{"id":"39473df9-9026-4997-82e6-74e5718b5701","arxiv_id":"1908.04773","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A new calibrated catalog of 1.6 million radial velocities from LAMOST medium-resolution spectra, with precision from 0.91 to 1.36 km/s depending on signal-to-noise.","lead":"This paper measures radial velocities for 1.6 million LAMOST medium-resolution spectra and releases a validated catalog. The catalog gives astronomers a large, uniformly processed set of stellar motions for studying the Milky Way, with typical precision near 1 km/s.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline accuracy 0.0227/0.0277 km/s is a residual on the same HY18 standards used for zero-point calibration, not an independent accuracy; independent catalogs show offsets of 0.17-0.46 km/s.","rationale":"The reader's verdict is CONDITIONAL, and I agree that conditions are needed. However, the reader's stated weakest assumption is that binary stars contaminate the repeated-observation precision estimate in Section 4.4. That is a real concern, but it affects the secondary 'precision' numbers and would likely inflate the quoted scatter, making the stated precision conservative. The more load-bearing concern is the circular accuracy claim: the headline '0.0227 km/s accuracy' is computed on the same sample used to calibrate the zero point, so it is a calibration residual rather than an independent validation. The paper does contain independent comparisons, and those show offsets an order of magnitude larger, which is the number that should appear in the abstract. Because the paper still provides a useful catalog and the independent comparisons allow a reader to recover the true accuracy, a conditional accept remains appropriate; the condition should be that the accuracy claim be reframed using held-out or independent references. I therefore keep the verdict unchanged rather than moving it, and I mark agreement as partial because the reader's formal weakest-assumption differs from mine, though the rationale already mentions circularity.","tokens_in":13378,"tokens_out":4228,"duration_ms":41112,"concrete_test":"Run a leave-half-out validation: split the 983 HY18 standards (or their 7,820 spectra) randomly by spectrograph and exposure, derive the per-spectrograph/lamp/band RVZP offsets from one half, and compute the mean RV difference on the held-out half. If the held-out mean offset is within ~0.03 km/s, the calibration residual is representative and the headline accuracy survives; if it rises to ~0.2-0.5 km/s, the abstract's accuracy claim is overstated and should be replaced by offsets from independent catalogs (Gaia RV-STD, APOGEE, RAVE, Gaia DR2). As a secondary check, recompute the mean of the HY18 residuals to reconcile the abstract value 0.0227 with Table 2's 0.0277.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central accuracy claim in the abstract is circular. In Section 4.1, the zero-point corrections are derived per spectrograph, lamp, and band from 7,820 spectra of 983 HY18 standard stars. In Section 4.3, Table 2 then reports the HY18 comparison using exactly the same 983 stars and 7,820 spectra. After subtracting the fitted zero-point offsets, the mean residual is expected to be near zero by construction, so the quoted 0.0277 km/s (abstract: 0.0227 km/s, summary: 0.03 km/s) measures internal calibration consistency, not absolute accuracy. The genuinely independent comparisons in Table 2 — Gaia RV-STD, APOGEE DR14, RAVE, and Gaia DR2 — show mean offsets of 0.17-0.46 km/s and RMS scatters of 1.7-4.1 km/s, suggesting the true systematic accuracy is roughly an order of magnitude worse than the headline. The paper's own Figure 7 confirms that the calibration residuals are centered at ~0.02 km/s after correction, which is expected. The load-bearing problem is that a reader of the abstract will take 0.0227 km/s as the external accuracy of the catalog, whereas the data in the paper support a more modest claim such as 'the zero-point is consistent with HY18 standards at the ~0.03 km/s level on the calibration sample, with independent checks showing offsets of 0.2-0.5 km/s.' This does not invalidate the catalog, but it requires a