pith. machine review for the scientific record. sign in

arxiv: 2605.03063 · v2 · submitted 2026-05-04 · ✦ hep-ph · hep-ex· stat.ML

Recognition: unknown

From Information Geometry to Jet Substructure: A Triality of Cumulant Tensors, Energy Correlators, and Hypergraphs

Aritra Bal , Markus Klute , Benedikt Maier , Michael Spannowsky
Authors on Pith no claims yet

Pith reviewed 2026-05-08 17:40 UTC · model grok-4.3

classification ✦ hep-ph hep-exstat.ML
keywords energy correlatorsjet substructureFisher tensorscumulantshypergraphsinformation geometryKullback-Leiblerexponential families
0
0 comments X

The pith

In a finite basis of binned energy correlators the local tensor simultaneously represents a KL coefficient, a connected cumulant, and a hyperedge weight.

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

The paper shows that higher-order Fisher tensors supply the structure missing from pairwise graphs by having three equivalent meanings in the right coordinates. This triality connects the local Kullback-Leibler expansion to cumulants of the observables and to weights on hyperedges. Readers should care because it identifies multi-observable radiation patterns that are genuinely connected rather than just collections of pairs. It also gives a way to build hypergraphs directly from correlator data and to compress bases while keeping higher-order information. The framework is illustrated in jet substructure classification and basis design tasks.

Core claim

In a finite basis of binned EECs, ECFs, or EFPs, and in the natural exponential-family coordinates generated by that basis, the same local tensor has three equivalent interpretations: a coefficient in the local Kullback-Leibler expansion, a connected cumulant of the chosen correlator observables, and a signed weight on a hyperedge linking those observables. This gives an exact Fisher-correlator-hypergraph triality in the local exponential-family embedding and provides a direct construction of physics-informed hypergraphs from correlator data.

What carries the argument

The higher-order local Fisher tensor in natural exponential-family coordinates, interpreted equivalently as KL coefficient, cumulant and hyperedge weight.

If this is right

  • The triality allows direct construction of physics-informed hypergraphs from correlator data.
  • Extending to the cubic tensor reduces KL truncation error and isolates dominant triplet structures.
  • It improves classification in compressed bases for two-versus-three prong jets.
  • It retains more third-order response in basis design with fewer observables.
  • The hyperedges serve as inductive bias for message passing in learning on observables.

Where Pith is reading between the lines

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

  • This approach could be tested in other collider observables to see if the triality holds beyond energy correlators.
  • The hypergraph construction might help in developing new jet clustering algorithms that account for higher-order correlations.
  • One could check if using these tensors improves performance in full event reconstruction tasks at the LHC.
  • The framework suggests a general method for turning information geometry objects into graph structures in other domains like network science.

Load-bearing premise

The finite basis of binned energy correlators must generate coordinates in which the tensor has identical meanings as a divergence coefficient, a cumulant, and a hyperedge weight.

What would settle it

Compute the three quantities separately from data in a given basis and observe whether the higher-order tensor components agree numerically; disagreement would show the triality does not hold.

read the original abstract

Pairwise Fisher graphs capture local covariance information, but they cannot distinguish an irreducible multi-observable radiation pattern from a collection of ordinary pairwise correlations. We show that this missing structure is naturally supplied by higher-order Fisher tensors. In a finite basis of binned EECs, ECFs, or EFPs, and in the natural exponential-family coordinates generated by that basis, the same local tensor has three equivalent interpretations: a coefficient in the local Kullback-Leibler expansion, a connected cumulant of the chosen correlator observables, and a signed weight on a hyperedge linking those observables. This gives an exact Fisher-correlator-hypergraph triality in the local exponential-family embedding. The triality provides a direct construction of physics-informed hypergraphs from correlator data. Extending the quadratic Fisher matrix to the first non-trivial higher tensor identifies genuinely connected multi-observable radiation patterns, supplies hyperedge weights for higher-order Laplacians and message passing, and gives a principled criterion for compressing observable bases beyond pairwise information. We develop these constructions and spell out why the exact cumulant interpretation is special to natural exponential-family coordinates. We illustrate the framework in four applications. In a minimal local-KL study, the cubic Fisher tensor reduces the KL truncation error and isolates the dominant triplet structure. In a two-versus-three prong jet substructure benchmark, the hypergraph selector improves compressed-basis classification. In a 33-observable basis-design problem, the Fisher hypergraph retains more third-order local response at twelve observables. A low-capacity learning benchmark then shows how the same Fisher hyperedges can be used as an interpretable inductive bias for message passing on correlator observables.

