Pith. Sign in

REVIEW

Positive-weight moment matching at NNLO and beyond

PaperPDF from arXiv

data files

ReviewThe objections and verdict, as JSON Paper breakdownThe authors' claims, conditions and measurements, as JSON Signed receiptProof this review has not been altered
X Bluesky LinkedIn Reddit Hacker News
Benoît Assi, Rene Poncelet, James Whitehead

High Energy Physics - Phenomenology · Submitted 28 Oct 2026 · 77 references

Pith's verdict: The reweighting works in practice and deserves peer review, but the abstract's "simultaneous NNLO accuracy" outruns what a finite tower of noisy moments plus an algebraic convergence rate establishes.

T0 · AI review

How this review was made: An AI referee (deepseek/deepseek-v4-flash) read the paper on October 31, 2026 and answered a fixed list of questions about it. T0 is the starting level every paper gets. The levels describe how deeply Pith has checked the paper, never the authors.

  1. T0 AI review An AI read the paper and answered a fixed list of questions.
  2. T1 Audit Automatic checks that the references and files the paper relies on exist and match.
  3. T2 Compute The paper's code or calculations re-run with the same results.
  4. T3 Formal statement The claims written out in a formal math language a computer can read.
  5. T4 Proof Those claims proved, and the proof checked by computer.

How Pith checks papers →

· 2 major objections · 6 minor · Ask about this paper · The compiled record · Read the full review →
Challenge this review Comment (0) Read on arXiv ↗
Cite ▾
arXiv ID2610.38712 (version 1)
Pith IDQHQCI356, Pith's permanent record of this paper and its review
CitationBeno\^it Assi, James Whitehead, Rene Poncelet. "Positive-weight moment matching at NNLO and beyond." arXiv:2610.38712 (2026). https://pith.science/paper/QHQCI356
Cite this reviewBibTeX for Pith's review of this paper
Claim paper

Why Pith says this

Highlight any text to comment on it.

Referee report

How Pith checks papers → · Full audit of this paper →

The paper

The paper claims

Maximum-entropy reweighting, constrained by towers of Chebyshev moments, yields the first positive-weight event sample with simultaneous NNLO accuracy for Drell-Yan and Z+jet observables.

The reading

State-of-the-art NNLO event samples are polluted by negative event weights, which inflate variance and make them impractical for the large-scale simulation programmes the HL-LHC will require. This paper claims a way to reach NNLO accuracy with no negative weights at all: take a cheap, all-positive leading-order parton-shower sample and tilt its event weights by a smooth exponential factor chosen, by maximum entropy, so that a finite tower of Chebyshev moments of NNLO calculations is reproduced exactly. Applied to a leading-order Pythia Drell-Yan sample, the method matches NNLO to sub-percent level on constrained observables while the same sample simultaneously carries NNLO-accurate Z+jet recoil, with effective sample size retained at 99.9%; even held-out observables that were never constrained move toward NNLO through correlations inherited from the prior. If the claim stands, the either/or between perturbative accuracy and statistical power disappears, and the method extends to higher orders, higher multiplicities, and any process with a fixed-order calculation.

What carries the argument

The load-bearing object is the maximum-entropy tilt, the unique minimal-entropy-distance change of the prior weights that reproduces the imposed expectation values, with closed form $q_i = p_i e^{\lambda\cdot m_i}/Z(\lambda)$. Three devices carry the argument. Chebyshev moment towers map each constrained observable monotonically to $[-1,1]$ and fix its low-order polynomial expectations, making the log-tilt a polynomial in that observable and keeping the reweighting smooth and free of binning artefacts. Smooth $C^2$ window functions blend fixed-order moment targets into the prior's own moments outside each window, so that below a matching scale ($v_m = 30$ GeV for $p_T^{\ell\ell}$) the prior's resummation is preserved and above $2v_m$ fixed order dominates. The convergence argument bounds the total-variation distance between reweight and target by the best degree-$N$ polynomial approximation error of the target-to-prior log-density ratio: geometric $O(\rho^{-N})$ when that ratio is analytic in a Bernstein ellipse, algebraic $O(N^{-3})$ for the windowed $p_T^{\ell\ell}$ target that is only $C^2$. Performance is measured by the Kish effective sample size, which stays near 100% because the tilt preserves positivity and leaves weights nearly uniform.

