pith. sign in

arxiv: 2605.15910 · v1 · pith:IHWCUKNUnew · submitted 2026-05-15 · ✦ hep-ph

Geometric algebra as the input language of collider foundation models

E. Abasov , L. Dudko , F. Grigoryev , P. Volkov , A. Zaborenko This is my paper

Pith reviewed 2026-05-20 16:55 UTC · model grok-4.3

classification ✦ hep-ph
keywords geometric algebracollider eventsmultivectorfoundation modelsLorentz equivarianceCP-odd observablesresonance topologyequivariant networks
0
0 comments X

The pith

Collider events can be encoded as single multivectors whose grades directly recover the observables used in particle analyses.

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

The paper proposes treating each hard hadron-collider event as one multivector in geometric algebra rather than a list of four-momenta and separate labels. This multivector lives in Cl(1,3) tensor a flavor space and encodes both kinematics and particle types. Its grade decomposition then lines up with the quantities analysts already use: grade zero for masses and dot products, grade one for momenta, grade two for decay planes, grade three for volumes, and grade four for the CP-odd pseudoscalar. The authors argue this unified object supplies a natural input layer for foundation models that respect spacetime and discrete symmetries. They supply an explicit dictionary of 34 observables by grade and demonstrate the idea on resonance-topology separation with a multivector transformer.

Core claim

A hard hadron-collider event is treated here as a single geometric object—the kinematics and the discrete object-type labels of all reconstructed final-state particles encoded in one multivector evMV∈Cl(1,3)⊗Vflav—rather than as the customary list of four-momenta with separate label fields attached. The natural mathematical setting for this view is geometric algebra, whose grade decomposition is shown to organise essentially every observable in current use for collider analyses: inner products and invariant masses at grade zero, four-momenta at grade one, decay-plane bivectors at grade two, oriented three-volumes at grade three, and the CP-odd pseudoscalar at grade four. The high-level invar

What carries the argument

The event multivector evMV ∈ Cl(1,3) ⊗ Vflav whose grade projections recover invariants, momenta, planes, and the CP-odd sign while supplying inputs to Lorentz-equivariant networks.

If this is right

  • The high-level invariants, low-level recipe, and equivariant-network inputs are recovered as projections onto specific grades.
  • An explicit per-grade dictionary of 34 classical observables organises all standard collider quantities.
  • The Cayley-Menger lemma shows no new Lorentz-invariant scalars exist beyond the usual dot products and masses.
  • The genuine new channel is the CP-odd sign carried by the grade-four pseudoscalar.
  • A grade-resolved pre-training strategy is outlined for foundation models of collider physics.

Where Pith is reading between the lines

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

  • The algebraic structure could let networks enforce Lorentz and flavor symmetries at the input layer without explicit augmentation.
  • Grade-resolved pre-training might improve performance on tasks that rely on both local particle properties and global event topology.
  • The same multivector encoding could be tested on processes with known CP violation to measure the practical value of the pseudoscalar component.

Load-bearing premise

That the grade decomposition of the multivector organises essentially every observable in current use for collider analyses without loss of information or the need for additional ad-hoc mappings.

What would settle it

Construct the multivector from a sample event and verify whether the grade-zero projection exactly reproduces the invariant mass of every particle pair and the grade-two projection reproduces the decay plane for every three-body resonance.

Figures

Figures reproduced from arXiv: 2605.15910 by A. Zaborenko, E. Abasov, F. Grigoryev, L. Dudko, P. Volkov.

Figure 1
Figure 1. Figure 1: Grade ladder of Cl(1, 3). The five homogeneous components have dimen￾sions (1, 4, 6, 4, 1), the binomial coefficients summing to 24 = 16; each grade carries a definite physical reading on collider four-momenta. The Hodge dual X ⋆ ≡ X I−1 pairs grade k with grade 4 − k (dashed arcs on the right). (eq. (13)) and the missing-transverse-momentum pseudo-object (eq. (14)) are catalogued there. The grade decompos… view at source ↗
Figure 2
Figure 2. Figure 2: Per-particle assembly of the event multivector for an [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The event multivector E as a hierarchical decomposition into the kinematic Clifford factor Cl(1, 3) and the Lorentz-trivial object-type factor Vflav. The per-grade counts on the right are for an event with N final-state detector objects (the parton￾level tW b demonstration of Sec. 7 uses N = 6; reconstructed-level single-lepton￾plus-jets samples typically extend to N ≲ 8 once ISR jets are included). The ob… view at source ↗
Figure 4
Figure 4. Figure 4: Two representative tree-level diagrams contributing to the [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Receiver-operating curves for GA (red) and REF (blue) on the parton￾level discrimination of single-resonant tW against double-resonant t¯t topologies of pp → tW b with p j4 T > 10 GeV. Both networks are evaluated on the same Monte-Carlo samples (see text and Tab. 3). The dashed line is the random-classifier baseline. What the algebraic representation buys. Three properties of E make it a natural foundation… view at source ↗
Figure 6
Figure 6. Figure 6: Discriminator output on the three independent parton-level samples: [PITH_FULL_IMAGE:figures/full_fig_p023_6.png] view at source ↗
read the original abstract