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

0 major / 3 minor

Summary. The manuscript claims an exact triality in the local exponential-family embedding of a finite basis of binned energy correlators (EECs, ECFs or EFPs): the higher-order Fisher tensor is simultaneously a coefficient in the local KL-divergence expansion, a connected cumulant of the chosen observables, and a signed hyperedge weight. This triality is said to follow from the structure of natural coordinates and is used to construct physics-informed hypergraphs. Four applications are presented: a minimal local-KL truncation study, a two-versus-three-prong jet classification benchmark, a 33-to-12 observable basis-compression task that retains third-order response, and a low-capacity message-passing learning benchmark that employs the hyperedges as inductive bias.

Significance. If the triality is rigorously established, the work supplies a principled, coordinate-specific route from correlator data to weighted hypergraphs that capture irreducible multi-observable radiation patterns. The explicit link between natural exponential-family coordinates and the cumulant interpretation is a useful clarification, and the four applications demonstrate concrete uses in jet substructure. The framework could improve interpretability of higher-order graph methods in QCD analyses and provide a systematic criterion for observable compression beyond pairwise Fisher information.

minor comments (3)
  1. The abstract states that the paper 'spells out why the exact cumulant interpretation is special to natural exponential-family coordinates,' yet the corresponding derivation or proof sketch should be highlighted with an explicit equation or proposition number for easy reference.
  2. In the basis-design application, the claim that the Fisher hypergraph 'retains more third-order local response at twelve observables' would be strengthened by reporting the precise numerical improvement (e.g., the ratio of retained third-order Fisher-tensor norms) rather than a qualitative statement.
  3. Notation for the binned correlator basis, the natural coordinates, and the resulting hyperedge weights should be introduced uniformly in an early section to avoid repeated re-definition when moving between the geometric, statistical, and graph-theoretic interpretations.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive and accurate summary of our manuscript, as well as for recommending minor revision. The referee correctly identifies the core contribution: the exact triality linking higher-order Fisher tensors in natural exponential-family coordinates of binned energy correlators to local KL coefficients, connected cumulants, and signed hyperedge weights, together with the four concrete applications in jet substructure. We appreciate the recognition that this supplies a principled route from correlator data to physics-informed hypergraphs.

Circularity Check

0 steps flagged

No significant circularity; triality follows from exponential family structure

full rationale

The paper presents the Fisher-correlator-hypergraph triality as a direct mathematical consequence of using natural coordinates in the exponential family whose sufficient statistics are the finite basis of binned energy correlators. In these coordinates the higher derivatives of the log-partition function are the connected cumulants by standard definition, which then serve as KL coefficients and hyperedge weights. The abstract states that the constructions are developed and the special role of the cumulant reading is spelled out, indicating an explanatory derivation rather than a reduction to fitted inputs or self-citations. No load-bearing self-citation chains, ansatzes smuggled via prior work, or predictions that collapse to the fit are described. The applications are downstream illustrations. The derivation chain is therefore self-contained against external benchmarks of information geometry.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the mathematical identification that holds only inside natural exponential-family coordinates generated by a finite basis of binned correlators; no free parameters or new entities are introduced in the abstract.

