{"id":"64c2dbb5-4b4d-41ef-b7f2-8842202ea259","arxiv_id":"2412.06892","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":18,"one_line_summary":"A one-loop EFT model matches ACCEL2 Lyman-alpha forest simulations well and predicts BAO scale shifts of roughly -0.2 to -0.3 percent at z=2.","lead":"Researchers fit a physics model to supercomputer simulations of the Lyman-alpha forest and find it matches the measured clustering to about 2 percent on small scales. They then calculate how nonlinear structure shifts the apparent cosmic ruler, giving DESI a theoretical error budget for its highest-redshift distance measurements.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"BAO-shift budget inherits uncontrolled one-loop truncation: Eq. (19) uses parameters fit to kmax=2 where loop/tree ratio exceeds unity (Fig. 2), and the quoted errors omit truncation uncertainty.","rationale":"The paper's central claim is that a one-loop EFT calibrated on ACCEL2 provides a reliable BAO shift error budget. This requires the fitted bias parameters entering Eq. (A1) to be unbiased estimates of the EFT parameters, not effective parameters contaminated by two-loop contributions. The paper itself undercuts this: Fig. 2 shows the loop/tree ratio crossing unity at k~1.5, the text admits the model is phenomenological there and defers a two-loop/field-level check, and Table II shows large stochastic terms at z=2. The BAO shift's dependence on kmax (Fig. 7, sign change in the radial shift) is a concrete symptom. My proposed test would settle whether the quoted numbers are robust. I agree with the reader's weakest_assumption; a conditional verdict remains appropriate. The quasar-bias issue is real but secondary, since the paper already labels those estimates indicative and the auto-correlation shift is the primary target.","tokens_in":30624,"tokens_out":5735,"duration_ms":62528,"concrete_test":"Recompute ∆α∥ and ∆α⊥ from the MCMC chains with kmax=1 and 1.5 h/Mpc (same model, priors, and stochastic inclusion) and also from the kmax=2 chain with Pshot, a0, a2 fixed to zero. If the central values shift by more than the quoted ±0.09% / ±0.11%, or if the kmax=1 and kmax=2 results disagree at ≳2σ, then Eq. (19) is sensitive to the uncontrolled UV regime and the error budget must be expanded accordingly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central error budget, Eq. (19), is a function of the fitted quadratic and cubic bias parameters through Eq. (A1). Those parameters are determined by fits extending to kmax=2 h/Mpc, yet Fig. 2 shows the one-loop-to-tree ratio crossing unity near k≈1.5 h/Mpc at z=2. The authors themselves (Sec. 6.1) describe the model as \"phenomenological\" beyond this scale and defer a definitive two-loop or field-level check. The stochastic terms, which are detected at >5σ at z=2 (Table II), are exactly the kind of counterterms that can absorb two-loop UV sensitivity; if they do, the inferred b2, bG2, bη2, etc. used in Eq. (A1) can be biased. Fig. 7 shows the radial shift changing sign as kmax is increased from 0.5 to 2 h/Mpc, demonstrating sensitivity to the arbitrary fit range. The MCMC errors quoted in Eq. (19) propagate parameter covariance only; they do not include the truncation error. If the one-loop expansion is not controlled at the kmax used, the claim that Eq. (19) provides a reliable theoretical error budget for DESI is premature.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper calibrates the one-loop effective field theory (EFT) power spectrum for the Lyman-α forest, as developed in Ref. [49], against the three-dimensional flux power spectrum measured from the ACCEL2 hydrodynamic simulations at five redshifts z = 2.0–4.0. An 18-parameter model (linear, quadratic, and cubic biases; counterterms; stochastic terms) is fitted with Gaussian priors and analytic marginalization over the linear parameters, and the authors report fit residuals at the 2% level for k ≳ 1 h/Mpc and 5–10% on large scales (k ≲ 0.5 h/Mpc). Using the MCMC chains, the fitted bias parameters are propagated through an analytic expression (Eq. A1) and a Fisher forecast to predict non-linear BAO peak shifts: Δα∥ = −0.20 ± 0.09% and Δα⊥ = −0.31 ± 0.11% at z = 2.0 (Eq. 19), corresponding to Δαiso = −0.28 ± 0.09% and Δαap = 0.11 ± 0.07%, with analogous numbers at z = 2.6 and for the Lyα–quasar cross-correlation using eBOSS quasar biases and analytic/simulation-based bias relations. The paper presents these numbers as a theoretical error budget for DESI Lyα BAO measurements and as informative priors for full-shape analyses.","tokens_in":30961,"tokens_out":11591,"duration_ms":117778,"significance":"The paper is a timely and useful