{"id":"4798c84e-8383-4083-96a7-05935a16585a","arxiv_id":"2412.12963","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"MCMC via BAT.jl reproduces Sherpa's Z+3 jets distributions for two observables, but full 13-dimensional phase space validation is missing.","lead":"Researchers linked the BAT.jl MCMC library to the Sherpa event generator to test whether MCMC can sample high-energy collision phase space. They find agreement with standard sampling for Z+3 jets, though full unbiased sampling is not yet proven.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The unbiased-sampling claim hinges on the logit transform's Jacobian, which the paper never states is included; if omitted, the chain samples the wrong target density.","rationale":"I read the paper as a feasibility study: it builds a BAT.jl-Sherpa interface and validates MCMC phase-space sampling for Z+3 jets against Sherpa on the leading-lepton pT and dilepton invariant mass distributions. The central claim is that this demonstrates unbiased sampling of the fully differential cross section. The reader's weakest-assumption analysis identifies the same point that I consider most load-bearing: the logit transformation used in Section 3 requires a Jacobian correction in the target density, and the paper never states that this correction is applied. If it is missing, the chain's stationary distribution is not the desired phase-space density, so all downstream distributions, including those used for validation in Fig. 4, are generated from the wrong target. The observed agreement with Sherpa is suggestive evidence that the implementation may be correct, but it is not a substitute for an explicit statement or testable code. The other concerns noted by the reader—limited validation to two projections, absent p-values, and lack of reproducibility—are real but secondary; they affect how strongly the conclusion is supported rather than identifying a specific mechanism by which the conclusion could be wrong. The Jacobian issue is that mechanism. I recommend keeping the reader's CONDITIONAL verdict: the paper should be accepted only after the authors clarify the Jacobian handling and provide either code or a direct test, and ideally broaden the validation. My stress-test does not move the verdict because the concern is exactly the one already flagged, and the claim remains plausible pending that clarification.","tokens_in":6405,"tokens_out":5072,"duration_ms":55386,"concrete_test":"Inspect the interface code (or ask the authors) to see whether the target function evaluated after the logit transform includes the Jacobian factor exp(log(abs(x*(1-x)))) or an equivalent term. As an independent numerical check, run the same BAT.jl-plus-Sherpa workflow on a simple one-dimensional target with known analytic form, e.g. p(x) proportional to x^a(1-x)^b, using the same logit transform. Compare the sampled histogram with p(x) and with the un-Jacobian-corrected target p(x(q)); if the histogram matches p(x), the Jacobian is handled, whereas if it matches p(x(q)) only, the manuscript must correct the description or the implementation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3 introduces the logit transformation q(x)=log(x/(1-x)) and states that sampling is performed on the logit-transformed phase space, but it never states whether the target density is multiplied by the Jacobian |dx/dq|=x(1-x). For an MCMC chain whose target is originally a density p(x) in the unit-hypercube variable x, the correct target in q-space is p(q)=p(x(q)) x(q)(1-x(q)). If this factor is omitted, the stationary distribution is biased by a factor 1/[x(1-x)] relative to p(x). The paper's central conclusion—that the interface generates differential cross-section distributions 'without bias'—is directly load-bearing on this point. The agreement with Sherpa in Fig. 4 suggests the Jacobian may be handled correctly in the implementation, but the text neither confirms it nor supplies the code, so a reader cannot verify the central claim. Because every presented distribution is produced through this transformed sampling, the missing Jacobian statement is the weakest link in the argument.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents an interface between the BAT.jl MCMC library and the Sherpa event generator, and uses it to sample the phase space of a Z+3 jets process (specifically the partonic channel gg -> e+e- d dbar g at 13 TeV) with a Metropolis-Hastings algorithm and RAM tuning. A logit transformation is applied to the phase-space variables to improve sampling efficiency, and the resulting samples are compared with a standalone Sherpa run via two one-dimensional distributions (leading-lepton pT and dilepton invariant mass). The authors conclude that MCMC algorithms can sample the fully differential cross section without bias.","tokens_in":6547,"tokens_out":6450,"duration_ms":61957,"significance":"If substantiated, this is a useful proof-of-principle for MCMC-based phase space sampling as an alternative to traditional importance-sampling event generators, with potential benefits for multi-modal and high-dimensional phase spaces. The work leverages existing, well-tested tools (BAT.jl, Sherpa) and performs a direct comparison against an independent generator, which is a sound