A hard hadron-collider event is treated here as a single geometric object - the kinematics and the discrete object-type labels of all reconstructed final-state particles encoded in one multivector $\evMV\in\Cl(1,3)\otimes\Vflav$ - rather than as the customary list of four-momenta with separate label fields attached. The natural mathematical setting for this view is geometric algebra, whose grade decomposition is shown to organise essentially every observable in current use for collider analyses: inner products and invariant masses at grade zero, four-momenta at grade one, decay-plane bivectors at grade two, oriented three-volumes at grade three, and the CP-odd pseudoscalar at grade four. The high-level invariants, the low-level recipe, and the equivariant-network inputs are recovered as projections onto specific grades. An explicit per-grade dictionary of $34$ classical observables is provided, and the spacetime, discrete and approximate symmetries acting on $\evMV$ are listed. The Cayley--Menger lemma settles the question of new Lorentz-invariant scalars: none are unlocked beyond $\{p_i\!\cdot\!p_j,\,m_i^2\}$; the genuine non-trivial channel is the CP-odd sign of the pseudoscalar. The event-as-geometric-object representation is intended as a uniform input layer for foundation models of collider physics, and a grade-resolved pre-training strategy is outlined. The methodology is illustrated on the resonance-topology separation with a Lorentz-equivariant multivector transformer type whose per-particle grade-$0\!\oplus\!1$ tokens are complemented by event-level pairing tokens that surface the grade-two and grade-three candidate-pairing content of the multi-resonance topology at the input layer.

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.

Circularity Check

0 steps flagged

No significant circularity; definitional proposal with independent representational content

full rationale

The paper advances a representational framework that encodes collider events as multivectors in Cl(1,3) ⊗ Vflav and maps grade projections to existing observables via an explicit 34-item dictionary. This mapping is constructed by definition from the chosen encoding rather than derived from fitted parameters or self-referential predictions. No load-bearing steps reduce to self-citations, uniqueness theorems imported from prior author work, or ansatzes smuggled through citations; the Cayley-Menger reference and symmetry listings are external mathematical facts applied to the new object. The introduction of supplementary per-particle and pairing tokens is stated explicitly to address combinatorial structure, keeping the central claim self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The framework rests on the algebraic properties of geometric algebra and the assumption that collider observables are fully captured by grade projections; no numerical free parameters are introduced.

axioms (2)
  • standard math The geometric algebra Cl(1,3) admits a grade decomposition that organises inner products, four-momenta, bivectors, and pseudoscalars in a manner compatible with Lorentz symmetry.
    Invoked in the abstract when stating that grade decomposition organises every observable in current use.
  • standard math The Cayley-Menger lemma implies that no new Lorentz-invariant scalars exist beyond pairwise dot products and squared masses.
    Used to settle the question of new scalars in the abstract.
invented entities (1)
  • evMV multivector in Cl(1,3)⊗Vflav no independent evidence
    purpose: To serve as a single geometric object encoding all kinematics and discrete labels of a collider event
    Central representational innovation introduced in the abstract; no independent experimental evidence provided.

pith-pipeline@v0.9.0 · 5856 in / 1499 out tokens · 74622 ms · 2026-05-20T16:55:47.558507+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

  • IndisputableMonolith/Foundation/AlexanderDuality.lean alexander_duality_circle_linking echoes
    ?
    echoes

    ECHOES: this paper passage has the same mathematical shape or conceptual pattern as the Recognition theorem, but is not a direct formal dependency.

    A hard hadron-collider event is treated here as a single geometric object—the kinematics and the discrete object-type labels of all reconstructed final-state particles encoded in one multivector evMV∈Cl(1,3)⊗Vflav—rather than as the customary list of four-momenta with separate label fields attached. The natural mathematical setting for this view is geometric algebra, whose grade decomposition is shown to organise essentially every observable...

