Pith. sign in

REVIEW 3 major objections 4 minor 66 references

ARCANE Reweighting: A Monte Carlo Technique to Tackle the Negative Weights Problem in Collider Event Generation

T0 review · 3 major / 4 minor · reviewed 2026-08-08 · deepseek-v4-flash

Pith's one-line read Adding a zero-integral correction to each Monte Carlo event weight removes negative weights without biasing any physical observable, by redistributing hidden generator pathways after the event is complete.

desk verdict Exact additive-reweighting identity with a clear, honest statement of what is unproven; the missing demonstration is explicitly deferred to a placeholder companion. read the letter →

arxiv 2502.08052 v1 pith:XMWODU7Z submitted 2025-02-12 hep-ph hep-ex

classification hep-phhep-ex
keywords negativeweightsMonteCarloeventgenerationMC@NLOARCANEreweightingadditivelatentattributessignproblemvariancereduction
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper introduces ARCANE reweighting, a Monte Carlo technique for reducing or eliminating negatively weighted events in collider event generation. The core move is to add a correction term to each event's weight that redistributes contributions among the hidden generator pathways leading to the same visible event, while leaving the marginal distribution of the visible event exactly unchanged. Because the correction is applied after the whole event has been generated, the technique escapes the usual forward-chain Monte Carlo structure and can reach combinations of event histories that earlier techniques cannot. If the redistribution function satisfies a simple sign condition and the visible distribution is non-negative, negative weights disappear entirely; in general the sign problem drops to its irreducible floor. The payoff would be fewer simulated events needed for a fixed precision, without changing matching or merging prescriptions and without inducing correlations between events.

What carries the argument

The central object is the ARCANE redistribution function $G_{(V,H)}$, a possibly signed function on visible-and-hidden event attributes whose marginal over hidden attributes $H$ is zero for every visible event $V$. It enters as an additive correction to the weight: $W_{ARCANE} = W_{MC} + G_{(V,H)}(V,H)/P^{MC}_{(V,H)}(V,H)$, which is exactly the control-variate construction generalized from a single sampling step to an entire event-generation pipeline. The function is required to have support only where $P^{MC}_{(V,H)}$ has support, and the optimal choice $G^*$ makes all events with the same $V$ carry the same weight, $E^{MC}[W | V]$. A weaker but sufficient sign condition, Eq. (25), guarantees the sign problem drops to $S/P[V]$, the irreducible floor coming from the visible quasi-density itself. A quasi-density is an unnormalized, possibly signed target distribution; the visible quasi-density $F^{MC}_V$ is what physical observables depend on.

What would settle it

Use the companion paper's electron-positron to quark-antiquark plus one-jet setup, where the exact prediction is known and the construction of $G$ is explicit, and compare a fine histogram of a physical observable from a large ARCANE-reweighted sample with the exact prediction; any deviation larger than the ordinary weighted-sample Monte Carlo error would refute the unbiasedness claim. For the independence claim, check that weights of different events have zero covariance and that confidence intervals from the standard independent-sample formulas cover at the nominal rate.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that any weighted Monte Carlo pipeline $P^{MC}_{(W,V,H)}$ can be reweighted by adding $W_\Delta = G_{(V,H)}(V,H) / P^{MC}_{(V,H)}(V,H)$ to every event weight, as long as the ARCANE redistribution function $G_{(V,H)}$ integrates to zero over the hidden attributes $H$ for each visible event $V$ and vanishes wherever the pipeline produces no events. The reweighted sample then models the same visible quasi-density $F^{MC}_V$, so no physical observable is biased and the events remain independent draws. If $G$ is chosen so that the sign of $F^{MC}_{(V,H)} + G_{(V,H)}$ agrees with the sign of $F^{MC}_V$ wherever $F^{MC}_V$ is nonzero, the sign problem measured by $S/P$ is reduced to its irreducible value $S/P[V]$; when $F^{MC}_V$ is non-negative, this removes all negative weights. The paper frames ARCANE as a deferred reweighting that operates after the entire event is generated, thereby escaping the 'forward chain Monte Carlo' paradigm that constrained earlier theory-side and single-step Monte Carlo-side fixes.