axioms (1)
  • domain assumption A finite basis of binned EECs, ECFs, or EFPs generates natural exponential-family coordinates in which the local Fisher tensor equals both the KL coefficient and the connected cumulant.
    Invoked as the setting where the three interpretations coincide exactly.

pith-pipeline@v0.9.0 · 5623 in / 1455 out tokens · 28012 ms · 2026-05-08T17:40:05.335123+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

44 extracted references · 35 canonical work pages · 1 internal anchor

  1. [1]

    C. L. Basham, L. S. Brown, S. D. Ellis, and S. T. Love. Energy Correlations in electron-Positron Annihilation in Quantum Chromodynamics: Asymptotically Free Perturbation Theory.Phys. Rev. D, 19:2018, 1979. doi: 10.1103/PhysRevD.19.2018

    work page doi:10.1103/physrevd.19.2018 2018

  2. [2]

    Energy Correlation Functions for Jet Substructure

    Andrew J. Larkoski, Gavin P. Salam, and Jesse Thaler. Energy Correlation Functions for Jet Substructure.JHEP, 06:108, 2013. doi: 10.1007/JHEP06(2013)108

    work page doi:10.1007/jhep06(2013)108 2013

  3. [3]

    Journal of High Energy Physics , volume =

    Patrick T. Komiske, Eric M. Metodiev, and Jesse Thaler. Energy flow polynomials: A complete linear basis for jet substructure.JHEP, 04:013, 2018. doi: 10.1007/JHEP04(2018)013. – 28 –

    work page doi:10.1007/jhep04(2018)013 2018

  4. [4]

    Conformal collider physics: Energy and charge correlations

    Diego M. Hofman and Juan Maldacena. Conformal collider physics: Energy and charge correlations.JHEP, 05:012, 2008. doi: 10.1088/1126-6708/2008/05/012

    work page doi:10.1088/1126-6708/2008/05/012 2008

  5. [5]

    Simplicity from Recoil: The Three-Loop Soft Function and Factorization for the Energy-Energy Correlation.JHEP, 08:160, 2018

    Ian Moult and Hua Xing Zhu. Simplicity from Recoil: The Three-Loop Soft Function and Factorization for the Energy-Energy Correlation.JHEP, 08:160, 2018. doi: 10.1007/JHEP08(2018)160

    work page doi:10.1007/jhep08(2018)160 2018

  6. [6]

    North-Holland, 1989

    Claude Berge.Hypergraphs: Combinatorics of Finite Sets, volume 45 ofNorth-Holland Mathematical Library. North-Holland, 1989

    1989

  7. [7]

    Mathematical Engineering

    Alain Bretto.Hypergraph Theory: An Introduction. Mathematical Engineering. Springer,

  8. [8]

    doi: 10.1007/978-3-319-00080-0

    work page doi:10.1007/978-3-319-00080-0

  9. [9]

    Learning with hypergraphs: Clustering, classification, and embedding

    Dengyong Zhou, Jiayuan Huang, and Bernhard Sch¨ olkopf. Learning with hypergraphs: Clustering, classification, and embedding. InAdvances in Neural Information Processing Systems 19, 2007

    2007

  10. [10]

    Hypergraph neural networks.CoRR, abs/1809.09401, 2018

    Yifan Feng, Haoxuan You, Zizhao Zhang, Rongrong Ji, and Yue Gao. Hypergraph neural networks.CoRR, abs/1809.09401, 2018. URLhttp://arxiv.org/abs/1809.09401

    work page arXiv 2018

  11. [11]

    Reconstructing short-lived particles using hypergraph representation learning.Phys

    Callum Birch-Sykes, Brian Le, Yvonne Peters, Ethan Simpson, and Zihan Zhang. Reconstructing short-lived particles using hypergraph representation learning.Phys. Rev. D, 111(3):032004, 2025. doi: 10.1103/PhysRevD.111.032004

    work page doi:10.1103/physrevd.111.032004 2025

  12. [12]