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

40 extracted references · 40 canonical work pages · 14 internal anchors

  1. [1]

    Doran and A

    C. Doran and A. Lasenby, Geometric Algebra for Physicists , Cambridge University Press, doi:10.1017/CBO9780511807497 (2003)

    work page doi:10.1017/cbo9780511807497 2003

  2. [2]

    Hestenes and G

    D. Hestenes and G. Sobczyk, Clifford Algebra to Geometric Calculus: A Unified Language for Mathematics and Physics , Reidel, Dordrecht, doi:10.1007/978-94-009-6292-7 (1984)

    work page doi:10.1007/978-94-009-6292-7 1984

  3. [3]

    Lounesto, Clifford Algebras and Spinors , vol

    P. Lounesto, Clifford Algebras and Spinors , vol. 286 of London Mathematical Society Lecture Note Series , Cambridge University Press, 2 edn., ISBN 9780521005517, doi:10.1017/CBO9780511526022 (2001)

    work page doi:10.1017/cbo9780511526022 2001

  4. [4]

    E. E. Boos, V. E. Bunichev, L. V. Dudko and A. A. Markina, Method of optimum observables and implementation of neural networks in physics investigations , Phys. Atom. Nucl. 71(2), 388 (2008), doi:10.1134/s1063778808020191

    work page doi:10.1134/s1063778808020191 2008

  5. [5]

    Dudko, G

    L. Dudko, G. Vorotnikov, P. Volkov, D. Ovchinnikov, M. Perfilov, A. Shporin and A. Chernoded, General recipe to form input space for deep learning analysis of HEP scattering processes , Int. J. Mod. Phys. A 35(21), 2050119 (2020), doi:10.1142/S0217751X20501195, 2002.09350

    work page doi:10.1142/s0217751x20501195 2020

  6. [6]

    Bogatskiy, B

    A. Bogatskiy, B. Anderson, J. T. Offermann, M. Roussi, D. W. Miller and R. Kondor, Lorentz Group Equivariant Neural Network for Particle Physics (2020), 2006.04780

    work page arXiv 2020

  7. [7]

    Bogatskiy, T

    A. Bogatskiy, T. Hoffman, D. W. Miller and J. T. Offermann, PELICAN: Permutation Equivariant and Lorentz Invariant or Covariant Aggregator Network for Particle Physics (2022), 2211.00454

    work page arXiv 2022

  8. [8]

    Estimation of line-of-sight velocities of individual galaxies using neural networks - I. Modelling redshift-space distortions at large scales

    J. Brehmer, P. de Haan, S. Behrends and T. Cohen, Geometric Algebra Transformer , In Advances in Neural Information Processing Systems , vol. 37 (2023), 2305.18415

    work page arXiv 2023

  9. [9]

    Lorentz-equivariant geometric algebra transformers for high-energy physics

    J. Spinner, V. Bres \'o , P. de Haan, T. Plehn, J. Thaler and J. Brehmer, Lorentz-Equivariant Geometric Algebra Transformers for High-Energy Physics , In 38th conference on Neural Information Processing Systems (2024), 2405.14806

    work page arXiv 2024

  10. [10]

    Brehmer, V

    J. Brehmer, V. Bres \'o , P. de Haan, T. Plehn, H. Qu, J. Spinner and J. Thaler, A Lorentz-equivariant transformer for all of the LHC , SciPost Phys. 19(4), 108 (2025), doi:10.21468/SciPostPhys.19.4.108, 2411.00446

    work page doi:10.21468/scipostphys.19.4.108 2025

  11. [11]

    E. E. Boos, V. E. Bunichev, P. V. Volkov, L. V. Dudko and M. A. Perfilov, Separation of Pair and Single Top Quark Production in tWb Associated Final State Using a Neural Network , Moscow Univ. Phys. Bull. 78(6), 707 (2023), doi:10.3103/S0027134923060024

    work page doi:10.3103/s0027134923060024 2023

  12. [12]

    E. E. Boos, V. E. Bunichev, L. V. Dudko and M. A. Perfilov, Application of the Subsidiary Fields Method to the Modeling of tW + t t Processes with the Anomalous Wtb Interactions , Phys. Atom. Nucl. 83(6), 989 (2020), doi:10.1134/S1063778820060095

    work page doi:10.1134/s1063778820060095 2020

  13. [13]

