{"id":"d344a2cc-3254-46fa-8425-bff852339032","arxiv_id":"2601.05332","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Using an information-theoretic bound on the halo velocity distribution, dark matter direct-detection limits vary from <10x to ~10,000x near threshold depending on the EFT operator, with higher velocity-weighting operators most affected.","lead":"This paper calculates how much our ignorance of the dark matter velocity distribution in the Milky Way changes the exclusion limits set by XENONnT and PICO60, for every type of dark matter-nucleus interaction in the non-relativistic effective theory. It finds that the uncertainty is tiny for some interaction types but up to 10,000x near the energy threshold for others, depending on how strongly the interaction rate weights high-speed dark matter.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (10)'s KL direction plus an unstated velocity support makes the 'most conservative' limits ill-defined; a 10% shell at v_esc is allowed at D_KL≈0.1 and changes O5-like rates by orders of magnitude.","rationale":"The reader's weakest assumption identifies the KL budget as the main concern. I agree that the chosen K is physically motivated rather than derived, but the more load-bearing and more specific problem is the combination of the divergence direction in Eq. (10) and the missing velocity-support constraint. With D_KL(fMB||f), distributions with a small high-speed component enter the allowed set at tiny KL cost, and without an explicit f(v)=0 for v>v_esc the optimization is formally unbounded. Since the paper never states the numerical bounds used in the CVXPY/WimPyDD implementation, the finite factors quoted in the abstract and conclusions are not reproducible from the manuscript alone. This does not destroy the qualitative hierarchy — operators with higher velocity weighting will generally be more sensitive — but it does undermine the central quantitative claim that the paper has bracketed the astrophysical uncertainty with 'most conservative' limits. The appropriate disposition remains CONDITIONAL: the paper should specify and justify the velocity support, report the KL divergences of the comparison simulation distributions, and release the adapted code so the optimization can be checked. I therefore recommend keeping the reader's CONDITIONAL verdict while strengthening the specific condition.","tokens_in":21469,"tokens_out":10808,"duration_ms":127628,"concrete_test":"Take O5 with XENONnT near threshold (e.g., mχ = 10 GeV). Construct f(v) = (1-ε)fMB(v) + ε δ(v-v_esc) (or a narrow Gaussian shell of width ~1 km/s at v_esc) and solve for ε such that D_KL(fMB||f) = 0.1 (ε ≈ 0.095). Compute the O5 rate ∫_{v_min} v^3 f and the implied 90% C.L. limit using the same exposure/background as the paper. Compare with the paper's maximized rate for K=0.1. If the constructed rate exceeds the paper's maximized rate, the optimizer either missed an allowed distribution or used an extra implicit constraint; if it is below, inspect the code for the velocity-grid maximum. Separately, rerun the optimization with v_max = v_esc and v_max = 2v_esc (or no cutoff) and check whether the O5 limit changes materially. If it does, the numerical hierarchy is a function of a hidden support parameter.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Eq. (10) defines D = ∫ fMB log(fMB/f), i.e., D_KL(fMB||f). For f = (1-ε)fMB + ε δ(v-v_esc), D ≈ -log(1-ε) ≈ ε (the δ term contributes zero because fMB(v_esc) is finite and the support width shrinks; the bulk term is -log(1-ε)). Thus a 10% shell at the escape speed is inside the K=0.1 ball. The O5 rate in Eq. (19) contains ∫ v^3 f, whose ε v_esc^3 term dominates near threshold, where the SHM integral over v ≥ v_min is tiny. The optimization maximum over the stated constraints (11)-(13) is therefore dominated by where that 10% mass is allowed to sit.\n\nThe paper says 'galactically-bounded only' when motivating K, but constraints (11)-(13) contain no f(v)=0 for v>v_esc and no other upper bound. If no bound is imposed, the maximum is unbounded by moving the shell to arbitrarily large v (D stays ≈ε), so the finite 'most conservative' curves in Figs. 1-3 necessarily come from an unstated grid cutoff or boundary in the WimPyDD/CVXPY implementation. All quoted uncertainty factors (two to four orders of magnitude) are conditional on that unstated choice. The claimed quantitative, model-independent quantification is therefore not yet well-posed unless the velocity support is specified and justified.