prouse, Patrick de Perio, Phillip Urquijo, Shuoyu Chen

Authors on Pith no claims yet

Pith reviewed 2026-05-10 12:45 UTC · model grok-4.3

classification ✦ hep-ex physics.comp-ph

keywords Hyper-Kamiokandemachine learningevent reconstructionResNetneutrino detectorsparticle identificationMonte Carlo simulation

0 comments

The pith

ResNet neural networks match traditional Hyper-Kamiokande reconstruction accuracy while running thousands of times faster on simulated single-particle events.

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

The paper shows that deep learning can handle the huge Monte Carlo samples needed for Hyper-Kamiokande's systematic studies by replacing slow likelihood fitting with fast neural-network regression and classification. Single-particle events are turned into 190 by 189 pixel images of photomultiplier charge and timing, then fed to ResNet models that both identify the particle type and estimate its vertex, direction, and momentum. The networks deliver momentum, angle, and vertex resolutions for muons and electrons that stay within the range achieved by conventional methods. At the same time they classify electrons against muons, photons, and neutral pions with high separation power. The decisive practical gain is inference speed: 1-2 milliseconds per event on a GPU, which is tens of thousands of times quicker than the likelihood approach.

Core claim

Averaged over the full kinematic range from Cherenkov threshold to 2 GeV, the regression models achieve momentum resolutions of 1.35% and 2.39%, angular resolutions of 1.25° and 1.94°, and vertex resolutions of 28.2 cm and 25.4 cm for muons and electrons respectively, broadly consistent with traditional methods. The classifier improves e-μ, e-γ, and e-π⁰ separation with ROC areas of 0.9999992, 0.633, and 0.9526. The networks run at 1-2 ms per event on a single GPU, delivering speed-ups of 3.2×10⁴ to 5.2×10⁴ relative to likelihood-based reconstruction.

What carries the argument

ResNet models inside the WatChMaL framework that take two-channel 190×189 images encoding photomultiplier-tube charge and timing as input and output both particle-type probabilities and regression values for vertex, direction, and momentum.

If this is right

Large-scale Monte Carlo campaigns required for Hyper-Kamiokande systematic uncertainty budgets become computationally feasible.
Event-by-event processing can move from offline clusters to near-real-time GPU hardware.
Particle identification between electrons, muons, photons, and neutral pions improves for single-ring events in the 0-2 GeV range.

Where Pith is reading between the lines

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

The same image-based approach could be extended to multi-particle final states or to events that include Michel electrons or other delayed activity.
Once trained, the networks could serve as a fast pre-filter that passes only ambiguous events to the slower likelihood fitter, combining the speed of machine learning with the precision of traditional methods.
Similar image-to-parameter networks might be portable to other large water-Cherenkov detectors facing comparable data-volume pressures.

Load-bearing premise

That the accuracy measured on Monte Carlo simulations will transfer to real detector data without large losses caused by unmodeled effects such as calibration drifts or noise.

What would settle it

A side-by-side comparison of the neural-network and likelihood reconstructions on the same set of real Hyper-Kamiokande calibration data, checking whether resolution and classification performance degrade beyond the Monte-Carlo expectations.

Figures

Figures reproduced from arXiv: 2604.13503 by Andrew Atta, Eric Thrane, Kimihiro Okumura, Nick Prouse, Patrick de Perio, Phillip Urquijo, Shuoyu Chen.

**Figure 1.** Figure 1: Example 190 × 189 input image used for network training and evaluation. Image is of the integrated-charge channel of a µ event. A second channel stores the PMT hit time relative to the start of the 1.35 µs readout window, but is not shown here. Charge is in photo-electrons. Training and validation samples were produced with the open-source package WCSim [24]. The full simulation configuration, including … view at source ↗

**Figure 3.** Figure 3: Vertex-reconstruction resolution for the ResNet-152 models (blue) and fiTQun results from Super-Kamiokande-IV [29] (red). Reconstruction performance for muons (solid) and electrons (dashed) are shown. The resolution is defined as the three-dimensional Euclidean distance below which 68% of events fall per 100 MeV/c true momentum bin. Shaded bands indicate 95% bootstrap confidence intervals on the ResNet res… view at source ↗

**Figure 4.** Figure 4: Direction reconstruction performance of the ResNet-152 models for [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗

**Figure 5.** Figure 5: ROC curves for electron signal versus background from other particle [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