    Baskakov, E

    A. Baskakov, E. Boos, V. Bunichev, L. Dudko, M. Perfilov and P. Volkov, Recommendations for the search of the Anomalous Wtb interactions in the tW -associated Single Top Quark Production , EPJ Web Conf. 222, 04010 (2019), doi:10.1051/epjconf/201922204010

    work page doi:10.1051/epjconf/201922204010 2019

  14. [14]

    de Haan, T

    P. de Haan, T. Cohen and J. Brehmer, Euclidean, Projective, Conformal: Choosing a Geometric Algebra for Equivariant Transformers , In International Conference on Artificial Intelligence and Statistics (2024), 2311.04744

    work page arXiv 2024

  15. [15]

    Cayley, A theorem in the geometry of position , Cambridge Math

    A. Cayley, A theorem in the geometry of position , Cambridge Math. J. 2, 267 (1841)

    work page

  16. [16]

    Menger, Untersuchungen \"uber allgemeine Metrik , Math

    K. Menger, Untersuchungen \"uber allgemeine Metrik , Math. Ann. 100, 75 (1928), doi:10.1007/BF01448840

    work page doi:10.1007/bf01448840 1928

  17. [17]

    Realizability of the Lorentzian (n,1)-Simplex

    K. Tate and M. Visser, Realizability of the Lorentzian (n,1)-Simplex , JHEP 01, 028 (2012), doi:10.1007/JHEP01(2012)028, 1110.5694

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1007/jhep01(2012)028 2012

  18. [18]

    CP Violation in Top Physics

    D. Atwood, S. Bar-Shalom, G. Eilam and A. Soni, CP violation in top physics , Phys. Rept. 347, 1 (2001), doi:10.1016/S0370-1573(00)00112-5, hep-ph/0006032

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1016/s0370-1573(00)00112-5 2001

  19. [19]

    Mikuni and B

    V. Mikuni and B. Nachman, Solving key challenges in collider physics with foundation models , Phys. Rev. D 111(5), L051504 (2025), doi:10.1103/PhysRevD.111.L051504, 2404.16091

    work page doi:10.1103/physrevd.111.l051504 2025

  20. [20]

    Single Top Quark Production at the LHC: Understanding Spin

    G. Mahlon and S. J. Parke, Single top quark production at the LHC: Understanding spin , Phys. Lett. B 476, 323 (2000), doi:10.1016/S0370-2693(00)00149-0, hep-ph/9912458

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1016/s0370-2693(00)00149-0 2000

  21. [21]

    Top quark spin correlations and polarization at the LHC: standard model predictions and effects of anomalous top chromo moments

    W. Bernreuther and Z.-G. Si, Top quark spin correlations and polarization at the LHC: standard model predictions and effects of anomalous top chromo moments , Phys. Lett. B 725, 115 (2013), doi:10.1016/j.physletb.2013.06.051, [Erratum: Phys.Lett.B 744, 413--413 (2015)], 1305.2066

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1016/j.physletb.2013.06.051 2013

  22. [22]

    A set of top quark spin correlation and polarization observables for the LHC: Standard Model predictions and new physics contributions

    W. Bernreuther, D. Heisler and Z.-G. Si, A set of top quark spin correlation and polarization observables for the LHC: Standard Model predictions and new physics contributions , JHEP 12, 026 (2015), doi:10.1007/JHEP12(2015)026, 1508.05271

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1007/jhep12(2015)026 2015

  23. [23]

    Identifying Boosted Objects with N-subjettiness

    J. Thaler and K. Van Tilburg, Identifying Boosted Objects with N-subjettiness , JHEP 03, 015 (2011), doi:10.1007/JHEP03(2011)015, 1011.2268

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1007/jhep03(2011)015 2011

  24. [24]

    A. J. Larkoski, S. Marzani, G. Soyez and J. Thaler, Soft Drop , JHEP 05, 146 (2014), doi:10.1007/JHEP05(2014)146, 1402.2657

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1007/jhep05(2014)146 2014

  25. [25]

    Angular Correlations in Top Quark Pair Production and Decay at Hadron Colliders

    G. Mahlon and S. J. Parke, Angular correlations in top quark pair production and decay at hadron colliders , Phys. Rev. D 53, 4886 (1996), doi:10.1103/PhysRevD.53.4886, hep-ph/9512264

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1103/physrevd.53.4886 1996

  26. [26]

    Hayrapetyan et al., Observation of a pseudoscalar excess at the top quark pair production threshold , Rept