contribution if the quoted error budget can be made robust. Its strengths include: the first precision calibration of the one-loop Lyα-forest EFT with full posterior distributions for the bias set; an analytic extension of the galaxy BAO-shift formalism of Ref. [71] to the anisotropic Lyα forest and to the Lyα–quasar cross-correlation (Appendix A); a transparent, quantified disagreement with the percent-level simulation-based shift of Ref. [72]; and a commendably explicit statement of its own limitations (possible phenomenological status of the model at kmax = 2, resolution differences between the two boxes, and the mismatched effective redshift of the eBOSS quasar biases). The concern about circularity raised in the stress test does not, in my reading, land: the BAO shift is never fitted to the simulations; it is propagated from power-spectrum-calibrated parameters through an independent analytic formula. However, the central deliverable is an error budget, and the omission of a truncation-error term, together with the ambiguity over which chains (minimal vs. stochastic) produce Eq.","major_comments":[{"comment":"The quoted error bars in Eqs. (19)–(22) propagate only the MCMC parameter covariance and do not include a truncation-error term, although the one-loop expansion is not manifestly controlled at the fitting scale kmax = 2 h/Mpc: Fig. 2 shows the one-loop-to-tree ratio crossing unity near k ≈ 1.5 h/Mpc at z = 2, and Sec. 6.1 explicitly leaves open the possibility that the model is only phenomenological there. This is directly relevant to the error budget: Fig. 7 shows that the radial shift changes sign as kmax is increased from 0.5 to 2 h/Mpc, a variation far larger than the quoted ±0.09% uncertainty. I request that a systematic truncation error (e.g., from varying kmax, or from a two-loop estimate) be added to Eq. (19), or that the quoted budget be restricted to scales on which the loop expansion is controlled.","section":"Sec. 6.2, Eq. (19), Fig. 7"},{"comment":"The abstract states that the EFT model \"fits the data with an accuracy of below 2 percent up to a wavenumber of k = 2 h/Mpc.\" The body (Sec. 6.1 and Fig. 1) instead reports residuals of 5–10% on large scales (k ≲ 0.5 h/Mpc), with sub-2% accuracy only for k ≳ 1 h/Mpc. The body is honest about this scale dependence, but the abstract overstates the accuracy and should be rephrased to describe the scale-dependent residuals, e.g., \"below 2% on small scales (k > 1 h/Mpc) and 5–10% on large scales.\"","section":"Abstract; Sec. 6.1, Fig. 1"},{"comment":"It is not specified which MCMC chains were used to produce the baseline BAO shifts in Eqs. (19)–(22): the minimal model or the model with stochastic terms (\"+st.\"). This matters because Table II shows the stochastic terms are detected at ≫5σ at z = 2.0–3.0, and the minimal versus +st. best fits differ substantially in exactly the quadratic biases entering Eq. (A1) (e.g., at z = 2, bη2 = −0.35 versus −1.18 and bG2 = 0.05 versus −0.17). Since Eq. (A1) is linear in these parameters, the predicted BAO shift will differ by more than the quoted uncertainty between the two model choices. Please state explicitly which configuration was used for Eqs. (19)–(22) and report the shift from both configurations as a robustness check.","section":"Sec. 6.2, Table II, Fig. 3"},{"comment":"The claim that the BAO shift estimates \"will not depend on a particular point of view on the role of the one-loop corrections\" is too strong. The fitted quadratic and cubic biases used in Eq. (A1) are obtained from fits at kmax = 2 h/Mpc, where the one-loop-to-tree ratio exceeds unity and the stochastic/counterterm contributions are large; any mis-absorption of two-loop UV physics into those biases propagates directly into Eq. (19) through the analytic relation in Eq. (A1). A concrete, feasible test would be to compute the shift using only fits with kmax ≤ 1 h/Mpc, where the loop expansion is controlled, and to compare the result with Eq. (19); this would place the \"phenomenological versus controlled\" debate on a quantitative footing for the BAO-shift claim.","section":"Sec. 6.1, discussion after Fig. 2"}],"minor_comments":[{"comment":"There are typos in the resolution units: \"resolution down to 100 h kpc−1\" should read \"100 h−1 kpc,\" and similarly \"25 hkpc−1\" should read \"25 h−1 kpc\" (the correct form appears elsewhere in the paper).","section":"Sec. 2"},{"comment":"The symbol kNL is used in Eq. (7) but is not defined at first use; please define it explicitly as the nonlinear scale (e.g., the scale where the dimensionless power spectrum is of order unity).","section":"Sec. 3, Eq. (7)"},{"comment":"The eBOSS quasar biases used for