pith. sign in

arxiv: 2607.00884 · v1 · pith:VVVQHQ3Hnew · submitted 2026-07-01 · ✦ hep-ex · physics.data-an

Modeling Falling Backgrounds with Exponential Mixtures

Austin Townsend , Marc Osherson , Mike Hildreth , Stefano Castruccio This is my paper

Pith reviewed 2026-07-02 02:56 UTC · model grok-4.3

classification ✦ hep-ex physics.data-an
keywords exponential mixturebackground modelingLHC searchesfalling distributionsextreme value theorysemi-parametric modelnew physics
0
0 comments X

The pith

Finite exponential mixtures model falling LHC backgrounds with performance comparable to existing methods on real and simulated data.

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

The paper establishes that finite mixtures of exponential distributions form a flexible semi-parametric family for approximating the smoothly falling backgrounds common in LHC new-physics searches. The motivation comes from extreme-value theory results on tail behavior, offering a way to avoid the extensive case-by-case parametric tuning that grows burdensome with larger datasets. Tests on two published datasets (28.6 million and 5,000 events) show the mixtures perform at least as well as standard approaches. In controlled simulations of 5,000 events the model produces bias small relative to statistical uncertainty and preserves nominal coverage across the bulk of the distribution. If the claim holds, background modeling in future searches could proceed with a single, reusable functional form rather than repeated ad-hoc development.

Core claim

Finite exponential mixtures constitute an effective semi-parametric class for modeling falling background distributions in LHC searches. On two published datasets the performance is comparable to existing methods for both small and large samples; in simulation studies the finite mixture exhibits small bias relative to the true statistical uncertainty while maintaining consistent nominal coverage in the bulk.

What carries the argument

The finite exponential mixture: a weighted sum of exponential densities whose form is justified by extreme-value theory for approximating falling tails.

If this is right

  • The same mixture form applies without major changes to both small and large published LHC datasets.
  • The model reduces the need for analysis-specific parametric families that require repeated development as data volumes increase.
  • Simulation results indicate that the approach keeps bias small relative to uncertainty and preserves nominal coverage in the bulk.
  • The method can be used directly in searches for localized excesses on falling backgrounds.

Where Pith is reading between the lines

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

  • The same construction might apply to background modeling in other collider experiments that encounter similar falling spectra.
  • If the approximation holds across many analyses, the mixture could serve as a default starting point that shortens the time spent on background validation.
  • Extensions could test whether adding a small number of mixture components suffices for the highest-mass tails encountered in Run 3 and HL-LHC data.

Load-bearing premise

Falling background distributions in LHC searches belong to a class that finite exponential mixtures approximate well without requiring analysis-specific validation or post-hoc adjustments.

What would settle it

A new dataset or simulation in which the exponential mixture produces bias exceeding the reported statistical uncertainty or coverage falling outside the nominal interval in the bulk region.

read the original abstract

Searches for new physics at the LHC often look for localized excesses on smoothly falling background distributions. Several classes of background models have been considered, including polynomials and other parametric families; however, these approaches can require extensive analysis-specific development as datasets grow. In this work, we motivate the finite exponential mixture as a flexible semi-parametric class of functions for approximating falling distributions, drawing on results from extreme value theory. Using two published datasets ($n=28,619,185$ and $n=5,036$), we show that the exponential mixture performance is comparable to existing methods for both small and large datasets. Finally, in simulation studies ($n = 5,036$), we find that the finite exponential mixture exhibits small bias relative to the true statistical uncertainty while maintaining consistent nominal coverage in the bulk.

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

2 major / 2 minor

Summary. The manuscript proposes finite exponential mixtures, motivated by extreme value theory, as a flexible semi-parametric model for smoothly falling backgrounds in LHC new-physics searches. It reports that the approach yields performance comparable to existing methods on two published datasets (n=28,619,185 and n=5,036) and, in simulation studies at n=5,036, exhibits small bias relative to statistical uncertainty while maintaining consistent nominal coverage in the bulk.