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends the KL-divergence-based method of Herrera & Rappelt (2024) to the full non-relativistic EFT basis for dark matter–nucleon scattering, studying XENONnT and PICO-60. For each operator, it solves a convex optimization over the local dark matter velocity distribution subject to D_KL(f_MB||f) ≤ K and normalization, obtaining 'most conservative' and 'most aggressive' 90% C.L. upper limits. It reports a clear hierarchy: O1 and O4 are relatively robust, while velocity- and momentum-suppressed operators such as O5, O7, O8, O14, and O15 can show two-to-three (or even four) orders of magnitude uncertainty near threshold. The paper also maps the rate integrals onto truncated velocity moments, providing an intuitive explanation of the hierarchy.","tokens_in":21782,"tokens_out":15652,"duration_ms":164964,"significance":"If the numerical results were well-posed, the paper would be a useful systematic extension of a recent method: it is the first KL-based, model-independent quantification of astrophysical uncertainties across the entire NR EFT operator basis, with a transparent moment interpretation. The qualitative hierarchy is physically sensible, and the method is in principle more flexible than traditional halo-independent analyses. However, as detailed below, the optimization problem is not well-posed as stated because the velocity support of f is unbounded, and the quoted numerical uncertainty factors rely on an unstated discretization cutoff. The paper's central quantitative claim (the 2–3 order-of-magnitude hierarchy) therefore needs a redefinition and re-derivation of the optimization domain before it can be accepted.","major_comments":[{"comment":"The optimization problem is not well-posed. Eq. (10) defines D_KL(f_MB || f), and f_MB is a Maxwellian truncated at v_esc (Eqs. (2)–(3)). Since f_MB(v)=0 for v>v_esc, the KL integrand vanishes there; the constraint (12) does not restrict the support of f above v_esc. For any v_L>v_esc and any ε≤1−e^{−K}, the normalized mixture f=(1−ε)f_MB+ε δ(v−v_L) satisfies D_KL(f_MB||f)=−log(1−ε)≤K and ∫f=1, yet ∫ v^n f ≥ ε v_L^n diverges as v_L→∞. Hence the 'max' in Eq. (11) is unbounded for every operator whose rate has a velocity weight n≥1 (which includes O1, O3, O5, O8, O14, etc.). The finite 'most aggressive' curves in the figures must therefore be artifacts of an unstated upper-velocity cutoff in the WimPyDD/CVXPY implementation. The authors must specify and justify the admissible velocity support (e.g., f=0 for v>v_esc, or an explicit v_max) and show the sensitivity of the quoted factors to th","section":"Sec. 3, Eqs. (10)–(13), Figs. 1–3"},{"comment":"The numerical upper limits cannot be reproduced because the statistical procedure is not specified. The paper quotes 90% C.L. upper limits on σ_DM−p in cm^2, but gives no likelihood, no observed event count, no background model, no exposure, no energy window, and no efficiency curves for XENONnT or PICO-60. The reference to 'an adapted version of WimPyDD' is insufficient. Please provide the exact experimental inputs and the limit-setting method (e.g., Poisson counting with the observed and expected background), and ideally compare the O1 SHM result with the published XENONnT limit for calibration.","section":"Sec. 3, Figs. 1–2"},{"comment":"The interpretation of the KL direction is stated backwards. The text says the KL divergence 'quantifies the amount of information lost when fMB is used to approximate the true velocity distribution f', but Eq. (10) is D_KL(f_MB || f) = ∫ f_MB log(f_MB/f), which measures the information lost when f is used to approximate f_MB (or, equivalently, the expected log-likelihood ratio under f_MB). This matters in a methodological paper because the chosen direction is asymmetric and affects which distributions are strongly penalized. Please correct the wording and explicitly note that the forward KL heavily penalizes f with suppressed support where f_MB is sizable.","section":"Sec. 3, Eq. (10)"}],"minor_comments":[{"comment":"There