self-consistency check. The paper also raises a legitimate practical issue, autocorrelation, and attempts to account for it via effective sample size matching. However, the central unbiasedness claim currently rests on a small amount of validation and on a missing technical detail (the Jacobian of the logit transformation), so the evidence is not yet commensurate with the strength of the conclusion.","major_comments":[{"comment":"The logit transformation is introduced without stating the Jacobian of the transformation. For a target density p(x) defined on the unit hypercube, sampling in q = log(x/(1-x)) requires the transformed density p(q) = p(x(q)) |dx/dq| = p(x(q)) x(q)(1-x(q)). If this Jacobian factor is omitted, the chain's stationary distribution is biased by a factor 1/[x(1-x)] relative to p(x). Since the unbiasedness claim in Section 4 rests on sampling in this transformed space, the paper must state explicitly how the Jacobian is handled (e.g., in the target function passed to the sampler) and provide the corresponding derivation or a precise reference to the implementation.","section":"Section 3"},{"comment":"The conclusion that the MCMC samples reproduce the differential cross section 'without bias' is supported only by visual agreement of two one-dimensional distributions. The p-values from the claimed chi2 test are not reported in the text or figure, and no other observables or multi-dimensional correlations are shown. For a 13-dimensional phase space, agreement in these two marginals is necessary but not sufficient to establish unbiased sampling. Please report the p-values and chi2/dof values, and add further validation such as jet transverse momenta, angular separations, and at least one two-dimensional distribution or a multivariate two-sample test.","section":"Section 3, Fig. 4"},{"comment":"The effective sample size formula is incorrect as written: ESS = 1/(1 + 2*N*sum_{k} rho(k)) should read ESS = N/(1 + 2*sum_{k} rho(k)). The current formula would give a dimensionally inconsistent result and would misestimate the number of independent samples. Because the comparison in Section 3 matches the number of Sherpa events to the ESS, this must be corrected and the impact on the uncertainty estimate and on the comparison checked.","section":"Section 2.2, Eq. (7)"}],"minor_comments":[{"comment":"The number of MCMC samples, burn-in length, number of chains, and the Gelman-Rubin convergence values are not reported; these details are needed to assess the reliability of the sampling and the convergence criterion.","section":"Section 3"},{"comment":"The p-values mentioned in the text and figure caption are not displayed; please provide the numerical p-values (and chi2/dof) either in the figure, its caption, or the text.","section":"Section 3, Fig. 4"},{"comment":"The paper states that the number of Sherpa events is matched to the ESS, but it does not specify the actual numbers or the target ESS used; please provide these values to make the statistical comparison reproducible.","section":"Section 3"},{"comment":"The text contains numerous LaTeX artifacts from the source, such as 'di fferent', 'di fferential', and 'S HERPA'; these should be cleaned before publication.","section":"Throughout"},{"comment":"The process definition should clarify whether the Z boson is treated as on-shell or whether the dilepton invariant mass distribution is generated directly with off-shell effects; the cut m(l+l-) > 66 GeV suggests the latter, but this is not stated.","section":"Section 3"},{"comment":"It is not stated whether the events written to the HEPMC file are unweighted (as the MCMC samples would be after convergence) or whether event weights are included; this should be clarified for the comparison with Sherpa.","section":"Section 3"}],"recommendation":"major_revision","confidential_remarks":"This is a proceedings-style paper, so the level of detail is understandably limited. However, the conclusions are currently stronger than the evidence: the missing Jacobian statement is a technical point that can be fixed by a short derivation, and the validation should be augmented with p-values and additional observables before the 'without bias' claim is supported. The paper does not appear to suffer from circularity, as it compares two independent sampling algorithms for the same target. No concerns about novelty or attribution; the main issue is the gap between claims and documentation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a feasibility study, and it reads like one. The genuinely new pieces are the BAT.jl–Sherpa interface and the use of a logit transform to keep the MCMC chain away from the boundary-dominated regions of phase space. The Z+3 jets test case is a reasonable first demonstration, and the fact that two kinematic distributions agree with Sherpa's own importance sampling is encouraging.\n\nThe main soft spot is exactly what the stress-test flagged: the Jacobian of the logit transformation is never mentioned. The paper says sampling is done on the logit-transformed phase space, but it doesn't state whether the target density has been multiplied by |dx/dq| = x(1-x). Without that correction, the chain's stationary distribution in x-space is biased by 1/[x(1-x)], and the \"without bias\" conclusion would be wrong. The agreement in Fig. 4 could mean the correction is in the code but not the text, or it could mean the bias happens to be small in these projections. I can't tell from the paper. That has to be clarified.