pith. sign in

arxiv: 2606.18403 · v1 · pith:YSG2ADKCnew · submitted 2026-06-16 · ✦ hep-ph

Stress testing of fast reconstruction pipelines using machine learning

Swagata Ghosh This is my paper

Pith reviewed 2026-06-26 23:31 UTC · model grok-4.3

classification ✦ hep-ph
keywords fast reconstructionmachine learningstress testinglocal assumptionregime mappingZ boson decayHL-LHCdetector simulation
0
0 comments X

The pith

The Z → ℓℓ decay violates the local assumption in fast reconstruction pipelines, causing bias and resolution loss that unsupervised regime mapping corrects.

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

Fast reconstruction pipelines rely on simplified models assuming that reconstruction uncertainty depends only on local input parameters rather than the full global data context. The paper introduces a domain-agnostic context-aware stress testing pipeline that relaxes this assumption and permits reconstruction responses to depend on global parameters. Applied to the Z → ℓℓ benchmark at HL-LHC conditions, the test shows that the local assumption fails, producing measurable reconstruction bias and degraded resolution. An unsupervised regime-mapping framework is then applied to restore peak stability and recover the original truth-level resolution. The overall result supplies a diagnostic method for validating and improving next-generation fast simulation pipelines across domains.

Core claim

The decay channel Z → ℓℓ, as a benchmark, violates the local assumption resulting in a significant reconstruction bias and resolution degradation. Using the unsupervised regime-mapping framework, this work also restores this peak stability and recovers the truth-level resolution, where a robust diagnostic tool for next-generation fast simulation pipelines is accommodated.

What carries the argument

Domain-agnostic context-aware stress testing pipeline that isolates global-parameter dependence in reconstruction response, paired with unsupervised regime-mapping for correction.

If this is right

  • The local assumption breaks for Z → ℓℓ, producing measurable bias and resolution degradation.
  • Unsupervised regime mapping restores peak stability and truth-level resolution.
  • The pipeline acts as a diagnostic that flags global-parameter violations in fast simulation.
  • The approach applies to any fast reconstruction pipeline that currently relies on local response models.

Where Pith is reading between the lines

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

  • The same stress-testing method could be run on other common HEP channels to map which processes are most sensitive to global effects.
  • Medical imaging pipelines that use analogous local approximations might benefit from identical regime-mapping corrections.
  • Embedding the stress test as a standard validation step during pipeline development could catch global dependencies before deployment.
  • Identifying the specific global parameters that drive the Z → ℓℓ violation would allow targeted improvements to the underlying response model.

Load-bearing premise

The introduced domain-agnostic context-aware stress testing pipeline accurately isolates global-parameter dependence without introducing its own reconstruction artifacts or selection biases.

What would settle it

Reconstructed Z mass peak position and width remaining unchanged when the stress test varies global parameters compared with the standard local model.

Figures

Figures reproduced from arXiv: 2606.18403 by Swagata Ghosh.

Figure 1
Figure 1. Figure 1: Left : The dilepton invariant mass distribution mℓℓ within 60 − 120 GeV reveals the Z-boson resonance peak. Right : Scatter plot of invariant mass mℓℓ as a function of the global context scale HT [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Left : Distribution of the scalar sum of lepton transverse momenta (HT ). This variable dictates the energy scale and is used as a primary coordinate in our regime partitioning. Right : Distribution of the absolute pseudorapidity separation |∆η| . High values of |∆η| characterize forward-scattering events, which often suffer from higher detector noise. 2 Analysis To describe the physical state of a system,… view at source ↗
Figure 3
Figure 3. Figure 3: Unsupervised clustering of the (HT , |∆η|) space into distinct context regimes. domain-agnostic, as this clustering converts the continuous phase space into discrete regions. This fa￾cilitates the evaluation of the stability of the reconstruction pipeline across varied kinematic topologies. The K-Means algorithm inspects the global energy scale HT and the absolute value of the separation angle |∆η| with eq… view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of relative error distributions for different context regimes. This shift in the Z-peak position is due to the influence of something outside the local parameters of the individual leptons, and we call this broader scenario as the “global context”. We have two plots in the [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Heatmap of mean absolute bias. Note the emergence of “hot spots” at high HT and high |∆η|, indicating systematic framework degradation. heatmap tells us where the reconstruction pipeline becomes unreliable. A clear transition is seen here – as we move from the low energy to high energy region, the mean absolute bias increases, which can be seen in the [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
read the original abstract