Load-bearing premise

The practical promise rests on being able to build a good redistribution function $G_{(V,H)}$ and to compute the sampling density $P^{MC}_{(V,H)}$ for every event in a real event generator; if that construction is not tractable, the method remains exact but offers no computational gain.

Editorial extensions

If this is right

  • MC@NLO-type negative weights can be removed or reduced without changing any physical observable and without touching the matching and merging prescription.
  • ARCANE reweighting followed by unweighting reduces the coefficient of variation of event weights to $S/P[V]$, so fewer generated events are needed to reach a given precision.
  • ARCANE attacks only the numerator $F^{MC}_{(V,H)}$ of the weight, while rejection reweighting attacks the denominator, so the two techniques are complementary and can be applied in either order or repeatedly.
  • A redistribution function satisfying the sign condition of Eq. (25) is already as good as the optimal one for eliminating negative weights, even though residual weight variance may remain.
  • Any target quasi-density that splits into a straightforwardly samplable part plus a computable correction can be sampled by additive reweighting, opening the door to formalisms like 'hard remainder spreading' that are not required to be forward-chain Monte Carlo-able.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural next step, which the paper identifies but does not quantify, is ARCANE-aware sampling: choose the proposal density $P^{MC}_V$ with knowledge that an additive correction will follow, potentially improving global unweighting efficiency beyond what a fixed generator allows.
  • If $G_{(V,H)}$ is learned from a finite training dataset, standard weighted-analysis formulas remain valid conditional on the trained function; one could extend the method by treating the learned function as an estimated object and propagating its training-set uncertainty into final predictions.
  • Because the technique only requires a zero-marginal correction at the chosen split, the same mechanism could be applied at later pipeline stages, such as detector simulation or pile-up overlay, whenever hidden sampling pathways create weight inefficiencies there.
  • The largest gains should appear in phase-space regions where the local sign problem is severe, so analyses that target those regions should benefit more than global unweighting-efficiency metrics would suggest.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper introduces ARCANE reweighting, an additive Monte Carlo reweighting technique for reducing negative weights in collider event generation. For a generator with weighted density P^MC(V,H) and weight W^MC = F^MC(V,H)/P^MC(V,H), the method adds W_Delta = G(V,H)/P^MC(V,H) to each weight, where G integrates to zero over the hidden variables H for each visible event V. The paper proves that this preserves the quasi density F_V of the visible attributes, identifies an optimal redistribution function G* that makes all weights for a given V equal to E[W|V], and shows that any G satisfying a sign condition reduces the sign-problem metric S/P[V,H] to the irreducible lower bound S/P[V]. If F_V is non-negative, such a G eliminates negative weights. The paper also discusses the interplay with rejection reweighting and unweighting, quantifies the sign problem, presents generalizations, and argues in Appendix A that positive resampling techniques induce mutual dependence among events.

Significance. The central mathematical identities are correct and are presented with unusual clarity: Eq. (17) together with Eq. (22) exactly preserves F_V; Eq. (23) gives the optimal conditional-expectation weight; and the proof in Eqs. (65)-(68) that a sign-condition-satisfying G reduces S/P[V,H] to S/P[V] is valid. The Appendix A analysis of non-IID behavior in positive resampling is a useful and largely correct caution. If the practical construction of G(V,H) and the computation of P^MC(V,H) can be made to work for MC@NLO-type pipelines, the technique would be a significant methodological advance because it is exact, does not introduce inter-event correlations, and does not require changes to matching or merging prescriptions. The paper is, however, a theoretical framework paper: the practical applicability is explicitly a hypothesis, and the only demonstration is promised in an external companion paper with a placeholder identifier.