    A. Hayrapetyan et al., Observation of a pseudoscalar excess at the top quark pair production threshold , Rept. Prog. Phys. 88(8), 087801 (2025), doi:10.1088/1361-6633/adf7d3, 2503.22382

    work page doi:10.1088/1361-6633/adf7d3 2025

  27. [27]

    Observation of a cross-section enhancement near the $t\bar{t}$ production threshold in $\sqrt{s}=13$ TeV $pp$ collisions with the ATLAS detector

    G. Aad et al., Observation of a cross-section enhancement near the t t production threshold in s =13 TeV pp collisions with the ATLAS detector (2026), 2601.11780

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  28. [28]

    Study of $t\bar{t}$ threshold effects in $e\mu$ differential distributions measured in $\sqrt{s}=13$\,TeV $pp$ collisions with the ATLAS detector

    G. Aad et al., Study of t t threshold effects in e differential distributions measured in s =13 TeV pp collisions with the ATLAS detector (2026), 2605.02341

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  29. [29]

    E. Boos, V. Bunichev, M. Dubinin, L. Dudko, V. Ilyin, A. Kryukov, V. Edneral, V. Savrin, A. Semenov and A. Sherstnev, CompHEP 4.4: Automatic computations from Lagrangians to events , Nucl. Instrum. Meth. A 534, 250 (2004), doi:10.1016/j.nima.2004.07.096, hep-ph/0403113

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1016/j.nima.2004.07.096 2004

  30. [30]

    J. A. Hanley and B. J. McNeil, The meaning and use of the area under a receiver operating characteristic (ROC) curve , Radiology 143(1), 29 (1982), doi:10.1148/radiology.143.1.7063747

    work page doi:10.1148/radiology.143.1.7063747 1982

  31. [31]

    HiGEN: companion code repository for the present paper , https://github.com/lev-dudko/higen (2026)

    work page 2026

  32. [32]

    Golling, L

    T. Golling, L. Heinrich, M. Kagan, S. Klein, M. Leigh, M. Osadchy and J. A. Raine, Masked particle modeling on sets: towards self-supervised high energy physics foundation models , Mach. Learn. Sci. Tech. 5(3), 035074 (2024), doi:10.1088/2632-2153/ad64a8, 2401.13537

    work page doi:10.1088/2632-2153/ad64a8 2024

  33. [33]

    B. M. Dillon, G. Kasieczka, H. Olischlager, T. Plehn, P. Sorrenson and L. Vogel, Symmetries, safety, and self-supervision , SciPost Phys. 12(6), 188 (2022), doi:10.21468/SciPostPhys.12.6.188, 2108.04253

    work page doi:10.21468/scipostphys.12.6.188 2022

  34. [34]

    H. Qu, C. Li and S. Qian, Particle Transformer for Jet Tagging (2022), 2202.03772

    work page arXiv 2022

  35. [35]

    Weyl, The Classical Groups: Their Invariants and Representations , Princeton University Press, Princeton, NJ (1939)

    H. Weyl, The Classical Groups: Their Invariants and Representations , Princeton University Press, Princeton, NJ (1939)

    work page 1939

  36. [36]

    R. H. Dalitz, On the analysis of -meson data and the nature of the -meson , Phil. Mag. Ser. 7 44, 1068 (1953), doi:10.1080/14786441008520365

    work page doi:10.1080/14786441008520365 1953

  37. [37]

    L. J. Dixon, Calculating scattering amplitudes efficiently , In Theoretical Advanced Study Institute in Elementary Particle Physics (TASI 95): QCD and Beyond , pp. 539--584 (1996), hep-ph/9601359

    work page internal anchor Pith review Pith/arXiv arXiv 1996

  38. [38]

    M. L. Mangano and S. J. Parke, Multiparton amplitudes in gauge theories , Phys. Rept. 200, 301 (1991), doi:10.1016/0370-1573(91)90091-Y

    work page doi:10.1016/0370-1573(91)90091-y 1991

  39. [39]

    S@M, a Mathematica Implementation of the Spinor-Helicity Formalism

    D. Maitre and P. Mastrolia, S@M, a Mathematica Implementation of the Spinor-Helicity Formalism , Comput. Phys. Commun. 179, 501 (2008), doi:10.1016/j.cpc.2008.05.002, 0710.5559

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1016/j.cpc.2008.05.002 2008

  40. [40]

    J. A. Aguilar-Saavedra, Single top quark production at LHC with anomalous Wtb couplings , Nucl. Phys. B 804, 160 (2008), doi:10.1016/j.nuclphysb.2008.06.013, 0803.3810

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1016/j.nuclphysb.2008.06.013 2008