are numerous typos and garbled renderings: 'suh as' (Sec. 3), 'angular averaded' (Appendix A), 'we show Further, we also plot' (Sec. 3), square-root symbols in Eqs. (1) and (16) appearing as 'q', and mis-rendered table headers in Table 2. These should be corrected in a revision.","section":"Throughout"},{"comment":"The inner momentum integral in Eq. (17) is written as ∫_0^∞ dq q F(v,q), but the kinematically allowed q-range depends on v and v_min (and on the nuclear response). The expression is schematic; please specify the integration limits or state that it is a formal representation.","section":"Eq. (17)"},{"comment":"The claim that rates are proportional to ∫ v^n f(v) is only heuristic: the (v^⊥)^2 decomposition in Eq. (14) introduces v_min-dependent subtractions, and the nuclear form factors do not factor out of the q-integral. The authors acknowledge this in Appendix A, but the main text should be more careful to avoid presenting these expressions as exact rate formulas.","section":"Sec. 3, Eqs. (15), (19)"},{"comment":"The statement 'without assuming any specific functional form for the velocity distribution' is overstated: the method assumes the true distribution lies within a KL ball around a specified truncated Maxwell-Boltzmann benchmark with fixed parameters. This is acknowledged later, but the abstract could be misleading.","section":"Abstract and Sec. 3"}],"recommendation":"major_revision","confidential_remarks":"The central idea is promising and the operator classification is physically transparent, but the unbounded-support problem in the optimization is a genuine flaw that invalidates the quantitative limits as presented. The fix—imposing an explicit velocity support (e.g., v≤v_esc) and re-running the analysis—is within the scope of a major revision, so I do not recommend rejection at this stage. The paper should also be pressed to provide sufficient experimental details for reproducibility; without those, the absolute normalization of the quoted limits cannot be checked."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper has a genuinely useful core: it extends the KL-divergence approach from the author's earlier work to all 15 non-relativistic EFT operators for XENONnT and PICO60, and it gives a clean physical classification of why some operators are more sensitive to halo uncertainties. The truncated-moment appendix is a nice addition. The hierarchy—O1/O4 robust, O5/O7/O8/O14/O15 fragile—is transparent and follows from the velocity/momentum structure of the operators, so I expect it to be qualitatively correct.\n\nThe problem is that the optimization as stated is not actually finite. Eq. (10) uses D_KL(fMB||f), but the constraints (11)–(13) contain no velocity support bound. For f = (1-ε)fMB + ε δ(v-v_esc), D_KL ≈ -log(1-ε) ≈ ε, so a 10% shell at the escape speed is inside the K=0.1 ball. Without any upper bound, you can place such a shell arbitrarily far, making ∫ v^3 f (which controls O5-like rates) arbitrarily large. The finite 'most conservative' curves in Figs. 1–3 therefore have to come from an unstated grid cutoff or boundary in the WimPyDD/CVXPY setup. All quoted uncertainty magnitudes are conditional on that hidden choice.\n\nThere is a related issue with the KL budget itself. K=0.1/1 is inherited from the earlier paper and motivated by 'galactically-bounded only' distributions from simulations and observations. But a thin high-speed component can have small KL and still strongly change v^3-weighted integrals, so the orders of magnitude in the abstract are not necessarily conservative. This should be stated explicitly wherever the numbers are quoted.\n\nAlso, the statistical pipeline is not shown: no likelihood, no background model, no exposure/efficiency tables, no code or commit hash. WimPyDD is public, but the adapted version is not; a referee cannot reproduce the limits from the manuscript alone. That is addressable but it is a real gap.