major comments (3)
  1. [Section 2.1, Section 2.4, and Ref. [7]] The practical promise of the method, as stated in the abstract, is that ARCANE can reduce or eliminate negative weights in collider event generation. Within this manuscript, that promise rests on an unverified premise: Section 2.1 calls the existence of good redistribution functions 'a hypothesis,' Section 2.4.1 describes the required probability tracking as 'somewhat tricky in practice,' and the only demonstration is relegated to Ref. [7], whose arXiv identifier is a placeholder (2501.YYYYY). Because the value of the technique for HEP depends on being able to construct a good G(V,H) and to compute P^MC(V,H) for realistic processes, this is a load-bearing gap rather than a presentation issue. The manuscript should either include a concrete, nontrivial construction (beyond the discrete two-pathway illustration of Eq. (28)) or be explicitly reframed as a conditional theoretical result with a real companion citation.
  2. [Section 2.3, Eq. (23)] The optimal redistribution function G* and the corresponding optimal weight E[W|V] require the pointwise value of the quasi density F^MC_V(v). In the generic collider setting after parton showering and hadronization, F^MC_V is a high-dimensional differential cross section that is not analytically available, which is precisely the object one is trying to sample. The engineering examples in Section 2.4.2 avoid this only when all discrete pathways leading to V can be enumerated, as in Eq. (28); no prescription is given for the continuous latent variables that appear in realistic generators. This is the concrete obstacle that the companion paper must resolve, and it should be acknowledged as a separate requirement rather than folded into 'tracking histories.'
  3. [Section 2.2, Eq. (21)] Condition (I) refers to the 'essential support of P^PH_{(V,H)}' but the density P^PH is not defined anywhere in the paper; the intended object is presumably F^PH_{(V,H)} or P^MC_{(V,H)}. As written, the condition does not guarantee that P^MC_{(V,H)} is nonzero on the support of G, which is what is needed for the additive term W_Delta = G/P^MC to be finite and for the reweighted procedure to be realizable. Please correct the notation and state the support condition with respect to the sampling density.
minor comments (4)
  1. [Table 1 caption] The caption says that ARCANE followed by unweighting reduces CV^2[W] 'down to /Sbbb/PbbbMC[W]'; this should be /Sbbb/PbbbMC[V] (or S/P[V]), in agreement with Eq. (71).
  2. [Abstract] The sentence 'Negatively weighted events ... significantly increases the computational resource requirements' has a subject-verb agreement error; 'increases' should be 'increase.'
  3. [Ref. [7]] The companion paper reference is incomplete: the arXiv identifier 2501.YYYYY is a placeholder and must be replaced with the actual identifier before publication.
  4. [Section 4.2, Eq. (45)] The equality condition for the lower bound on CV^2[W] is stated as 'P^MC_X proportional to |F^MC_X| and W fully determined by X.' This is correct, but it may be worth adding a sentence explaining why no separate sign-constancy condition is needed, since the point is a common source of confusion.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the ARCANE exactness argument is a self-contained algebraic identity, and the companion-paper self-citation is a promised demonstration rather than a load-bearing input to the proof.

full rationale

The paper's derivation chain is self-contained. The core construction, Eq. (16) with Eq. (22), defines an additive redistribution G(V,H) with zero H-integral for every V, which directly gives F_ARCANE_V = F_MC_V; this is an identity shown by construction, not a fitted quantity disguised as a prediction. The optimal redistribution function G* in Eq. (23) is obtained algebraically as the conditional expectation E[W|V] and is not fit to external data. The sign-problem reduction in Section 4.3 follows from the sign condition (25) by a direct integral argument, and the lower bound S/P[V] is a consequence of the definition of the metric, not an imported conclusion. The only self-citation is the companion paper Ref. [7], which is invoked as a demonstration rather than as a premise of any proof. The paper is explicit that practical constructibility of G is a hypothesis (Section 2.1) and that tracking probabilities and Jacobians is 'somewhat tricky in practice' (Section 2.4.1); these are deferred-evidence or correctness concerns, not circularity. The mathematical claims therefore do not reduce to their inputs or to the cited companion work.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The ledger is small because the paper introduces no fitted constants and no new physical objects. The single largest unprovided ingredient is the redistribution function G, which is a free function rather than a numeric parameter. The remaining assumptions are standard measure-theoretic conventions and realistic modeling assumptions about collider simulation pipelines.