**Figure 6.** Figure 6: Electron selection efficiency at fixed background-rejection operating points of 99.9% for µ, 80% for γ, and 95% for π 0 candidates. The blue curve corresponds to the (e + γ + π 0 ) signal versus µ background discriminator, the green curve to the (e + γ) signal versus π 0 background, and the orange curve to the e signal versus γ background, as defined in Eqs. 2–4. The top panel shows the efficiency as a fun… view at source ↗

read the original abstract

The forthcoming Hyper-Kamiokande experiment requires substantially larger Monte Carlo datasets than previous experiments to satisfy stringent systematic-uncertainty requirements. While traditional maximum-likelihood reconstruction provides high-quality results, its per-event computational cost makes processing these large samples increasingly impractical. We demonstrate a neural-network-based reconstruction approach for the Hyper-Kamiokande far detector using simulated data. Single-particle events with kinetic energies from the Cherenkov threshold up to 2 GeV are propagated through the detector, with PMT charge and timing information mapped to $190\times189$ two-channel images serving as inputs to ResNet models in the WatChMaL framework. These models (i) classify events into four particle hypotheses ($e$, $\mu$, $\gamma$, $\pi^{0}$) and (ii) regress the vertex, direction, and momentum of electrons and muons. Averaged over the full kinematic range, the regression models achieve momentum resolutions of $1.35\%$ and $2.39\%$, angular resolutions of $1.25^\circ$ and $1.94^\circ$, and vertex resolutions of $28.2$ cm and $25.4$ cm, for muons and electrons respectively, broadly consistent with traditional methods. The classifier improves $e$-$\mu$, $e$-$\gamma$, and $e$-$\pi^{0}$ separation, with ROC curve areas of $0.9999992$, $0.633$, and $0.9526$. Crucially, our networks achieve inference times of 1-2 ms per event on a single GPU, yielding speed-ups of $3.2\times10^{4}$-$5.2\times10^{4}$ relative to likelihood-based reconstruction, highlighting deep learning as a scalable alternative for Hyper-Kamiokande event reconstruction.

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.

Desk Editor's Note private letter to a colleague

ResNet matches traditional resolutions on Hyper-K MC with huge speed gains, but everything rests on simulation and real-data transfer is untested.

read the letter

The main point is that a ResNet inside WatChMaL produces momentum, angle, and vertex resolutions on single-particle Hyper-K Monte Carlo that sit close to likelihood-based reconstruction while running 30,000 to 50,000 times faster per event on a GPU. The four-class particle ID also shows strong e-mu separation and decent gamma and pi0 performance on the same simulated sample. Those concrete numbers and the speed claim are what the paper actually adds for this detector geometry and energy range up to 2 GeV. The work is straightforward and internally consistent on the MC side. They take charge and timing hits, turn them into 190 by 189 two-channel images, and train the regression and classification heads separately. The quoted resolutions (1.35 percent momentum for muons, 1.25 degrees angle, 28 cm vertex) line up with what traditional methods achieve on the identical events, and the inference time of 1-2 ms is a clear practical win for the multi-billion-event samples Hyper-K needs. The soft spot is that none of this has been checked on real data. The paper gives no tests with PMT calibration drifts, water quality changes, or electronics noise that the simulation may not capture perfectly. Without that step, the speed advantage stays theoretical for actual analysis use. Training details such as validation splits and regularization are also missing from the abstract, though the results do not look overfit on the reported metrics. This is useful reading for the Hyper-K reconstruction team and for anyone running large water-Cherenkov Monte Carlo campaigns who needs a faster alternative to full likelihood fits. A reader already familiar with WatChMaL or similar image-based approaches will pick up the implementation choices and the exact speed-up factors quickly. The paper deserves a serious referee because the claims are narrow, the method is reproducible on the provided MC, and the central speed question is easy to verify or challenge with additional tests.

Referee Report

2 major / 1 minor

Summary. The manuscript describes the application of ResNet models within the WatChMaL framework to reconstruct single-particle events in the Hyper-Kamiokande detector using simulated data. PMT charge and timing information is converted into 190×189 two-channel images, which are used to classify events into e, μ, γ, π⁰ hypotheses and to regress vertex position, direction, and momentum for electrons and muons. The reported performance includes momentum resolutions of 1.35% (muons) and 2.39% (electrons), angular resolutions of 1.25° and 1.94°, vertex resolutions of 28.2 cm and 25.4 cm, and high ROC areas for classification, with inference times of 1-2 ms per event providing speed-ups of 3.2×10⁴ to 5.2×10⁴ over likelihood-based methods.