    Radhakrishna Rao

    C. Radhakrishna Rao. Information and the accuracy attainable in the estimation of statistical parameters.Bull. Calcutta Math. Soc., 37:81–91, 1945

    1945

  13. [13]

    A differential geometric approach to statistical inference on the basis of contrast functionals.Hiroshima Mathematical Journal, 15:341–391, 1985

    Shinto Eguchi. A differential geometric approach to statistical inference on the basis of contrast functionals.Hiroshima Mathematical Journal, 15:341–391, 1985. doi: 10.32917/hmj/1206130775

    work page doi:10.32917/hmj/1206130775 1985

  14. [14]

    Geometry of minimum contrast.Hiroshima Mathematical Journal, 22(3): 631–647, 1992

    Shinto Eguchi. Geometry of minimum contrast.Hiroshima Mathematical Journal, 22(3): 631–647, 1992. doi: 10.32917/hmj/1206128508

    work page doi:10.32917/hmj/1206128508 1992

  15. [15]

    Springer, 1985

    Shun-ichi Amari.Differential-Geometrical Methods in Statistics, volume 28 ofLecture Notes in Statistics. Springer, 1985. doi: 10.1007/978-1-4612-5056-2

    work page doi:10.1007/978-1-4612-5056-2 1985

  16. [16]

    American Mathematical Society and Oxford University Press, 2000

    Shun-ichi Amari and Hiroshi Nagaoka.Methods of Information Geometry, volume 191 of Translations of Mathematical Monographs. American Mathematical Society and Oxford University Press, 2000

    2000

  17. [17]

    Naturalgradientworksefficientlyinlearning

    Shun-ichi Amari. Natural gradient works efficiently in learning.Neural Computation, 10(2): 251–276, 1998. doi: 10.1162/089976698300017746

    work page doi:10.1162/089976698300017746 1998

  18. [18]

    Nihat Ay, J¨ urgen Jost, Hˆ ong Vˆ an Le, and Lorenz Schwachh¨ ofer.Information Geometry, volume 64 ofErgebnisse der Mathematik und ihrer Grenzgebiete. 3. Folge. Springer, 2017. doi: 10.1007/978-3-319-56478-4

    work page doi:10.1007/978-3-319-56478-4 2017

  19. [19]

    Chapman and Hall, 1987

    Peter McCullagh.Tensor Methods in Statistics. Chapman and Hall, 1987

    1987

  20. [20]

    Barndorff-Nielsen.Information and Exponential Families in Statistical Theory

    Ole E. Barndorff-Nielsen.Information and Exponential Families in Statistical Theory. Wiley Series in Probability and Mathematical Statistics. Wiley, 1978

    1978

  21. [21]

    N. N. Cencov.Statistical Decision Rules and Optimal Inference, volume 53 ofTranslations of Mathematical Monographs. American Mathematical Society, Providence, RI, 1982. English translation of the 1972 Russian original

    1982

  22. [22]

    Dixon, Ian Moult, and Hua Xing Zhu

    Lance J. Dixon, Ian Moult, and Hua Xing Zhu. Collinear limit of the energy-energy correlator.Phys. Rev. D, 100(1):014009, 2019. doi: 10.1103/PhysRevD.100.014009. – 29 –

    work page doi:10.1103/physrevd.100.014009 2019

  23. [23]

    Three point energy correlators in the collinear limit: symmetries, dualities and analytic results.JHEP, 08(08):028, 2020

    Hao Chen, Ming-Xing Luo, Ian Moult, Tong-Zhi Yang, Xiaoyuan Zhang, and Hua Xing Zhu. Three point energy correlators in the collinear limit: symmetries, dualities and analytic results.JHEP, 08(08):028, 2020. doi: 10.1007/JHEP08(2020)028

    work page doi:10.1007/jhep08(2020)028 2020

  24. [24]