free parameters (1)
  • Redistribution function G(V,H)
    The method's performance depends entirely on this unspecified function. The paper gives sufficient conditions (Section 2.2) and the optimal form G* (Section 2.3), but no explicit construction; the companion paper is expected to supply one.
assumptions (4)
  • standard math All statistical properties exist and are finite, and reference measures exist for all data attributes.
    Stated as caveats in Section 1.1 and used implicitly throughout all subsequent equations.
  • domain assumption The downstream simulation and analysis stages depend on the visible attributes V and the event weight, not on the hidden attributes H.
    Used throughout, e.g., Sections 1.2 and 2, to justify that redistributing contributions over H at fixed V leaves physical observables unchanged. This is a modeling assumption about the simulation pipeline.
  • ad hoc to paper A suitable redistribution function G can be constructed for processes of interest, and P^MC(V,H) can be computed for each event.
    Stated as a hypothesis in Section 2.1 and essential for the technique to have practical value. It is not proven in this paper and is deferred to the companion paper.
  • domain assumption For the optimality results, the event weight is a deterministic function of (V,H).
    Section 2.4.1 notes this is required to fully eliminate weight-variance within bins of V; Section 5.1 relaxes it, but then the optimality statements become approximate.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ARCANE Reweighting: A Monte Carlo Technique to Tackle the Negative Weights Problem in Collider Event Generation." pith.science (2026). https://pith.science/paper/XMWODU7Z

@misc{pith2026250208052,
  author       = {Pith},
  title        = {Pith review of: ARCANE Reweighting: A Monte Carlo Technique to Tackle the Negative Weights Problem in Collider Event Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XMWODU7Z}},
  note         = {Machine review of arXiv:2502.08052}
}
read the original abstract

Negatively weighted events, which appear in the Monte Carlo (MC) simulation of particle collisions, significantly increases the computational resource requirements of current and future collider experiments. This paper introduces and theoretically discusses an MC technique called ARCANE reweighting for reducing or eliminating negatively weighted events. The technique works by redistributing (via an additive reweighting) the contributions of different pathways within an event generator that lead to the same final event. The technique is exact and does not introduce any biases in the distributions of physical observables. A companion paper demonstrates the technique for a physics example.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 10 canonical work pages

  1. [45]

    Müller , F

    T . Müller , F . Rousselle, A. Keller and J. Novák,Neural control variates, ACM Trans. Graph. 39(6) (2020), doi: 10.1145/3414685.3417804

  2. [7]

    Shyamsundar ,A demonstration of ARCANE reweighting: Reducing the sign problem i n the MC@NLO generation of e+e− → q¯ q+ 1 jet events (2025), 2501.YYYYY

    P . Shyamsundar ,A demonstration of ARCANE reweighting: Reducing the sign problem i n the MC@NLO generation of e+e− → q¯ q+ 1 jet events (2025), 2501.YYYYY

  3. [1]

    Computational challenges for MC event generation

    A. Buckley ,Computational challenges for MC event generation, J. Phys. Conf. Ser .1525(1), 012023 (2020), doi: 10.1088/1742-6596/1525/1/012023, 1908.00167

  4. [2]

    Amoroso et al., Challenges in Monte Carlo Event Generator Software for High-Luminosity LHC, Comput

    S. Amoroso et al., Challenges in Monte Carlo Event Generator Software for High-Luminosity LHC, Comput. Softw . Big Sci. 5(1), 12 (2021), doi: 10.1007/s41781-021-00055-1, 2004.13687

  5. [3]

    Albrecht et al., A Roadmap for HEP Software and Computing R&D for the 2020s, Comput

    J. Albrecht et al., A Roadmap for HEP Software and Computing R&D for the 2020s, Comput. Softw . Big Sci. 3(1), 7 (2019), doi: 10.1007/s41781-018-0018-8, 1712.06982

  6. [4]

    Frixione and B

    S. Frixione and B. R. Webber ,Matching NLO QCD computations and parton shower simula- tions, JHEP 06, 029 (2002), doi: 10.1088/1126-6708/2002/06/029, hep-ph/0204244. 44The analogue of this downsampling approach for the related g enerative-model based techniques in Refs. [40, 41] is to make the size of the training dataset of weighted events much larger th...