the cross-correlation shifts are measured at z_eff = 1.48, while the Lyα parameters are at z = 2.0; the authors appropriately caution that the result \"should only be taken indicatively,\" but the headline cross-correlation numbers (Eqs. 27–30) could be misinterpreted. A direct statement of the implied systematic from the redshift mismatch, or a matched-redshift estimate, would remove ambiguity.","section":"Sec. 6.2, Eqs. (24)–(30)"},{"comment":"The likelihood uses a diagonal Gaussian covariance derived from the number of Fourier modes per bin. Given the very small quoted parameter uncertainties (e.g., b1 to 0.003 in Table II), a short test of sensitivity to off-diagonal or non-Gaussian covariance (e.g., a split-box or jackknife estimate) would increase confidence that the uncertainties are not underestimated.","section":"Sec. 4, Eq. (13)"},{"comment":"The figure legend distinguishes \"kmax = 2.0 h/Mpc,\" \"kmax = 3.0 h/Mpc,\" and \"kmax = 3.0 h/Mpc + st.,\" which implies the baseline kmax = 2 chains do not include stochastic terms. This is also relevant to the major comment about which chains produce Eq. (19); a consistent labeling or an explicit statement in Sec. 6.2 would help the reader.","section":"Fig. 3 and Sec. 6.2"},{"comment":"The simulation name is written inconsistently as \"ACCEL2\" and \"ACCL2\"; please unify.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The main technical risk to the central claim is the truncation error in the BAO-shift budget: Fig. 2 shows the loop-to-tree ratio crossing unity near k ≈ 1.5 h/Mpc and the stochastic terms are detected at high significance, so the fitted quadratic biases may absorb uncontrolled two-loop UV physics. The authors consciously defer a two-loop or field-level check, which is acceptable for a calibration paper provided the error budget includes a kmax-truncation systematic, a clarification of which chains produce Eq. (19), and a minimal-versus-stochastic comparison. These are all addressable with machinery already present in the paper. The cross-correlation section is appropriately hedged, and the reliance on same-group EFT papers (Refs. [49] and [71]) is standard practice in this field, not a novelty concern. The abstract's accuracy claim should be corrected before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a useful calibration paper. The authors measure the one-loop EFT bias parameters of the Lyman-alpha forest on the ACCEL2 hydro sims and derive analytic formulas for the nonlinear BAO shift in the forest auto-correlation and the forest-quasar cross-correlation. The central result—a shift of about -0.2 to -0.3 percent in the radial and transverse BAO distances at z=2—is plausible and roughly consistent with their own kmax and box-size checks. The paper deserves a serious referee, but the abstract oversells the fit, and the quoted shift errors leave out truncation uncertainty.\n\nWhat is genuinely new: the first precision fits of the one-loop EFT model to ACCEL2, across five snapshots with two box sizes, and the first analytic BAO-shift expressions for the forest and cross-correlation (Appendix A). The fitting procedure is careful: MCMC with analytic marginalization of the linear parameters, consistency between the small and large boxes, and agreement with earlier b1 and beta measurements from the same simulations. The bO-b1 relations also match the Sherwood results, which adds confidence. The informative priors for the EFT parameters will be directly useful for full-shape DESI analyses.\n\nThe soft spots are real but not fatal. First, the abstract says the fit is accurate to below 2 percent up to k=2, but the residuals shown in Fig. 1 are 5-10 percent on large scales; the body is honest about this, so the abstract should be corrected. Second, the model is pushed to kmax=2 where Fig. 2 shows the one-loop to tree-level ratio crossing unity near k=1.5 at z=2. The authors acknowledge this and offer an untested accidental-suppression argument, but they do not quantify the associated systematic. The BAO shift errors quoted in Eqs. (19)-(22) are MCMC parameter errors only; the kmax dependence in Fig. 7 shows the radial shift changing sign as the fit range increases from 0.5 to 2, which is a systematic larger than the quoted error bars. They caution the reader but do not fold this into the final budget. Third, the cross-correlation shift relies on quasar biases from eBOSS at zeff=1.48, not the forest redshift, and on analytic bias relations; the authors flag this, but the numbers should be treated as indicative. Finally, no code or data are released, which makes the fits hard to reproduce.