The fast reconstruction and detector-simulation pipelines are widely used in different scientific domains, such as, in High Energy Physics (HEP) and medical imaging, where the full experimental or the device-level simulation is computationally challenging. Instead of the use of the global data context, these pipelines use simplified response models which assume reconstruction uncertainty relies on local input parameters. To probe the robustness of the local assumption, a domain-agnostic context-aware stress testing pipeline is introduced, and a reconstruction response depending on the global parameters is also allowed. For an instance in HEP at High-Luminosity LHC (HL-LHC) simulation, this work shows that the decay channel $Z \rightarrow \ell\ell$, as a benchmark, violates the local assumption resulting in a significant reconstruction bias and resolution degradation. Using the unsupervised regime-mapping framework, this work also restores this peak stability and recovers the truth-level resolution, where a robust diagnostic tool for next-generation fast simulation pipelines is accommodated.

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 manuscript introduces a domain-agnostic, context-aware stress-testing pipeline to probe the local-parameter assumption in fast reconstruction and detector-simulation pipelines. Using the Z → ℓℓ decay channel as a benchmark in an HL-LHC context, it reports that the local assumption is violated, producing measurable reconstruction bias and resolution degradation; an unsupervised regime-mapping framework is then applied to restore peak stability and recover truth-level resolution.

Significance. If the pipeline is shown not to introduce its own selection or reconstruction artifacts, the approach could supply a practical diagnostic for next-generation fast simulations in HEP and related fields where full simulation is prohibitive. The work directly addresses a common modeling assumption whose violation would affect downstream analyses, but the absence of quantitative validation details makes the practical impact difficult to gauge at present.

major comments (1)
  1. [Abstract] Abstract: The central claim that the introduced pipeline 'accurately isolates global-parameter dependence without introducing its own reconstruction artifacts or selection biases' is load-bearing for both the reported violation and the subsequent recovery. No closure test on a purely local toy model, comparison of selected vs. unselected distributions, or ablation of the regime-mapping step is described, leaving open the possibility that observed non-local effects are induced by the diagnostic itself rather than by the reconstruction model under test.
minor comments (2)
  1. [Abstract] Abstract, sentence 3: 'For an instance in HEP at High-Luminosity LHC (HL-LHC) simulation' is grammatically awkward and should be rephrased for clarity (e.g., 'As an example in the context of HL-LHC simulation in HEP').
  2. [Abstract] Abstract, final clause: 'where a robust diagnostic tool ... is accommodated' is unclear; the intended meaning appears to be that the framework provides such a tool, but the phrasing should be revised.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. The concern regarding validation of the diagnostic pipeline is well taken, and we address it directly below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that the introduced pipeline 'accurately isolates global-parameter dependence without introducing its own reconstruction artifacts or selection biases' is load-bearing for both the reported violation and the subsequent recovery. No closure test on a purely local toy model, comparison of selected vs. unselected distributions, or ablation of the regime-mapping step is described, leaving open the possibility that observed non-local effects are induced by the diagnostic itself rather than by the reconstruction model under test.

    Authors: We agree that the manuscript would be strengthened by explicit validation tests supporting the claim that the pipeline does not induce artifacts. The current text relies on the design principles of the stress-testing framework but does not present a dedicated closure test on a purely local model, selected vs. unselected distribution comparisons, or an ablation of the regime-mapping step. In the revised manuscript we will add a new subsection with (i) a closure test on a synthetic dataset constructed to obey the local assumption exactly, (ii) direct comparisons of kinematic distributions before and after any selection cuts to confirm the absence of induced biases, and (iii) an ablation study quantifying the contribution of the regime-mapping step. These additions will be incorporated in the next version. revision: yes

Circularity Check

0 steps flagged

No circularity in derivation chain

full rationale

The paper introduces a domain-agnostic stress-testing pipeline and an unsupervised regime-mapping framework to probe local assumptions in fast reconstruction, using Z→ℓℓ as benchmark. No equations, parameter fits, self-citations, or uniqueness theorems appear in the provided text. The central claims rest on empirical application of the new pipeline rather than any derivation that reduces to its own inputs by construction; the method is presented as an external diagnostic tool whose validity is not justified via self-reference.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that fast pipelines rely on local parameters and that a context-aware test can isolate global effects without new artifacts.