  7. [5]

    Frixione, P

    S. Frixione, P . Nason and B. R. Webber , Matching NLO QCD and parton showers in heavy flavor production , JHEP 08, 007 (2003), doi: 10.1088/1126-6708/2003/08/007, hep-ph/0305252

  8. [6]

    Frixione and B

    S. Frixione and B. R. Webber , The MC@NLO 2.0 event generator (2003), hep-ph/0307146

Show all 66 references
  1. [8]

    Hoeche, F

    S. Hoeche, F . Krauss, M. Schonherr and F . Siegert,A critical appraisal of NLO+PS matching methods, JHEP 09, 049 (2012), doi: 10.1007/JHEP09(2012)049, 1111.1220

  2. [9]

    S. Höche, Introduction to parton-shower event generators , In Theoretical Advanced Study Institute in Elementary Particle Physics: Journeys Throug h the Precision Frontier: Amplitudes for Colliders , pp. 235–295, doi: 10.1142/9789814678766_0005 (2015), 1411.4085

  3. [10]

    Platzer and M

    S. Platzer and M. Sjodahl, The Sudakov V eto Algorithm Reloaded, Eur . Phys. J. Plus 127, 26 (2012), doi: 10.1140/epjp/i2012-12026-x, 1108.6180

  4. [11]

    Lönnblad, Fooling Around with the Sudakov V eto Algorithm, Eur

    L. Lönnblad, Fooling Around with the Sudakov V eto Algorithm, Eur . Phys. J. C73(3), 2350 (2013), doi: 10.1140/epjc/s10052-013-2350-9, 1211.7204

  5. [12]

    Kleiss and R

    R. Kleiss and R. V erheyen, Competing Sudakov V eto Algorithms , Eur . Phys. J. C 76(7), 359 (2016), doi: 10.1140/epjc/s10052-016-4231-5, 1605.09246

  6. [13]

    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 pa rton shower simulations , ...

  7. [14]

    Bierlich et al., A comprehensive guide to the physics and usage of PYTHIA 8.3 , SciPost Phys

    C. Bierlich et al., A comprehensive guide to the physics and usage of PYTHIA 8.3 , SciPost Phys. Codeb. 2022, 8 (2022), doi: 10.21468/SciPostPhysCodeb.8, 2203.11601

  8. [15]

    Bothmann et al

    E. Bothmann et al. , Event generation with Sherpa 3 , JHEP 12, 156 (2024), doi:10.1007/JHEP12(2024)156, 2410.22148

  9. [16]

    Weinzierl, Introduction to Monte Carlo methods (2000), hep-ph/0006269

    S. Weinzierl, Introduction to Monte Carlo methods (2000), hep-ph/0006269

  10. [17]

    G. P . Lepage, Adaptive multidimensional integration: VEGAS enhanced , J. Comput. Phys. 439, 110386 (2021), doi: 10.1016/j.jcp.2021.110386, 2009.05112

  11. [18]

    Bothmann, T

    E. Bothmann, T . Janßen, M. Knobbe, T . Schmale and S. Schumann , Exploring phase space with Neural Importance Sampling , SciPost Phys. 8(4), 069 (2020), doi:10.21468/SciPostPhys.8.4.069, 2001.05478

  12. [19]

    C. Gao, J. Isaacson and C. Krause, i-flow: High-dimensional Integration and Sampling with Normalizing Flows , Mach. Learn. Sci. Tech. 1(4), 045023 (2020), doi:10.1088/2632-2153/abab62, 2001.05486

  13. [20]

    C. Gao, S. Höche, J. Isaacson, C. Krause and H. Schulz, Event Generation with Normal- izing Flows, Phys. Rev . D 101(7), 076002 (2020), doi: 10.1103/PhysRevD.101.076002, 2001.10028. 40 SciPost Physics Submission

  14. [21]

    Heimel, R

    T . Heimel, R. Winterhalder , A. Butter , J. Isaacson, C. Krause, F . Maltoni, O. Mattelaer and T . Plehn, MadNIS - Neural multi-channel importance sampling , SciPost Phys. 15(4), 141 (2023), doi: 10.21468/SciPostPhys.15.4.141, 2212.06172

  15. [22]

    Heimel, N

    T . Heimel, N. Huetsch, F . Maltoni, O. Mattelaer , T . Plehn andR. Winterhalder , The Mad- NIS reloaded , SciPost Phys. 17(1), 023 (2024), doi: 10.21468/SciPostPhys.17.1.023, 2311.01548

  16. [23]

    Heimel, O

    T . Heimel, O. Mattelaer , T . Plehn and R. Winterhalder ,Differentiable MadNIS-Lite, SciPost Phys. 18, 017 (2025), doi: 10.21468/SciPostPhys.18.1.017, 2408.01486