\n\nWho this is for: anyone working on Lyman-alpha BAO or full-shape analyses in DESI. It deserves peer review, with the expectation that the abstract be fixed, a truncation error term be added to the shift budget, and the chains or code be made available. The central conclusion, that the nonlinear BAO shift is a few tenths of a percent, is probably right.","headline":"Useful calibration paper with an inflated abstract and a shift error budget that omits truncation sensitivity; the central result is plausible and deserves peer review after revisions.","tokens_in":31557,"tokens_out":4212,"would_cite":true,"duration_ms":41626,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["98.80.-k"],"model":"deepseek-v4-flash","headline":"One-loop EFT puts the Lyman-alpha BAO shift near 0.3 percent.","keywords":["Lyman-alpha forest","effective field theory of large-scale structure","baryon acoustic oscillations","BAO shift","flux power spectrum","hydrodynamic simulations","bias parameters","DESI"],"falsifier":"Compute a two-loop Lyman-alpha flux power spectrum at the ACCEL2 cosmology, or perform a field-level comparison on the same simulation boxes, and refit the one-loop model: if the best-fit bias parameters move by more than their quoted uncertainties, the derived BAO-shift budget is not reliable. A second check is to fit the same model to an independent hydrodynamic simulation with different intergalactic-medium thermal physics and see whether the 0.1–0.3 percent shift values persist.","tokens_in":30421,"feed_emoji":"🔭","tokens_out":10558,"duration_ms":98160,"temperature":0.7,"pith_summary":"The paper's central claim is that the one-loop effective field theory of large-scale structure (EFT), calibrated on the ACCEL2 hydrodynamic simulations, describes the three-dimensional Lyman-$\\alpha$ flux power spectrum accurately enough to serve as the theoretical backbone for baryon acoustic oscillation (BAO) measurements: below 2 percent on small scales up to $k=2\\,h\\,\\mathrm{Mpc}^{-1}$, and at the 5–10 percent level on large scales. From the fitted bias parameters, the paper derives an analytic prediction for how nonlinear clustering shifts the apparent BAO scale. At $z=2$, the radial and transverse dilation parameters shift by $\\Delta\\alpha_\\parallel=-0.20\\pm0.09\\%$ and $\\Delta\\alpha_\\perp=-0.31\\pm0.11\\%$, corresponding to $\\Delta\\alpha_{\\rm iso}=-0.28\\pm0.09\\%$ and $\\Delta\\alpha_{\\rm ap}=0.11\\pm0.07\\%$. If these numbers hold, current and future Lyman-$\\alpha$ BAO analyses can attach a small, quantifiable theory error budget instead of relying on unquantified systematics. The paper also estimates the shift in the Lyman-$\\alpha$–quasar cross-correlation at roughly $-0.1\\%$, with similar values for two different quasar bias models.","feed_headline":"One-loop EFT puts Lyman-alpha BAO shift near 0.3 percent","feed_subtitle":"The same model matches simulated forest power to 2 percent, giving DESI a concrete theory error budget.","key_machinery":"The load-bearing object is the one-loop EFT flux power spectrum $P^{\\rm th}(k,\\mu)=P_{\\rm tree}+P_{\\rm 1\\text{-}loop}+P_{\\rm ct}+P_{\\rm st}$, with redshift-space kernels from the rotationally invariant bias expansion and IR resummation of the BAO wiggle. The BAO shift is carried by the mode-coupling integral $P^{(22)}$, whose dominant piece is $\\langle\\Psi\\delta\\rangle\\xi'(x)$: long-wavelength displacements contract the BAO sphere around overdensities where the forest signal is suppressed, producing an analytic shift proportional to $\\sigma_d^2\\,k\\,P_w'(k)$ with a bias-dependent prefactor. A Fisher formalism converts this shift into predicted offsets in $\\alpha_\\parallel$ and $\\alpha_\\perp$ for a survey of DESI-like volume.","core_discovery":"The paper establishes that a one-loop EFT model for the Lyman-alpha forest—built from the bias expansion invariant under rotations around the line of sight, with counterterms, stochastic terms, and infrared resummation—fits the ACCEL2 simulated flux power spectrum to sub-2% accuracy on small scales and about 10% on large scales across five snapshots at redshifts $z=2$ to $4$. Using the same fitted parameters, the paper derives a closed-form expression for the nonlinear BAO shift and finds that nonlinearities displace the BAO scale by only a few tenths of a percent, with the sign and size set by the quadratic bias coefficients. This result does not support earlier simulation-based claims of a percent-level redshift-space BAO shift, and it supplies both informative priors and a concrete error budget for full-shape and compressed analyses of the forest.","pith_inferences":["If the small BAO shift relies on an accidental suppression of tree-level terms, a two-loop computation or