\n\nRelated to that, the validation is thin. Only two 1D projections are shown, and although the text says p-values from a chi2-test were computed, the values aren't actually reported. No multidimensional correlations are checked, and no timing or efficiency numbers are given. The paper is upfront that timing is future work, so that's not a flaw by itself, but the unbiasedness claim would be stronger if the comparison covered more observables.\n\nThe paper is well written and honest about the autocorrelation and ESS challenges. The self-consistency check against Sherpa is not circular; it's comparing two independent sampling algorithms on the same target. I don't see any fit parameter or forced agreement.\n\nBottom line: this is a legitimate incremental contribution, but the missing Jacobian statement and the narrow validation keep me from endorsing the central conclusion as it stands. I'd send it to peer review, but it needs a revision that explicitly addresses the Jacobian and broadens the comparison. If that's done, it's a solid proceedings paper for the CHEP audience.","headline":"Useful feasibility study of MCMC phase-space sampling, but the unbiased-sampling claim needs the logit-Jacobian addressed before I'd trust it.","tokens_in":7116,"tokens_out":4072,"would_cite":false,"duration_ms":36541,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"MCMC can sample the fully differential cross section of Z+3 jets without bias","keywords":["Markov Chain Monte Carlo","phase space sampling","event generation","differential cross section","Z+3 jets","autocorrelation","effective sample size","logit transformation"],"falsifier":"Run the same interface on a low-dimensional toy target with a known density concentrated near the boundary, apply the logit transformation with and without the Jacobian factor, and compare the recovered marginal distribution to the true one; the match identifies the correct implementation. Alternatively, rerun the Z+3-jet comparison with the Jacobian intentionally omitted and check whether the leading-lepton $p_T$ and $m(\\ell\\ell)$ distributions shift away from the standalone-generator reference.","tokens_in":6182,"feed_emoji":"🎲","tokens_out":7993,"duration_ms":66754,"temperature":0.7,"pith_summary":"This paper aims to establish that Markov Chain Monte Carlo (MCMC) methods can serve as an unbiased sampler for the fully differential cross section of a high-energy scattering process, not just as an integrator. The authors connect a general-purpose MCMC toolkit to an existing event generator and use a Metropolis–Hastings chain to explore the 13-dimensional phase space of Z+3-jet production at the LHC. After transforming variables with a logit map and correcting for autocorrelation through the effective sample size, they find that the resulting distributions of the leading lepton transverse momentum and the dilepton invariant mass agree with the generator's own samples within statistical uncertainties. If correct, the result opens a route to event generation that avoids explicit multi-channel importance-sampling weights and can be driven by any MCMC algorithm in the toolkit.","feed_headline":"MCMC can sample collider phase space without bias","feed_subtitle":"A new interface generates Z+3-jet events whose distributions match a standard event generator within uncertainties.","key_machinery":"The central object is the interface between the BAT.jl MCMC package and the Sherpa event generator. A modified Sherpa executable initializes the process, then halts and waits for random numbers; BAT.jl proposes a point in the (logit-transformed) unit hypercube, Sherpa maps those numbers to four-momenta via Rambo or its multi-channel phase-space mapping, evaluates the matrix element and phase-space factors, and returns the event weight as the MCMC target density. The logit transformation $q(x)=\\log(x/(1-x))$ is the main device for making the target tractable, and the effective-sample-size correction with randomized resampling is the device that converts correlated chain output into usable, approximately independent events.","core_discovery":"The paper claims that MCMC algorithms can be used to sample the phase space of a scattering process and produce distributions of differential cross sections without bias. For the concrete case of $gg\\to e^+e^- d\\bar d g$ at $\\sqrt{s}=13$ TeV, the authors run a Metropolis–Hastings sampler with RAM proposal tuning and the Gelman–Rubin convergence criterion, using the event generator to evaluate the matrix element for every proposed phase-space point. Sampling is performed on logit-transformed variables $q(x)=\\log(x/(1-x))$ to move probability mass away from the boundaries, and the chain is run long enough that the effective sample size, estimated from autocorrelation, reaches the target. A final randomized resampling turns the correlated chain into approximately independent events; comparing the $p_T$ of the leading lepton and the invariant mass $m(\\ell\\ell)$ against a standalone generator run shows agreement within statistical uncertainties, with