  17. [24]

    Bewick et al

    G. Bewick et al. , Herwig 7.3 release note , Eur . Phys. J. C 84(10), 1053 (2024), doi:10.1140/epjc/s10052-024-13211-9, 2312.05175

  18. [25]

    Alioli, P

    S. Alioli, P . Nason, C. Oleari and E. Re, A general framework for implementing NLO cal- culations in shower Monte Carlo programs: the POWHEG BOX , JHEP 06, 043 (2010), doi:10.1007/JHEP06(2010)043, 1002.2581

  19. [26]

    Agostinelli et al

    S. Agostinelli et al. , GEANT4 - A Simulation T oolkit , Nucl. Instrum. Meth. A 506, 250 (2003), doi: 10.1016/S0168-9002(03)01368-8

  20. [27]

    Bothmann, T

    E. Bothmann, T . Childers, W . Giele, F . Herren, S. Hoeche, J. I saacson, M. Knobbe and R. Wang, Efficient phase-space generation for hadron collider event simulation , SciPost Phys. 15(4), 169 (2023), doi: 10.21468/SciPostPhys.15.4.169, 2302.10449

  21. [28]

    Bothmann, T

    E. Bothmann, T . Childers, W . Giele, S. Höche, J. Isaacson and M. Knobbe, A portable parton-level event generator for the high-luminosity LHC , SciPost Phys. 17(3), 081 (2024), doi:10.21468/SciPostPhys.17.3.081, 2311.06198

  22. [29]

    Nason, A new method for combining NLO QCD with shower Monte Carlo algorith ms, JHEP 11, 040 (2004), doi: 10.1088/1126-6708/2004/11/040, hep-ph/0409146

    P . Nason, A new method for combining NLO QCD with shower Monte Carlo algorith ms, JHEP 11, 040 (2004), doi: 10.1088/1126-6708/2004/11/040, hep-ph/0409146

  23. [30]

    Frixione, P

    S. Frixione, P . Nason and C. Oleari, Matching NLO QCD computations with Parton Shower simulations: the POWHEG method , JHEP 11, 070 (2007), doi:10.1088/1126-6708/2007/11/070, 0709.2092

  24. [31]

    Jadach, W

    S. Jadach, W . Płaczek, S. Sapeta, A. Siódmok and M. Skrzypek, Matching NLO QCD with parton shower in Monte Carlo scheme — the KrkNLO method , JHEP 10, 052 (2015), doi:10.1007/JHEP10(2015)052, 1503.06849

  25. [32]

    Frederix, S

    R. Frederix, S. Frixione, S. Prestel and P . Torrielli, On the reduction of neg- ative weights in MC@NLO-type matching procedures , JHEP 07, 238 (2020), doi:10.1007/JHEP07(2020)238, 2002.12716

  26. [33]

    Danziger , S

    K. Danziger , S. Höche and F . Siegert, Reducing negative weights in Monte Carlo event generation with Sherpa (2021), 2110.15211

  27. [34]

    J. R. Andersen, C. Gütschow , A. Maier and S. Prestel, A Positive Resampler for Monte Carlo events with negative weights , Eur . Phys. J. C 80(11), 1007 (2020), doi:10.1140/epjc/s10052-020-08548-w, 2005.09375

  28. [35]

    Nachman and J

    B. Nachman and J. Thaler , Neural resampler for Monte Carlo reweighting with preserved uncertainties, Phys. Rev . D102(7), 076004 (2020), doi: 10.1103/PhysRevD.102.076004, 2007.11586. 41 SciPost Physics Submission

  29. [36]

    J. R. Andersen and A. Maier , Unbiased elimination of negative weights in Monte Carlo samples, Eur . Phys. J. C 82(5), 433 (2022), doi: 10.1140/epjc/s10052-022-10372-3, 2109.07851

  30. [37]

    J. R. Andersen, A. Maier and D. Maître, Efficient negative-weight elimination in large high-multiplicity Monte Carlo event samples , Eur . Phys. J. C 83(9), 835 (2023), doi:10.1140/epjc/s10052-023-11905-0, 2303.15246