field-level fit could break the cancellation and enlarge the shift beyond the quoted budget.","The cross-correlation estimates use quasar bias parameters measured at an effective redshift near $z\\simeq1.48$, while the DESI forest measurement sits near $z\\simeq2.33$; a bias measurement at the matching redshift could move the cross-correlation shift by an amount comparable to its size.","The quoted uncertainties on the shifts include only EFT parameter posteriors; they omit the spread across quasar bias models and across choices of $k_{\\max}$, so the true cross-correlation error budget may be larger.","A clean test would be to predict the one-dimensional flux power spectrum or the flux bispectrum from the same bias parameters; either observable would exercise the model in a regime that the BAO-shift calculation does not directly use."],"forward_implications":["DESI Lyman-alpha BAO analyses can add a theory error term of order 0.1–0.3 percent to their covariance, comparable to or below the current statistical precision.","The measured bias parameters provide informative priors for full-shape analyses of the three-dimensional Lyman-alpha power spectrum.","The isotropic shift of about $-0.3\\%$ is far smaller than the percent-level shift suggested by some simulation-based studies, so the BAO scale in the forest is largely robust to nonlinear modeling.","The Lyman-alpha–quasar cross-correlation shift is only about $-0.1\\%$, so cross-correlation BAO measurements need only a modest theory correction.","The same EFT framework carries over to future surveys and to field-level analyses, where the bias parameters can be checked against simulation output."],"supporting_citations":[{"why":"supplies the ACCEL2 hydrodynamic simulation power spectra used as the fit data.","marker":"[63]"},{"why":"provides the one-loop EFT power spectrum and redshift-space kernels for the Lyman-alpha forest.","marker":"[49]"},{"why":"supplies the analytic BAO-shift formalism and Fisher approach that the paper extends to the forest.","marker":"[71]"},{"why":"provides the phenomenological fitting function used as a comparison for the fitted bias parameters.","marker":"[65]"},{"why":"provides previous simulation-fit methodology and priors that the analysis builds on.","marker":"[66]"},{"why":"supplies the eBOSS quasar bias parameters used for the cross-correlation shift estimate.","marker":"[75]"},{"why":"provides simulation-based halo bias relations used as an alternative for the quasar nonlinear biases.","marker":"[76]"},{"why":"provides the quasar linear bias redshift relation used to evaluate the cross-correlation shift as a function of redshift.","marker":"[110]"},{"why":"defines the DESI Lyman-alpha BAO measurement that motivates the theory error budget.","marker":"[38]"}],"fun_headline_variants":["One-loop EFT: Lyman-alpha BAO shifts stay under 0.3%","Simulations peg Lyman-alpha BAO shift at 0.3%","EFT fits forest power to 2%, BAO shift 0.3%","Lyman-alpha BAO bias tiny in one-loop EFT","Precision EFT for Lyman-alpha: BAO shift ~0.3%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The one-loop EFT is assumed to remain controlled up to $k_{\\max}=2\\,h\\,\\mathrm{Mpc}^{-1}$, even though the loop-to-tree ratio crosses unity near $k\\sim1.5\\,h\\,\\mathrm{Mpc}^{-1}$, so the fitted bias parameters and the BAO shift built from them could absorb uncontrolled higher-order corrections.","fun_headline_variants_meta":{"raw":{"variants":["One-loop EFT: Lyman-alpha BAO shifts stay under 0.3%","Simulations peg Lyman-alpha BAO shift at 0.3%","EFT fits forest power to 2%, BAO shift 0.3%","Lyman-alpha BAO bias tiny in one-loop EFT","Precision EFT for Lyman-alpha: BAO shift ~0.3%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000281,"raw_usage":{"total_tokens":1723,"prompt_tokens":1061,"completion_tokens":662,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":677,"completion_tokens_details":{"reasoning_tokens":558}},"tokens_in":677,"tokens_out":662,"duration_ms":6443,"temperature":1.0,"reasoning_tokens":558,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T19:21:06.778527+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute a two-loop Lyman-alpha flux power spectrum at the ACCEL2 cosmology, or perform a field-level comparison on the same simulation boxes, and refit the one-loop model: if the best-fit bias parameters move by more than their quoted uncertainties, the derived BAO-shift budget is not reliable. A second check is to fit the same model to an independent hydrodynamic simulation with different intergalactic-medium thermal physics and see whether the 0.1–0.3 percent shift values persist.","supporting_citations":[],"review_version":1}