    Thaler and K

    Jesse Thaler and Ken Van Tilburg. Identifying Boosted Objects with N-subjettiness.JHEP, 03:015, 2011. doi: 10.1007/JHEP03(2011)015

    work page doi:10.1007/jhep03(2011)015 2011

  25. [25]

    Jet substructure at the Large Hadron Collider: A review of recent advances in theory and machine learning

    Andrew J. Larkoski, Ian Moult, and Benjamin Nachman. Jet Substructure at the Large Hadron Collider: A Review of Recent Advances in Theory and Machine Learning.Phys. Rept., 841:1–63, 2020. doi: 10.1016/j.physrep.2019.11.001

    work page doi:10.1016/j.physrep.2019.11.001 2020

  26. [26]

    Metric space of collider events

    Patrick T. Komiske, Eric M. Metodiev, and Jesse Thaler. Metric Space of Collider Events. Phys. Rev. Lett., 123(4):041801, 2019. doi: 10.1103/PhysRevLett.123.041801

    work page doi:10.1103/physrevlett.123.041801 2019

  27. [27]

    Ngairangbam, and Michael Spannowsky

    Partha Konar, Vishal S. Ngairangbam, and Michael Spannowsky. Hypergraphs in LHC phenomenology — the next frontier of IRC-safe feature extraction.JHEP, 01:113, 2024. doi: 10.1007/JHEP01(2024)113

    work page doi:10.1007/jhep01(2024)113 2024

  28. [28]

    Alwall, R

    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 differential cross sections, and their matching to parton shower simulations.JHEP, 07:079, 2014. doi: 10.1007/JHEP07(2014)079

    work page doi:10.1007/jhep07(2014)079 2014

  29. [29]

    Parton distributions from high-precision collider data

    Richard D. Ball et al. Parton distributions from high-precision collider data.Eur. Phys. J. C, 77(10):663, 2017. doi: 10.1140/epjc/s10052-017-5199-5

    work page Pith review doi:10.1140/epjc/s10052-017-5199-5 2017

  30. [30]

    Hartland, and Juan Rojo

    Valerio Bertone, Stefano Carrazza, Nathan P. Hartland, and Juan Rojo. Illuminating the photon content of the proton within a global PDF analysis.SciPost Phys., 5(1):008, 2018. doi: 10.21468/SciPostPhys.5.1.008

    work page doi:10.21468/scipostphys.5.1.008 2018

  31. [31]

    Sj¨ ostrand, S

    Torbj¨ orn Sj¨ ostrand, Stefan Ask, Jesper R. Christiansen, Richard Corke, Nishita Desai, Philip Ilten, Stephen Mrenna, Stefan Prestel, Christine O. Rasmussen, and Peter Z. Skands. An introduction toPythia8.2.Comput. Phys. Commun., 191:159–177, 2015. doi: 10.1016/j.cpc.2015.01.024

    work page doi:10.1016/j.cpc.2015.01.024 2015

  32. [32]

    Extraction and validation of a new set of CMS PYTHIA8 tunes from underlying-event measurements

    Albert M Sirunyan et al. Extraction and validation of a new set of CMSPythia8tunes from underlying-event measurements.Eur. Phys. J. C, 80(1):4, 2020. doi: 10.1140/epjc/s10052-019-7499-4

    work page doi:10.1140/epjc/s10052-019-7499-4 2020

  33. [33]

    DELPHES 3, A modular framework for fast simulation of a generic collider experiment

    J. de Favereau, C. Delaere, P. Demin, A. Giammanco, V. Lemaˆ ıtre, A. Mertens, and M. Selvaggi.Delphes3: a modular framework for fast simulation of a generic collider experiment.JHEP, 02:057, 2014. doi: 10.1007/JHEP02(2014)057

    work page internal anchor Pith review doi:10.1007/jhep02(2014)057 2014

  34. [34]