  31. [38]

    J. R. Andersen, A. Cueto, S. P . Jones and A. Maier , A Cell Resampler study of Negative Weights in Multi-jet Merged Samples (2024), 2411.11651

  32. [39]

    T . M. Cover and J. A. Thomas, Elements of Information The- ory, chap. 2, pp. 13–55, John Wiley & Sons, Ltd, ISBN 9780471748823, doi: https://doi.org/10.1002/047174882X.ch2 (2005), https://onlinelibrary .wiley .com/doi/pdf/10.1002/047174882X.ch2

  33. [40]

    Butter , T

    A. Butter , T . Plehn and R. Winterhalder , How to GAN Event Subtraction , SciPost Phys. Core 3, 009 (2020), doi: 10.21468/SciPostPhysCore.3.2.009, 1912.08824

  34. [41]

    Backes, A

    M. Backes, A. Butter , T . Plehn and R. Winterhalder , How to GAN Event Unweighting , SciPost Phys. 10(4), 089 (2021), doi: 10.21468/SciPostPhys.10.4.089, 2012.07873

  35. [42]

    Butter , T

    A. Butter , T . Heimel, S. Hummerich, T . Krebs, T . Plehn, A. Rou sselot and S. V ent, Generative networks for precision enthusiasts , SciPost Phys. 14(4), 078 (2023), doi:10.21468/SciPostPhys.14.4.078, 2110.13632

  36. [43]

    Nason,MINT: A Computer program for adaptive Monte Carlo integration a nd generation of unweighted distributions (2007), 0709.2085

    P . Nason,MINT: A Computer program for adaptive Monte Carlo integration a nd generation of unweighted distributions (2007), 0709.2085

  37. [44]

    Frederix and P

    R. Frederix and P . Torrielli, A new way of reducing negative weights in MC@NLO , Eur . Phys. J. C 83(11), 1051 (2023), doi: 10.1140/epjc/s10052-023-12243-x, 2310.04160

  38. [46]

    D. P . Kroese, T . Taimre and Z. I. Botev , Handbook of Monte Carlo Methods , pp. i–xix, John Wiley & Sons, Ltd, ISBN 9781118014967, doi: https://doi.org/10.1002/9781118014967.fmatter (2011), https://onlinelibrary .wiley .com/doi/pdf/10.1002/9781118014967.fmatter

  39. [47]

    Lemieux, Control V ariates , pp

    C. Lemieux, Control V ariates , pp. 1–8, John Wiley & Sons, Ltd, ISBN 9781118445112, doi: https://doi.org/10.1002/9781118445112.stat07947 (2017), https://onlinelibrary .wiley .com/doi/pdf/10.1002/9781118445112.stat07947

  40. [48]

    Botev and A

    Z. Botev and A. Ridder , V ariance Reduction , pp. 1–6, John Wiley & Sons, Ltd, ISBN 9781118445112, doi: https://doi.org/10.1002/9781118445112.stat07975 (2017), https://onlinelibrary .wiley .com/doi/pdf/10.1002/9781118445112.stat07975

  41. [49]

    S. M. Ross, Introduction to probability and statistics for engineers and sc ientists (sixth edition), pp. 1–9. Academic Press, sixth edition edn., ISBN 978-0-12 -824346-6, doi:https://doi.org/10.1016/B978-0-12-824346-6.00010-7 (2021)

  42. [50]

    J. S. Gainer , J. Lykken, K. T . Matchev , S. Mrenna and M. Park, Exploring Theory Space with Monte Carlo Reweighting , JHEP 10, 078 (2014), doi: 10.1007/JHEP10(2014)078, 1404.7129. 42 SciPost Physics Submission

  43. [51]

    Bellm, S

    J. Bellm, S. Plätzer , P . Richardson, A. Siódmok and S. Webste r , Reweighting Parton Showers, Phys. Rev . D 94(3), 034028 (2016), doi: 10.1103/PhysRevD.94.034028, 1605.08256

  44. [52]

    Mrenna and P

    S. Mrenna and P . Skands, Automated Parton-Shower V ariations in Pythia 8 , Phys. Rev . D 94(7), 074005 (2016), doi: 10.1103/PhysRevD.94.074005, 1605.08352