Significance. If the empirical results generalize, the method could reduce the need for extensive analysis-specific background modeling as dataset sizes increase, providing a more standardized semi-parametric alternative to polynomials. The use of two real published datasets for direct comparison is a strength.

major comments (2)
  1. [Abstract] Abstract: the central assertion that finite exponential mixtures approximate arbitrary falling LHC backgrounds 'without requiring analysis-specific validation' is not supported by a general characterization of the function class, a proof that typical LHC spectra lie in it, or a procedure for detecting when the approximation fails; the evidence consists only of performance on two specific datasets plus simulations at a single size.
  2. [Simulation studies] Simulation studies section: bias and coverage are demonstrated only for n=5,036; it is unclear whether these properties extend to the n=28M regime or to other falling spectra encountered in LHC analyses, which is load-bearing for the claim of consistent nominal coverage.
minor comments (2)
  1. The connection to extreme value theory is invoked but no specific EVT result or reference is supplied to justify the exponential-mixture form over other semi-parametric families.
  2. The procedure for selecting the number of mixture components and the fitting algorithm (including any regularization or convergence criteria) should be stated explicitly for reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Thank you for the detailed review and constructive feedback. We address each major comment below and indicate the revisions we will make to improve the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central assertion that finite exponential mixtures approximate arbitrary falling LHC backgrounds 'without requiring analysis-specific validation' is not supported by a general characterization of the function class, a proof that typical LHC spectra lie in it, or a procedure for detecting when the approximation fails; the evidence consists only of performance on two specific datasets plus simulations at a single size.

    Authors: We agree that the manuscript provides no general characterization of the function class, no proof that typical LHC spectra lie within it, and no procedure for detecting approximation failure. The motivation draws on extreme value theory results for exponential mixtures, but the support remains empirical, based on the two datasets and simulations at n=5,036. We will revise the abstract to remove any implication of applicability to arbitrary backgrounds without validation and instead state the EVT motivation together with the specific empirical comparisons performed. revision: yes

  2. Referee: [Simulation studies] Simulation studies section: bias and coverage are demonstrated only for n=5,036; it is unclear whether these properties extend to the n=28M regime or to other falling spectra encountered in LHC analyses, which is load-bearing for the claim of consistent nominal coverage.

    Authors: The simulation studies are performed at n=5,036 to match the smaller real dataset and permit direct evaluation of bias and coverage against a known truth. For the n≈28M dataset the true background is unknown, precluding the same assessment; performance is instead compared to existing methods. We acknowledge that the reported nominal coverage is demonstrated only in the n=5,036 regime and that extension to larger samples or other spectra is not shown. We will revise the simulation section to state this scope explicitly, note the limitation for the large-n regime, and add a brief discussion of why the EVT motivation suggests the properties may generalize, while making clear that further verification would be required. revision: partial

Circularity Check

0 steps flagged

No significant circularity; claims rest on external empirical comparisons

full rationale

The paper motivates the finite exponential mixture class from external extreme value theory results and then reports performance on two published external datasets plus separate simulation studies. No equations, fitted parameters, or self-citations are shown that would reduce any reported performance metric or coverage claim to an input by construction. The central assertions are framed as direct empirical comparisons against existing methods on independent data, making the derivation self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review performed on abstract only; no explicit free parameters, axioms, or invented entities are stated beyond the general appeal to extreme value theory.

axioms (1)
  • domain assumption Results from extreme value theory justify finite exponential mixtures as approximations to falling distributions.
    Abstract states the motivation is drawn from extreme value theory.

pith-pipeline@v0.9.1-grok · 5662 in / 1076 out tokens · 54601 ms · 2026-07-02T02:56:32.020772+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

19 extracted references · 6 canonical work pages · 3 internal anchors

  1. [1]

    Searches for Dijet Resonances at Hadron Colliders

    R.M. Harris and K. Kousouris,Searches for Dijet Resonances at Hadron Colliders, International Journal of Modern Physics A26(2011) 5005 [1110.5302]. [4]CMScollaboration,Measurements of Higgs boson properties in the diphoton decay channel in proton-proton collisions at √s= 13TeV,Journal of High Energy Physics11(2018) 185. [5]CMScollaboration,Search for New ...

    work page internal anchor Pith review Pith/arXiv arXiv 2011

  2. [2]

    Dauncey, M

    P.D. Dauncey, M. Kenzie, N. Wardle and G.J. Davies,Handling Uncertainties in Background Shapes: The Discrete Profiling Method,Journal of Instrumentation10(2015) P04015

    2015

  3. [3]

    Modeling Smooth Backgrounds and Generic Localized Signals with Gaussian Processes

    M. Frate, K. Cranmer, S. Kalia, A. Vandenberg-Rodes and D. Whiteson,Modeling Smooth Backgrounds and Generic Localized Signals with Gaussian Processes,1709.05681

    work page internal anchor Pith review Pith/arXiv arXiv

  4. [4]