Significance. If the reported performance holds under realistic conditions, the approach could enable efficient processing of the very large Monte Carlo samples needed for Hyper-Kamiokande systematic studies, providing a practical complement to traditional reconstruction.

major comments (2)

[Abstract] Abstract: the quoted momentum, angular, and vertex resolutions are presented without any description of the training/validation split, regularization strategy, or whether the figures include systematic uncertainties arising from the Monte Carlo simulation itself.
[Results] Results section: all resolution and ROC metrics are obtained solely on single-particle Monte Carlo events; no tests with real detector data or with injected unmodeled effects (PMT calibration drifts, water attenuation variations, electronics noise) are reported, which is load-bearing for the claim that the method constitutes a scalable alternative for Hyper-Kamiokande operations.

minor comments (1)

[Methods] Clarify the precise ResNet architecture (number of layers, residual blocks) and any data-augmentation or preprocessing steps applied to the 190×189 images.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful review and constructive feedback. We address each major comment below and have revised the manuscript to improve clarity and acknowledge limitations.

read point-by-point responses

Referee: [Abstract] Abstract: the quoted momentum, angular, and vertex resolutions are presented without any description of the training/validation split, regularization strategy, or whether the figures include systematic uncertainties arising from the Monte Carlo simulation itself.

Authors: We agree that the abstract would benefit from additional context. In the revised manuscript we have added a brief description of the training/validation split used on the simulated dataset, the regularization strategy employed, and clarified that the quoted resolutions represent statistical uncertainties obtained from the Monte Carlo sample without incorporation of systematic effects from the simulation model. Expanded details on these aspects remain in the Methods section. revision: yes
Referee: [Results] Results section: all resolution and ROC metrics are obtained solely on single-particle Monte Carlo events; no tests with real detector data or with injected unmodeled effects (PMT calibration drifts, water attenuation variations, electronics noise) are reported, which is load-bearing for the claim that the method constitutes a scalable alternative for Hyper-Kamiokande operations.

Authors: The present work benchmarks the ResNet approach on controlled single-particle Monte Carlo events, which is the standard first step for developing and validating reconstruction algorithms prior to deployment. We have revised the Results and Conclusions sections to explicitly note this scope and to state that the demonstrated inference speed-up is intended to address the computational demands of generating the large Monte Carlo samples required for Hyper-Kamiokande systematic studies. We agree that validation against real data and robustness to unmodeled detector effects are essential for operational use and have added text identifying these as priorities for future work. The current results support the claim of a practical complement for MC processing as described in the introduction. revision: partial

Circularity Check

0 steps flagged

No circularity: performance metrics are direct empirical comparisons on held-out Monte Carlo

full rationale

The paper trains ResNet models on simulated single-particle events mapped to 190x189 charge-timing images and reports regression resolutions and classification ROC areas obtained by direct comparison to truth labels and to likelihood-based reconstruction on the same held-out MC sample. No parameter is fitted to a target quantity and then renamed as a prediction, no self-citation supplies a load-bearing uniqueness theorem or ansatz, and the speed-up claim is a straightforward wall-clock timing measurement. All central numbers are therefore independent of the paper's own outputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The performance claims rest on the assumption that Monte Carlo simulations faithfully reproduce real detector response and that the network generalizes beyond the training distribution.

axioms (1)

domain assumption Monte Carlo simulations accurately model real PMT charge and timing response across the full kinematic range
All training and evaluation is performed exclusively on simulated events; no real-data validation is mentioned.

pith-pipeline@v0.9.0 · 5658 in / 1322 out tokens · 43010 ms · 2026-05-10T12:45:04.644844+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

45 extracted references · 31 canonical work pages · 3 internal anchors

[1]

Hyper-Kamiokande Design Report

K. Abe, et al., Hyper-Kamiokande Design Report (5 2018). arXiv:1805.04163

work page Pith review arXiv 2018
[2]

Abe, et al., Physics potential of a long-baseline neutrino oscillation experiment using a J-PARC neutrino beam and Hyper-Kamiokande, PTEP 2015 (2015) 053C02

K. Abe, et al., Physics potential of a long-baseline neutrino oscillation experiment using a J-PARC neutrino beam and Hyper-Kamiokande, PTEP 2015 (2015) 053C02. arXiv: 1502.05199,doi:10.1093/ptep/ptv061

work page doi:10.1093/ptep/ptv061 2015
[3]

Abe, et al., Sensitivity of the Hyper-Kamiokande exper- iment to neutrino oscillation parameters using accelerator neutrinos, Eur