    Jet substructure as a new Higgs search channel at the LHC

    Jonathan M. Butterworth, Adam R. Davison, Mathieu Rubin, and Gavin P. Salam. Jet substructure as a new Higgs search channel at the LHC.Phys. Rev. Lett., 100:242001, 2008. doi: 10.1103/PhysRevLett.100.242001

    work page doi:10.1103/physrevlett.100.242001 2008

  35. [35]

    Soper and Michael Spannowsky

    Davison E. Soper and Michael Spannowsky. Finding physics signals with shower deconstruction.Phys. Rev. D, 84:074002, 2011. doi: 10.1103/PhysRevD.84.074002

    work page doi:10.1103/physrevd.84.074002 2011

  36. [36]

    Butteret al.,The Machine Learning landscape of top taggers, SciPost Phys.7, 014 (2019), doi:10.21468/SciPostPhys.7.1.014,1902.09914

    Gregor Kasieczka et al. The Machine Learning landscape of top taggers.SciPost Phys., 7: 014, 2019. doi: 10.21468/SciPostPhys.7.1.014

    work page doi:10.21468/scipostphys.7.1.014 2019

  37. [37]

    Kaplan, Keith Rehermann, Matthew D

    David E. Kaplan, Keith Rehermann, Matthew D. Schwartz, and Brock Tweedie. Top Tagging: A Method for Identifying Boosted Hadronically Decaying Top Quarks.Phys. Rev. Lett., 101:142001, 2008. doi: 10.1103/PhysRevLett.101.142001. – 30 –

    work page doi:10.1103/physrevlett.101.142001 2008

  38. [38]

    Kogleret al.,Jet Substructure at the Large Hadron Collider: Experimental Review, Rev

    Roman Kogler et al. Jet Substructure at the Large Hadron Collider: Experimental Review. Rev. Mod. Phys., 91(4):045003, 2019. doi: 10.1103/RevModPhys.91.045003

    work page doi:10.1103/revmodphys.91.045003 2019

  39. [39]

    Cacciari, G.P

    Matteo Cacciari, Gavin P. Salam, and Gregory Soyez. The anti-k t jet clustering algorithm. JHEP, 04:063, 2008. doi: 10.1088/1126-6708/2008/04/063

    work page doi:10.1088/1126-6708/2008/04/063 2008

  40. [40]

    Larkoski, Jesse Thaler, and Wouter J

    Andrew J. Larkoski, Jesse Thaler, and Wouter J. Waalewijn. Gaining (Mutual) Information about Quark/Gluon Discrimination.JHEP, 11:129, 2014. doi: 10.1007/JHEP11(2014)129

    work page doi:10.1007/jhep11(2014)129 2014

  41. [41]

    Qu and L

    Huilin Qu and Loukas Gouskos. ParticleNet: Jet Tagging via Particle Clouds.Phys. Rev. D, 101(5):056019, 2020. doi: 10.1103/PhysRevD.101.056019

    work page doi:10.1103/physrevd.101.056019 2020

  42. [42]

    Ngairangbam, and Michael Spannowsky

    Partha Konar, Vishal S. Ngairangbam, and Michael Spannowsky. Energy-weighted message passing: an infra-red and collinear safe graph neural network algorithm.JHEP, 02:060, 2022. doi: 10.1007/JHEP02(2022)060

    work page doi:10.1007/jhep02(2022)060 2022

  43. [43]

    Anomaly Detection for Physics Analysis and Less than Supervised Learning

    Benjamin Nachman. Anomaly Detection for Physics Analysis and Less than Supervised Learning. 10 2020

    2020

  44. [44]

    Adversarially-trained autoencoders for robust unsupervised new physics searches.JHEP, 10:047, 2019

    Andrew Blance, Michael Spannowsky, and Philip Waite. Adversarially-trained autoencoders for robust unsupervised new physics searches.JHEP, 10:047, 2019. doi: 10.1007/JHEP10(2019)047. – 31 –

    work page doi:10.1007/jhep10(2019)047 2019