\n\nNone of this destroys the qualitative message, and no internal contradiction in the physics is visible. But the central quantitative claim—'uncertainties up to three or four orders of magnitude'—is not well-posed as written. I would send this to peer review, with the request that the author add an explicit velocity support constraint (e.g., f=0 for v>v_esc, or a justified v_max), describe the numerical grid and convergence, and rerun with sensitivity to that boundary. With those changes it could be a solid reference for the direct-detection community; without them, the numbers shouldn't be cited.","headline":"Useful systematic extension of KL-bounded halo uncertainties to the full NR EFT basis, but the optimization lacks a velocity support bound, so the quoted orders of magnitude are not well-defined as stated.","tokens_in":22326,"tokens_out":4212,"would_cite":false,"duration_ms":43761,"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":"Direct-detection limits can vary by up to four orders of magnitude from halo-velocity uncertainty, depending on the dark matter interaction operator.","keywords":["dark matter direct detection","non-relativistic effective field theory","Kullback-Leibler divergence","halo velocity distribution","astrophysical uncertainties","velocity moments","WIMP-nucleon scattering","convex optimization"],"falsifier":"Compute the KL divergence between the Maxwell–Boltzmann benchmark and a representative suite of galactic dark-matter speed distributions inferred from simulations and stellar kinematics; if any plausible distribution has D_KL > 0.1, or if a distribution with D_KL < 0.1 yields a v^3-weighted rate integral outside the paper's optimized band by more than a factor of a few, then the uncertainty hierarchy understates the true astrophysical spread.","tokens_in":21259,"feed_emoji":"🌌","tokens_out":6896,"duration_ms":66819,"temperature":0.7,"pith_summary":"The paper asks how much the unknown velocity distribution of galactic dark matter can change the upper limits experiments set on dark matter–nucleon scattering, for every operator in the non-relativistic effective theory. It finds that the impact is highly operator-dependent: simple spin-independent or spin-dependent operators (O1, O4) give limits that move by only about an order of magnitude even for entropically large deviations from the standard halo model, while operators that scale with higher powers of the dark matter velocity or with momentum-suppressed nuclear responses (O5, O7, O8, O14, O15) can move by two to four orders of magnitude near the experimental threshold. The analysis is parameter-free: it bounds the Kullback–Leibler divergence between the true halo distribution and the Maxwell–Boltzmann benchmark, then optimizes the event rate under that bound. The paper also shows that each operator effectively probes a specific truncated raw moment of the speed distribution—mean-like, variance-like, or skewness-like—which explains the hierarchy and tells experimenters which features of the halo matter for which model.","feed_headline":"Halo uncertainty shifts dark matter limits by 4 orders","feed_subtitle":"A KL-divergence analysis finds simple operators robust, velocity-dependent ones highly uncertain near threshold.","key_machinery":"The load-bearing tool is the Kullback–Leibler divergence between the true dark-matter speed distribution and the Maxwell–Boltzmann benchmark, used as a budget in a convex optimization: minimize and maximize the scattering rate subject to D_KL ≤ K and normalization. Solving this for each operator yields the range of 90% C.L. limits compatible with any velocity distribution within K bits of the standard halo. The analytic handle is the decomposition of each operator's rate into truncated velocity-weighted integrals ∫_{v_min}^∞ v^n f(v) dv, whose combination mirrors the conditional mean, variance, and skewness of the distribution above threshold. This moment mapping turns the numerical hierarch","core_discovery":"Using a KL-divergence budget of D_KL ≤ 0.1 (or 1) around the Maxwell–Boltzmann halo, the author computes, for all 15 Galilean-invariant NR operators, the most conservative and most aggressive 90% C.L. limits on the dark-matter–nucleon coupling from current direct-detection data. The central result is a clear hierarchy: operators whose rates are dominated by the first velocity moment, O1 and O4, are robust to astrophysical unknowns (factors of a few to ~10), whereas operators dominated by v^2 or v^3 weightings or by momentum-suppressed nuclear responses—O5, O7, O8, O14, O15—show uncertainties up to two, three, or even four