Extended reading notes

Core claim

The central claim is that simultaneous NNLO accuracy can be imposed on a strictly positive-weight event sample directly, without rewriting NNLO subtraction calculations in a manifestly positive form and without interfacing NNLO codes to a parton shower. From a prior of positive-weight events, the paper takes the normalized reweighting $q_i = p_i e^{\lambda\cdot m_i}/Z(\lambda)$ that reproduces the fixed-order expectations of a vector of measurement functions and is closest to the prior in relative entropy; this weight is positive by construction, unique, and smooth in the constrained observables. The constraints are towers of Chebyshev moments of IR-safe observables — dilepton mass, rapidity, and pair transverse momentum, plus Z+jet recoil — mapped monotonically to $[-1,1]$, with smooth windows that blend each fixed-order target into the prior's own moments outside its region of validity, so the shower's resummation is kept in singular regions while fixed order takes over above the matching scale. On a $2.1\times10^6$-event leading-order Pythia sample, the result is a sample whose constrained Drell-Yan distributions agree with NNLO to sub-percent level and whose Z+jet recoil is simultaneously NNLO-accurate, with per-event tilts between 0.96 and 3.95 and effective sample size at 99.9% of the input. Held-out observables not used in the reweighting — the leading-lepton transverse momentum and $\phi^*_\eta$ — also move from the LO prior toward NNLO, which the authors attribute to kinematic correlations the prior already encodes; the same machinery applied to gluon-fusion Higgs production yields NNLO-accurate rapidity with NLO H+jet recoil under an N$^3$LO normalisation.

Load-bearing premise

The method changes weights, not events, so its accuracy on anything outside the constrained observables rests on finitely many moments capturing the target-to-prior log-density ratio and on correlations the prior already carries; the authors acknowledge that the analyticity needed for fast convergence 'should not be expected to hold for arbitrary IR-safe observables,' and for the windowed $p_T^{\ell\ell}$ target convergence is only algebraic, $O(N^{-3})$.

Editorial extensions

If this is right

  • Because the constraints come from independent fixed-order calculations, the method applies to any process, multiplicity, and perturbative order for which moments can be computed, with no need for positive-weight NNLO subtraction schemes or a direct NNLO-to-shower interface.
  • Positive weights and near-full effective sample size remove the statistical bottleneck that makes current NNLO samples impractical at scale; on a CPU-per-effective-event basis the method beats NLO and NNLO matched generators, and the fixed cost of computing NNLO moments is amortized by an effective sample size of about $10^8$ relative to NLO methods.
  • Existing samples are systematically improvable: when higher-order or better-converged moments become available, the tilt is re-solved and the same events acquire higher accuracy, so precision is no longer fixed at generation time.
  • The windowed blending lets a single sample carry the shower's all-orders resummation in singular regions and NNLO accuracy in the hard region, demonstrated for $p_T^{\ell\ell}$ down to 30 GeV against 13 TeV LHC data.
  • The machinery transfers beyond vector-boson production: the gluon-fusion Higgs demonstration, with N$^3$LO normalisation and 88% effective sample size, shows the method works for other processes and final states.