K. Abe, et al., Sensitivity of the Hyper-Kamiokande exper- iment to neutrino oscillation parameters using accelerator neutrinos, Eur. Phys. J. C 86 (2026) 170. arXiv:2505. 15019,doi:10.1140/epjc/s10052-025-14938-9

work page doi:10.1140/epjc/s10052-025-14938-9 2026
[4]

M. Koshiba, et al., KAMIOKANDE — Kamioka Nucleon Decay Experiment: Status and Performance, in: Low En- ergy Tests of Conservation Laws in Particle Physics, AIP Conference Proceedings, 1984

1984
[5]

Fukuda et al

Y . Fukuda, et al., Evidence for Oscillation of At- mospheric Neutrinos, Phys. Rev. Lett. 81 (1998) 1562–1567. arXiv:hep-ex/9807003, doi:10.1103/ PhysRevLett.81.1562

work page arXiv 1998
[6]

Fukuda, et al., The Super-Kamiokande detector, Nucl

S. Fukuda, et al., The Super-Kamiokande detector, Nucl. Instrum. Meth. A 501 (2003) 418–462. doi:10.1016/ S0168-9002(03)00425-X

2003
[7]

Abe, et al., The T2K Experiment, Nucl

K. Abe, et al., The T2K Experiment, Nucl. Instrum. Meth. A 659 (2011) 106–135. arXiv:1106.1238, doi: 10.1016/j.nima.2011.06.067

work page doi:10.1016/j.nima.2011.06.067 2011
[8]

Abe, et al., Measurements of neutrino oscillation param- eters from the T2K experiment using 3 .6 × 1021 protons on target, Eur

K. Abe, et al., Measurements of neutrino oscillation param- eters from the T2K experiment using 3 .6 × 1021 protons on target, Eur. Phys. J. C 83 (9) (2023) 782. arXiv:2303. 03222,doi:10.1140/epjc/s10052-023-11819-x

work page doi:10.1140/epjc/s10052-023-11819-x 2023
[9]

Abe, et al., J-PARC Neutrino Beamline Upgrade Tech- nical Design Report (2019).arXiv:1908.05141

K. Abe, et al., J-PARC Neutrino Beamline Upgrade Tech- nical Design Report (2019).arXiv:1908.05141

work page arXiv 2019
[10]

P. A. ˇCerenkov, Visible Radiation Produced by Electrons Moving in a Medium with Velocities Exceeding that of Light, Phys. Rev. 52 (1937) 378–379. doi:10.1103/ PhysRev.52.378

1937
[11]

Electromagnetic effects and weak form factors

O. Lalakulich, U. Mosel, K. Gallmeister, Neutrino energy reconstruction in quasielastic-like scattering in the Mini- BooNE and T2K experiments, Phys. Rev. C 86 (2012) 054606. arXiv:1208.3678, doi:10.1103/PhysRevC. 86.054606

work page doi:10.1103/physrevc 2012
[12]

Kose, The Hyper-Kamiokande experiment: Status and prospect, SciPost Phys

U. Kose, The Hyper-Kamiokande experiment: Status and prospect, SciPost Phys. Proc. 17 (2025) 019. doi:10. 21468/SciPostPhysProc.17.019

2025
[13]

Interferometric signatures of black holes with multiple photon spheres.Phys

S. Navas, et al., Review of Particle Physics, Phys. Rev. D 110 (3) (2024) 030001. doi:10.1103/PhysRevD.110. 030001

work page doi:10.1103/physrevd.110 2024
[14]

S. M. Bilenky, B. Pontecorvo, Lepton Mixing and Neutrino Oscillations, Phys. Rept. 41 (1978) 225–261. doi:10. 1016/0370-1573(78)90095-9

1978
[15]

Prouse, P

N. Prouse, P. de Perio, W. Fedorko, Machine Learning Techniques to Enhance Event Reconstruction in Water Cherenkov Detectors, Phys. Sci. Forum 8 (1) (2023) 63. doi:10.3390/psf2023008063

work page doi:10.3390/psf2023008063 2023
[16]

Krizhevsky, I

A. Krizhevsky, I. Sutskever, G. E. Hinton, ImageNet Classi- fication with Deep Convolutional Neural Networks, in: Ad- vances in Neural Information Processing Systems, V ol. 25, 2012

2012
[17]