  45. [53]

    Bothmann, M

    E. Bothmann, M. Schönherr and S. Schumann, Reweighting QCD matrix- element and parton-shower calculations , Eur . Phys. J. C 76(11), 590 (2016), doi:10.1140/epjc/s10052-016-4430-0, 1606.08753

  46. [54]

    Bierlich, P

    C. Bierlich, P . Ilten, T . Menzo, S. Mrenna, M. Szewc, M. K. Wil kinson, A. Youssef and J. Zupan, Reweighting Monte Carlo predictions and automated fragmentation variations in Pythia 8 , SciPost Phys. 16(5), 134 (2024), doi: 10.21468/SciPostPhys.16.5.134, 2308.13459

  47. [55]

    Bierlich, P

    C. Bierlich, P . Ilten, T . Menzo, S. Mrenna, M. Szewc, M. K. Wil kinson, A. Youssef and J. Zupan, T owards a data-driven model of hadronization using normalizin g flows , SciPost Phys. 17(2), 045 (2024), doi: 10.21468/SciPostPhys.17.2.045, 2311.09296

  48. [56]

    Bierlich, P

    C. Bierlich, P . Ilten, T . Menzo, S. Mrenna, M. Szewc, M. K. Wil kinson, A. Youssef and J. Zupan, Describing Hadronization via Histories and Observables for Mont e-Carlo Event Reweighting (2024), 2410.06342

  49. [57]

    Kleiss and R

    R. Kleiss and R. Pittau, Weight optimization in multichannel Monte Carlo , Comput. Phys. Commun. 83, 141 (1994), doi: 10.1016/0010-4655(94)90043-4, hep-ph/9405257

  50. [58]

    Kobyzev , S

    I. Kobyzev , S. J. Prince and M. A. Brubaker , Normalizing flows: An introduction and re- view of current methods , IEEE Transactions on Pattern Analysis and Machine Intelli gence 43(11), 3964 (2021), doi: 10.1109/TPAMI.2020.2992934, 1908.09257

  51. [59]

    K. T . Matchev and P . Shyamsundar , OASIS: Optimal Analysis-Specific Im- portance Sampling for event generation , SciPost Phys. 10(2), 034 (2021), doi:10.21468/SciPostPhys.10.2.034, 2006.16972

  52. [60]

    N. A. Weiss, A Course in Probability , Addison–Wesley , Boston (2005)

  53. [61]

    W . H. Press and G. R. Farrar , RECURSIVE STRATIFIED SAMPLING FOR MULTIDIMEN- SIONAL MONTE CARLO INTEGRATION (1989)

  54. [62]

    K. T . Matchev , A. Roman and P . Shyamsundar , Uncertainties associated with GAN-generated datasets in high energy physics , SciPost Phys. 12(3), 104 (2022), doi:10.21468/SciPostPhys.12.3.104, 2002.06307

  55. [63]

    Butter , S

    A. Butter , S. Diefenbacher , G. Kasieczka, B. Nachman and T . Plehn, GANplifying event sam- ples, SciPost Phys. 10(6), 139 (2021), doi:10.21468/SciPostPhys.10.6.139, 2008.06545

  56. [64]

    Bieringer , A

    S. Bieringer , A. Butter , S. Diefenbacher , E. Eren, F . Gaede,D. Hundhausen, G. Kasieczka, B. Nachman, T . Plehn and M. Trabs, Calomplification — the power of generative calorime- ter models , JINST 17(09), P09028 (2022), doi: 10.1088/1748-0221/17/09/P09028, 2202.07352

  57. [65]

    Bieringer , S

    S. Bieringer , S. Diefenbacher , G. Kasieczka and M. Trabs, Calibrating Bayesian generative machine learning for Bayesiamplification , Mach. Learn. Sci. Tech. 5(4), 045044 (2024), doi:10.1088/2632-2153/ad9136, 2408.00838. 43 SciPost Physics Submission

  58. [66]

    Shyamsundar ,The Bias–V ariance–Correlation tradeoff and some of its implicat ions for machine learning in high energy physics (in preparation)

    P . Shyamsundar ,The Bias–V ariance–Correlation tradeoff and some of its implicat ions for machine learning in high energy physics (in preparation). 44

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.