Reading between the lines

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

  • A natural stress test the authors leave implicit: apply the method to a genuinely high-dimensional final state such as top-pair production, where no Born-level kinematic relation locks held-out observables to the constrained ones; the transfer of accuracy to unconstrained distributions should degrade, and mapping that degradation would delimit the method's practical domain.
  • Because tower depth is set by a signal-to-noise rule on the input moments, the method would benefit from a community convention of publishing complete moment towers with Monte Carlo and scale errors, turning moment data into a portable precision interface between fixed-order codes and shower samples.
  • The paper notes energy-flow polynomials as a candidate extension; carrying that through — constraining joint EFP moments of the full event — would generalise the method to jet-substructure and multi-differential correlation observables that single-variable towers cannot reach, and would directly probe how far moment matching scales.
Original abstract

Parton-shower simulations at state-of-the-art perturbative accuracy carry many negative event weights, which dilute their statistical power. For many purposes, this makes them unusable. We formulate a method for the generation of positive-weighted event samples which reproduce, up to a controllable approximation, suites of IR-safe observables with the highest perturbative accuracy of any available calculation. This method is arbitrarily differential in phase space, and free from unphysical artefacts such as binning-dependence. Applied to a leading-order Pythia sample, the method gives simultaneous NNLO accuracy for Drell--Yan and $Z+\text{jet}$ observables in a positive-weighted event sample, for the first time. The method generalises immediately to higher orders of perturbation theory, higher multiplicities, and other processes, observables, and event generators.

The compiled record

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-10-30: 1 request, 4,941 input tokens.

What the abstract itself states, as probabilities: claims first 0.88 · states limitation 0.77 · human participants 0.01 · controlled comparison 0.09 · randomized assignment 0.01 · states quantitative result 0.15 · compares with prior results 0.77 · mentions code or data availability 0.03.

The abstract, sentence by sentence

arXiv listing abstract v1. Results and the main claim are underlined; .

Parton-shower simulations at state-of-the-art perturbative accuracy carry many negative event weights, which dilute their statistical power. For many purposes, this makes them unusable. We formulate a method for the generation of positive-weighted event samples which reproduce, up to a controllable approximation, suites of IR-safe observables with the highest perturbative accuracy of any available calculation. This method is arbitrarily differential in phase space, and free from unphysical artefacts such as binning-dependence. Applied to a leading-order Pythia sample, the method gives simultaneous NNLO accuracy for Drell--Yan and $Z+\text{jet}$ observables in a positive-weighted event sample, for the first time. The method generalises immediately to higher orders of perturbation theory, higher multiplicities, and other processes, observables, and event generators.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#sentencerolethe authors' own workstates the main conclusionstrength
s0 Parton-shower simulations at state-of-the-art perturbative accuracy carry many negative event weights, which dilute their statistical power. background 1.00 0.11 0.03 asserted 1.00
s1 For many purposes, this makes them unusable. background 0.93 0.08 0.03 asserted 0.96
s2 We formulate a method for the generation of positive-weighted event samples which reproduce, up to a controllable approximation, suites of IR-safe observables with the highest perturbative accuracy of any available calculation. contribution 0.85 0.94 0.73 asserted 0.99
s3 This method is arbitrarily differential in phase space, and free from unphysical artefacts such as binning-dependence. contribution 0.59 0.80 0.09 asserted 1.00
s4 Applied to a leading-order Pythia sample, the method gives simultaneous NNLO accuracy for Drell--Yan and $Z+\text{jet}$ observables in a positive-weighted event sample, for the first time. result 0.98 0.93 0.40 asserted 1.00
s5 The method generalises immediately to higher orders of perturbation theory, higher multiplicities, and other processes, observables, and event generators. interpretation 0.43 0.58 0.17 asserted 0.99

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from simulation or computation 0.95.

What was examined, and what was not

sources read
  • arXiv listing abstract v1
not examined
  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers
levels run
  • P abstract
not run
  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

Figures
Figure 1. FIG. 1. Ratios to NNLO of the differential cross-sections of
Figure 2. FIG. 2. MaxEnt, the LO+PS prior, fixed-order NNLO DY and
Figure 3. FIG. 3. Performance comparison of alternative generation methods (a) per marginal additional effective event, and (b) for
Figure 4. FIG. 4. Ratio to NNLO of the dilepton mass, constrained,
Figure 5. FIG. 5. Gluon-fusion Higgs production: the diphoton trans

Figures reproduced from arXiv:2610.38712 by the authors.

References

Discussion

Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.