orders of magnitude near the threshold. This hierarchy is explained by","pith_inferences":["If the local speed distribution's high-velocity tail were measured more precisely, the hierarchy predicts that operators O5, O8, O14, and O15 would benefit most; conversely, resources spent refining the bulk distribution would mostly affect the already-robust O1/O4 limits.","The quoted uncertainties are upper bounds for galactically bounded distributions because the KL ball also contains artificial distributions; adding physical constraints (e.g., a hard escape velocity or smoothness) inside the optimization would likely shrink the ranges.","The same KL-constrained moment optimization transfers directly to other observables that are functionals of an unknown distribution, such as the solar neutrino floor, cosmic-ray fluxes, or gravitational capture of dark matter in the Sun.","Repeating the optimization at smaller KL budgets (e.g., 0.01) would test whether the ordering of operators by sensitivity persists or whether some operators become robust faster than others; the paper reports only D_KL=0.1 and 1."],"forward_implications":["For operators O1 and O4, whose rates track the first velocity moment, limits computed under the Standard Halo Model remain valid to within about an order of magnitude, so their exclusion regions are essentially robust to astrophysical unknowns.","For operators O5, O7, O8, O14, and O15, near-threshold limits can weaken by two to four orders of magnitude when the halo is allowed to deviate within the KL budget; any exclusion claim with these operators is strongly halo-dependent.","Operators that share the same velocity and momentum dependence show nearly identical ratios of optimized to SHM limits, indicating the hierarchy is governed by the operator's moment index rather than by target or experiment.","For dark matter masses above roughly 20 GeV, all operators converge to a common modest uncertainty factor (about 3 for one experiment and 10 for another at D_KL=0.1), because the threshold velocity is low and the rate samples the bulk of the distribution.","The KL-constrained framework can be applied without parametric halo assumptions to re-interpret any current or future direct-detection dataset, including in the neutrino-floor regime."],"fun_headline_variants":["Dark matter limits swing 4 orders under halo shape uncertainty","Simple operators robust, velocity ones fragile to halo unknowns","KL divergence ranks dark matter operators by astrophysical risk","Halo unknowns: some dark matter limits shift by 10000x"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The central assumption is that all physically plausible galactic halo speed distributions lie within a KL divergence of 0.1 (or 1) from the Maxwell–Boltzmann benchmark, so that the quoted uncertainty ranges are valid; a true distribution with a small but fast-moving component could have a much smaller KL yet strongly alter the high-velocity moments that dominate operators like O5 or O14, underestimating the uncertainty.","fun_headline_variants_meta":{"raw":{"variants":["Dark matter limits swing 4 orders under halo shape uncertainty","Simple operators robust, velocity ones fragile to halo unknowns","KL divergence ranks dark matter operators by astrophysical risk","Halo unknowns: some dark matter limits shift by 10000x"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00026,"raw_usage":{"total_tokens":1449,"prompt_tokens":788,"completion_tokens":661,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":532,"completion_tokens_details":{"reasoning_tokens":593}},"tokens_in":532,"tokens_out":661,"duration_ms":8479,"temperature":1.0,"reasoning_tokens":593,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T11:41:01.399639+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the KL divergence between the Maxwell–Boltzmann benchmark and a representative suite of galactic dark-matter speed distributions inferred from simulations and stellar kinematics; if any plausible distribution has D_KL > 0.1, or if a distribution with D_KL < 0.1 yields a v^3-weighted rate integral outside the paper's optimized band by more than a factor of a few, then the uncertainty hierarchy understates the true astrophysical spread.","supporting_citations":[],"review_version":1}