    H.F. Tsoi, D. Rankin, C. Caillol, M. Cranmer, S. Dasu, J. Duarte et al.,SymbolFit: Automatic Parametric Modeling with Symbolic Regression,Computing and Software for Big Science9(2025) 12 [2411.09851]

    work page arXiv 2025

  5. [5]

    Pickands, III,Statistical Inference Using Extreme Order Statistics,The Annals of Statistics3(1975) 119

    J. Pickands, III,Statistical Inference Using Extreme Order Statistics,The Annals of Statistics3(1975) 119

    1975

  6. [6]

    Balkema and L

    A.A. Balkema and L. de Haan,Residual Life Time at Great Age,The Annals of Probability 2(1974) 792

    1974

  7. [7]

    Kiefer and J

    J. Kiefer and J. Wolfowitz,Consistency of the Maximum Likelihood Estimator in the Presence of Infinitely Many Incidental Parameters,The Annals of Mathematical Statistics 27(1956) 887

    1956

  8. [8]

    Lindsay,The Geometry of Mixture Likelihoods: A General Theory,The Annals of Statistics11(1983) 86

    B.G. Lindsay,The Geometry of Mixture Likelihoods: A General Theory,The Annals of Statistics11(1983) 86

    1983

  9. [9]

    Bernstein,Sur les fonctions absolument monotones,Acta Mathematica52(1929) 1

    S. Bernstein,Sur les fonctions absolument monotones,Acta Mathematica52(1929) 1

    1929

  10. [10]

    McGlinn,Uniform Approximation of Completely Monotone Functions by Exponential Sums,Journal of Mathematical Analysis and Applications65(1978) 211

    R.J. McGlinn,Uniform Approximation of Completely Monotone Functions by Exponential Sums,Journal of Mathematical Analysis and Applications65(1978) 211

    1978

  11. [11]

    Resnick,Extreme Values, Regular Variation and Point Processes, Springer New York (1986), 10.1007/978-0-387-75953-1

    S.I. Resnick,Extreme Values, Regular Variation and Point Processes, Springer New York (1986), 10.1007/978-0-387-75953-1

    work page doi:10.1007/978-0-387-75953-1 1986

  12. [12]

    Maguire, L

    E. Maguire, L. Heinrich and G. Watt,HEPData: A Repository for High Energy Physics Data,Journal of Physics: Conference Series898(2017) 102006

    2017

  13. [13]

    The RooFit toolkit for data modeling

    W. Verkerke and D.P. Kirkby,The RooFit Toolkit for Data Modeling,eConfC0303241 (2003) MOLT007 [physics/0306116]

    work page internal anchor Pith review Pith/arXiv arXiv 2003

  14. [14]

    Brun and F

    R. Brun and F. Rademakers,ROOT: An Object Oriented Data Analysis Framework,Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment389(1997) 81

    1997

  15. [15]

    Hatlo, F

    M. Hatlo, F. James, P. Mato, L. Moneta, M. Winkler and A. Zsenei,Developments of Mathematical Software Libraries for the LHC Experiments,IEEE Transactions on Nuclear Science52(2005) 2818

    2005

  16. [16]

    McLachlan and D

    G.J. McLachlan and D. Peel,Finite Mixture Models, Wiley Series in Probability and Statistics, Wiley (2004), 10.1002/0471721182. [23]ATLAScollaboration,Search for New Resonances in Mass Distributions of Jet Pairs Using 139 fb −1 ofppCollisions at √s= 13TeV with the ATLAS Detector,Journal of High Energy Physics03(2020) 145 [1910.08447]. – 17 – [24]ATLAScoll...

    work page doi:10.1002/0471721182 2004

  17. [17]

    Search for new physics in high-mass diphoton events from proton-proton collisions at √s= 13 TeV

    CMS Collaboration, “Search for new physics in high-mass diphoton events from proton-proton collisions at √s= 13 TeV.” HEPData (collection), 2024

    2024

  18. [18]

    Efron and R

    B. Efron and R. Tibshirani,Bootstrap Methods for Standard Errors, Confidence Intervals, and Other Measures of Statistical Accuracy,Statistical Science1(1986) 54

    1986

  19. [19]

    N. Ueda, R. Nakano, Z. Ghahramani and G.E. Hinton,Split and Merge EM Algorithm for Improving Gaussian Mixture Density Estimates,Journal of VLSI Signal Processing Systems for Signal, Image and Video Technology26(1998) 133. – 18 –

    1998