K. He, X. Zhang, S. Ren, J. Sun, Deep Residual Learning for Image Recognition, in: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 770–778.doi:10.1109/CVPR.2016.90

work page doi:10.1109/cvpr.2016.90 2016
[18]

Yankelevich, Machine Learning Methods for Super- Kamiokande Solar Neutrino Classification, Phys

A. Yankelevich, Machine Learning Methods for Super- Kamiokande Solar Neutrino Classification, Phys. Sci. Fo- rum 8 (1) (2023) 42.doi:10.3390/psf2023008042. 10

work page doi:10.3390/psf2023008042 2023
[19]

Jamieson, M

B. Jamieson, M. Stubbs, S. Ramanna, J. Walker, N. Prouse, R. Akutsu, P. de Perio, W. Fedorko, Using Machine Learn- ing to Improve Neutron Identification in Water Cherenkov Detectors, Front. Big Data 5 (2022) 978857. arXiv: 2206.12954,doi:10.3389/fdata.2022.978857

work page doi:10.3389/fdata.2022.978857 2022
[20]

Aguilar, et al., Classification of electron and muon neutrino events for the ESS νSB near water Cherenkov detector using Graph Neural Networks, JINST 20 (08) (2025) P08030

J. Aguilar, et al., Classification of electron and muon neutrino events for the ESS νSB near water Cherenkov detector using Graph Neural Networks, JINST 20 (08) (2025) P08030. arXiv:2503.15247, doi:10.1088/ 1748-0221/20/08/P08030

work page arXiv 2025
[21]

Qian, et al., Vertex and energy reconstruction in JUNO with machine learning methods, Nucl

Z. Qian, et al., Vertex and energy reconstruction in JUNO with machine learning methods, Nucl. Instrum. Meth. A 1010 (2021) 165527. arXiv:2101.04839, doi:10. 1016/j.nima.2021.165527

work page arXiv 2021
[22]

Aurisano, et al., A Convolutional Neu- ral Network Neutrino Event Classifier, JINST 11 (09) (2016) P09001

A. Aurisano, et al., A Convolutional Neu- ral Network Neutrino Event Classifier, JINST 11 (09) (2016) P09001. arXiv:1604.01444, doi:10.1088/1748-0221/11/09/P09001

work page doi:10.1088/1748-0221/11/09/p09001 2016
[23]

WatChMaL Collaboration, WatChMaL: Water Cherenkov Machine Learning, https://github.com/WatChMaL (2025)

2025
[24]

WCSim Collaboration, WCSim: Water Cherenkov Simula- tor,https://github.com/WCSim/WCSim(2025)

2025
[25]

P. J. Huber, Robust Estimation of a Location Parameter, Ann. Math. Statist. 35 (1) (1964) 73–101. doi:10.1214/ aoms/1177703732

work page arXiv 1964
[26]

K. He, X. Zhang, S. Ren, J. Sun, Delving Deep into Rectifiers: Surpassing Human-Level Performance on Im- ageNet Classification, in: IEEE International Conference on Computer Vision (ICCV 2015), 2015, pp. 1026–1034. doi:10.1109/ICCV.2015.123

work page doi:10.1109/iccv.2015.123 2015
[27]

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

S. Ioffe, C. Szegedy, Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift, in: Proceedings of the 32nd International Conference on Machine Learning, 2015.arXiv:1502.03167

work page internal anchor Pith review arXiv 2015
[28]

Decoupled Weight Decay Regularization

I. Loshchilov, F. Hutter, Decoupled Weight Decay Regular- ization, in: International Conference on Learning Repre- sentations, 2019.arXiv:1711.05101

work page internal anchor Pith review Pith/arXiv arXiv 2019
[29]

Jiang, et al., Atmospheric Neutrino Oscillation Analysis with Improved Event Reconstruction in Super-Kamiokande IV, PTEP 2019 (5) (2019) 053F01

M. Jiang, et al., Atmospheric Neutrino Oscillation Analysis with Improved Event Reconstruction in Super-Kamiokande IV, PTEP 2019 (5) (2019) 053F01. arXiv:1901.03230, doi:10.1093/ptep/ptz015

work page doi:10.1093/ptep/ptz015 2019
[30]

Gao, The Calibration of the Super-Kamiokande Outer- Detector and Particle Kinematics Reconstruction in Hyper- Kamiokande, Ph.D

J. Gao, The Calibration of the Super-Kamiokande Outer- Detector and Particle Kinematics Reconstruction in Hyper- Kamiokande, Ph.D. thesis, King’s College London (2024)

2024
[31]