axioms (1)
  • domain assumption Reconstruction uncertainty in fast pipelines depends only on local input parameters
    This is the assumption the stress test is designed to probe.

pith-pipeline@v0.9.1-grok · 5681 in / 1055 out tokens · 33481 ms · 2026-06-26T23:31:24.292785+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

10 extracted references · 8 canonical work pages · 3 internal anchors

  1. [1]

    Pattern Recognition and Event Reconstruction in Particle Physics Experiments

    R. Mankel. “Pattern recognition and event reconstruction in particle physics experiments.” Rept. Prog. Phys.67, 553 (2004)doi:10.1088/0034-4885/67/4/R03[arXiv:physics/0402039 [physics]]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0034-4885/67/4/r03 2004

  2. [2]

    Full event interpretation with machine-learning-based particle-flow reconstruction in the CMS detector

    A. Hayrapetyanet al.[CMS]. “Full event interpretation with machine-learning-based particle-flow reconstruction in the CMS detector.” [arXiv:2601.17554 [physics.ins-det]]

    arXiv

  3. [3]

    Review of top quark mass measurements in CMS

    A. Hayrapetyanet al.[CMS]. “Review of top quark mass measurements in CMS.” Phys. Rept. 1115, 116-218 (2025)doi:10.1016/j.physrep.2024.12.002[arXiv:2403.01313 [hep-ex]]

    work page doi:10.1016/j.physrep.2024.12.002 2025

  4. [4]

    Kalvov´ a

    J. Bardhan, T. Mandal, S. Mitra, C. Neeraj and M. Patra. “Unsupervised and lightly supervised learning in particle physics.” Eur. Phys. J. ST233, no.15-16, 2559-2596 (2024)doi:10.1140/epjs/ s11734-024-01235-x[arXiv:2403.13676 [hep-ph]]

    work page doi:10.1140/epjs/ 2024

  5. [5]

    JUNIPR: a Framework for Unsupervised Machine Learning in Particle Physics

    A. Andreassen, I. Feige, C. Frye and M. D. Schwartz. “JUNIPR: A Framework for Unsupervised Machine Learning in Particle Physics.” Eur. Phys. J. C79, no.2, 102 (2019)doi:10.1140/epjc/ s10052-019-6607-9[arXiv:1804.09720 [hep-ph]]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1140/epjc/ 2019

  6. [6]

    org/10.1007/s41781-018-0018-8

    J. Albrechtet al.[HEP Software Foundation]. “A Roadmap for HEP Software and Computing R&D for the 2020s.” Comput. Softw. Big Sci.3(2019) no.1, 7doi:10.1007/s41781-018-0018-8 [arXiv:1712.06982 [physics.comp-ph]]

    work page doi:10.1007/s41781-018-0018-8 2019

  7. [7]

    The automated computation of tree-level and next-to-leading order differential cross sections, and their matching to parton shower simulations

    J. Alwall, R. Frederix, S. Frixione, V. Hirschi, F. Maltoni, O. Mattelaer, H. S. Shao, T. Stelzer, P. Torrielli and M. Zaro. “The automated computation of tree-level and next-to-leading order dif- ferential cross sections, and their matching to parton shower simulations.” JHEP07(2014), 079 doi:10.1007/JHEP07(2014)079[arXiv:1405.0301 [hep-ph]]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1007/jhep07(2014)079 2014

  8. [8]

    Learning the latent structure of collider events

    B. M. Dillon, D. A. Faroughy, J. F. Kamenik and M. Szewc. “Learning the latent structure of collider events.” JHEP10(2020), 206doi:10.1007/JHEP10(2020)206[arXiv:2005.12319 [hep-ph]]

    work page doi:10.1007/jhep10(2020)206 2020

  9. [9]

    Learning from many collider events at once

    B. Nachman and J. Thaler. “Learning from many collider events at once.” Phys. Rev. D103(2021) no.11, 116013doi:10.1103/PhysRevD.103.116013[arXiv:2101.07263 [physics.data-an]]

    work page doi:10.1103/physrevd.103.116013 2021

  10. [10]

    A Perspective on Deep Imaging

    Ge Wang. “A Perspective on Deep Imaging.” (2016). [arXiv:1609.04375 [q-bio.QM]]. 6

    Pith/arXiv arXiv 2016