no bias observed.","pith_inferences":["The paper does not state whether the target density used by the sampler includes the Jacobian factor $|\\mathrm{d}x/\\mathrm{d}q|$ of the logit transformation; if that factor is missing, the chain would converge to a biased density, so reproducing the result requires checking this implementation detail.","The reported autocorrelation lags of 300–1000 to reach 10% decorrelation imply that higher-multiplicity final states will need substantially longer chains, making wall-clock time per independent event the real test of practical viability.","The same randomized-resampling recipe could be used as a principled unweighting procedure for MCMC samples, drawing events according to their weight instead of relying on a maximum-weight unweighting step."],"forward_implications":["MCMC phase-space sampling can produce physical distributions without constructing an explicit importance-sampling channel for every resonance or singularity.","The same interface works with any sampler provided by the MCMC toolkit, so Metropolis–Hastings is only one of several algorithms that can drive the exploration.","Autocorrelation is the main price: raw chain samples are not independent, and the number of useful events equals the effective sample size, not the chain length.","Because output is written in HEPMC format, MCMC-generated events can enter the same detector-simulation and analysis chain as conventional generator events."],"supporting_citations":[{"why":"Supplies the MCMC sampling algorithms, convergence diagnostics, and the software environment in which the chains are run.","marker":"[9]"},{"why":"Sherpa event generator; provides the matrix elements, phase-space mappings, and the executable that evaluates the target density for each proposed point.","marker":"[6-8]"},{"why":"Introduces the Metropolis algorithm, the base sampling mechanism on which the study's chains rely.","marker":"[5]"},{"why":"Generalizes the acceptance criterion to arbitrary proposal distributions, forming the Metropolis–Hastings sampler used here.","marker":"[15]"},{"why":"Defines the RAM adaptive proposal-tuning scheme used to set the Metropolis–Hastings step sizes.","marker":"[23]"},{"why":"Supplies the Gelman–Rubin criterion used to judge convergence of the multiple chains.","marker":"[24]"},{"why":"Rambo mapping, one of the phase-space parameterizations used to translate sampled random numbers into on-shell four-momenta.","marker":"[10]"}],"fun_headline_variants":["MCMC samples collider phase space unbiasedly","No bias: MCMC handles Z+3 jet phase space","MCMC joins Sherpa for unbiased event generation","Unbiased Z+3 jet sampling via MCMC"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The unbiasedness result rests on the MCMC target density being exactly proportional to the differential cross section after the logit coordinate change; the paper never states whether the required Jacobian factor $|\\mathrm{d}x/\\mathrm{d}q|$ was included, so if it was omitted the chain would converge to a biased distribution.","fun_headline_variants_meta":{"raw":{"variants":["MCMC samples collider phase space unbiasedly","No bias: MCMC handles Z+3 jet phase space","MCMC joins Sherpa for unbiased event generation","Unbiased Z+3 jet sampling via MCMC"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000791,"raw_usage":{"total_tokens":3413,"prompt_tokens":799,"completion_tokens":2614,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":415,"completion_tokens_details":{"reasoning_tokens":2551}},"tokens_in":415,"tokens_out":2614,"duration_ms":18298,"temperature":1.0,"reasoning_tokens":2551,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T13:32:20.386193+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same interface on a low-dimensional toy target with a known density concentrated near the boundary, apply the logit transformation with and without the Jacobian factor, and compare the recovered marginal distribution to the true one; the match identifies the correct implementation. Alternatively, rerun the Z+3-jet comparison with the Jacobian intentionally omitted and check whether the leading-lepton $p_T$ and $m(\\ell\\ell)$ distributions shift away from the standalone-generator reference.","supporting_citations":[{"cited_title":"BAT.jl -- A Julia-based tool for Bayesian inference","cited_arxiv_id":"2008.03132","evidence_quote":"Supplies the MCMC sampling algorithms, convergence diagnostics, and the software environment in which the chains are run."},{"cited_title":"Metropolis, A.W","cited_arxiv_id":null,"evidence_quote":"Introduces the Metropolis algorithm, the base sampling mechanism on which the study's chains rely."},{"cited_title":"Hastings, Biometrika 57, 97 (1970)","cited_arxiv_id":null,"evidence_quote":"Generalizes the acceptance criterion to arbitrary proposal distributions, forming the Metropolis–Hastings sampler used here."},{"cited_title":"Vihola, Statistics and Computing 22, 997 (2012)","cited_arxiv_id":null,"evidence_quote":"Defines the RAM adaptive proposal-tuning scheme used to set the Metropolis–Hastings step sizes."},{"cited_title":"Gelman, D.B","cited_arxiv_id":null,"evidence_quote":"Supplies the Gelman–Rubin criterion used to judge convergence of the multiple chains."}],"review_version":1}