I. M. Frank, I. E. Tamm, Coherent visible radiation of fast electrons passing through matter, Compt. Rend. Acad. Sci. URSS 14 (3) (1937) 109–114. doi:10.3367/UFNr. 0093.196710o.0388

work page doi:10.3367/ufnr 1937
[32]

Canetti, M

L. Canetti, M. Drewes, M. Shaposhnikov, Matter and anti- matter in the universe, New J. Phys. 14 (2012) 095012. arXiv:1204.4186, doi:10.1088/1367-2630/14/9/ 095012

work page doi:10.1088/1367-2630/14/9/ 2012
[33]

A. D. Sakharov, Violation of CP Invariance, C asym- metry, and baryon asymmetry of the universe, Pisma Zh. Eksp. Teor. Fiz. 5 (1967) 32–35. doi:10.1070/ PU1991v034n05ABEH002497

1967
[34]

A. D. Sakharov, Violation of CP invariance, C asym- metry, and baryon asymmetry of the universe, Sov. Phys. Usp. 34 (5) (1991) 392–393. doi:10.1070/ PU1991v034n05ABEH002497

1991
[35]

Fukugita, T

M. Fukugita, T. Yanagida, Baryogenesis Without Grand Unification, Phys. Lett. B 174 (1986) 45–47. doi:10. 1016/0370-2693(86)91126-3

1986
[36]

Z. Maki, M. Nakagawa, S. Sakata, Remarks on the Uni- fied Model of Elementary Particles, Prog. Theor. Phys. 28 (1962) 870–880.doi:10.1143/PTP.28.870

work page doi:10.1143/ptp.28.870 1962
[37]

Pontecorvo, Neutrino Experiments and the Problem of Conservation of Leptonic Charge, Sov

B. Pontecorvo, Neutrino Experiments and the Problem of Conservation of Leptonic Charge, Sov. Phys. JETP 26 (1968) 984–988

1968
[38]

Kobayashi, T

M. Kobayashi, T. Maskawa, CP-Violation in the Renormal- izable Theory of Weak Interaction, Prog. Theor. Phys. 49 (1973) 652–657.doi:10.1143/PTP.49.652

work page doi:10.1143/ptp.49.652 1973
[39]

M. A. Acero, et al., New constraints on oscillation parame- ters from νe appearance and νµ disappearance in the NOvA experiment, Phys. Rev. D 98 (3) (2018) 032012. arXiv: 1806.00096,doi:10.1103/PhysRevD.98.032012

work page doi:10.1103/physrevd.98.032012 2018
[40]

M. A. Acero, et al., First Measurement of Neu- trino Oscillation Parameters using Neutrinos and An- tineutrinos by NOvA, Phys. Rev. Lett. 123 (15) (2019) 151803. arXiv:1906.04907, doi:10.1103/ PhysRevLett.123.151803

work page arXiv 2019
[41]

NuFit-6.0: Updated global analysis of three-flavor neutrino oscillations

I. Esteban, M. C. Gonzalez-Garcia, M. Maltoni, T. Schwetz, A. Zhou, NuFit-6.0: updated global analysis of three-flavor neutrino oscillations, JHEP 09 (2024) 178. arXiv:2410.05380, doi:10.1007/ JHEP09(2024)178

work page internal anchor Pith review arXiv 2024
[42]

Abubakar et al

S. Abubakar, et al., Joint neutrino oscillation analysis from the T2K and NOvA experiments, Nature 646 (8086) (2025) 818–824. arXiv:2510.19888, doi:10.1038/ s41586-025-09599-3

work page arXiv 2025
[43]

R. B. Patterson, E. M. Laird, Y . Liu, P. D. Meyers, I. Stancu, H. A. Tanaka, The Extended-track event reconstruction for MiniBooNE, Nucl. Instrum. Meth. A 608 (2009) 206–

2009
[44]

arXiv:0902.2222, doi:10.1016/j.nima.2009. 06.064. 11

work page doi:10.1016/j.nima.2009 2009
[45]

Sakharov conditions

A. D. Missert, Improving the T2K Oscillation Analysis With fiTQun: A New Maximum-Likelihood Event Recon- struction for Super-Kamiokande, in: J. Phys. Conf. Ser., V ol. 888, 2017, p. 012066. doi:10.1088/1742-6596/ 888/1/012066. Appendix A. Physics Motivation Understanding the matter–antimatter asymmetry of the Uni- verse [32] remains one of the foremost ch...

work page doi:10.1088/1742-6596/ 2017