corrected statement of the central claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper describes the measurement of radial velocities for 1,594,956 LAMOST-II medium-resolution (R≈7500) spectra using template matching against Kurucz model spectra. A zero-point calibration is derived per spectrograph, lamp, and waveband from 7,820 spectra of 983 HY18 RV standard stars, and the calibrated RVs are compared with HY18, Gaia RV-STD, APOGEE DR14, RAVE DR5, and Gaia RVS. The authors report an accuracy of 0.0227 km/s (abstract) or 0.0277 km/s (Table 2) with respect to HY18 standards, and an 'intrinsic precision' of 1.36, 1.08, and 0.91 km/s at S/N 10, 20, and 50 based on repeated observations. A public catalog is provided.","tokens_in":13716,"tokens_out":8552,"duration_ms":75172,"significance":"The work is potentially valuable: it delivers a large, homogeneous RV catalog from LAMOST MRS, with per-spectrograph zero-point corrections and validation against several independent surveys. The external comparisons with APOGEE, RAVE, and Gaia RVS are a genuine strength and give a realistic picture of the systematic floor. However, the headline accuracy number in the abstract is computed on the calibration sample itself and is therefore not an independent accuracy measurement; the independent comparisons show offsets of order 0.2-0.4 km/s. With corrected claims, this would be a useful methods-and-catalog paper for the LAMOST community.","major_comments":[{"comment":"The quoted accuracy of 0.0277 km/s (HY18 row of Table 2; abstract 0.0227 km/s) is derived from exactly the same 983 stars and 7,820 spectra used in Section 4.1 to fit the RVZP offsets per spectrograph, lamp, and waveband. After applying the fitted offsets, the mean residual is near zero by construction; this statistic measures internal calibration consistency, not absolute accuracy. The independent comparisons in Table 2 (Gaia RV-STD, APOGEE, RAVE, Gaia RVS) show mean offsets of -0.36 to +0.003 km/s (absolute values 0.22-0.36 km/s except RAVE) with RMS scatters of 1.7-4.1 km/s, which is the more honest statement of external accuracy. The abstract and summary should be revised to state that the zero point is consistent with HY18 at the ~0.03 km/s level on the calibration sample, while independent catalogs indicate systematic offsets at the ~0.2-0.4 km/s level. As written, a reader of the abstract will overestimate the absolute accuracy of the catalog by an order of magnitude.","section":"§4.1, §4.3, Table 2"},{"comment":"The precision values (1.36, 1.08, 0.91 km/s at S/N 10, 20, 50) are computed from the sample standard deviation of repeated observations of the same stars without any exclusion of known binaries, pulsating variables, or other RV-variable objects. The text says only that the estimate is 'based on the RV measurements of multiple observations in different epoch for the same stars.' If the repeat-observation sample contains a non-negligible fraction of binaries, the reported scatter is an upper limit on single-epoch measurement precision, not the intrinsic precision claimed. The authors should either remove or flag RV-variable stars (e.g., by iterative sigma clipping or by using a subsample of RV standards) or explicitly re-label the statistic as 'repeatability including astrophysical variability.'","section":"§4.4, Eq. (1), Fig. 13"},{"comment":"The accuracy numbers are internally inconsistent: the abstract gives 0.0227 km/s, Section 4.3 says 0.028 km/s for HY18 and 0.107 km/s for Gaia RV-STD, Table 2 lists 0.0277 and -0.2797 km/s, and the summary states 0.03 and 0.28 km/s. In addition, Section 4.3 reports a HY18 cross-match of 1,106 stars/8,336 spectra and a Gaia RV-STD cross-match of 52 stars/326 spectra, while Table 2 uses 983/7,820 and 46/261; the selection criteria for the table rows are not explained. These discrepancies must be reconciled before the paper can be accepted, because the central claim concerns the accuracy of the measurements.","section":"Abstract, §4.3, §5, Table 2"}],"minor_comments":[{"comment":"The abstract has typographical errors: 'km s/1' should be 'km s−1', and there are missing spaces in '1.36 km s−1,1.08 km s−1,0.91 km s−1'.","section":"Abstract"},{"comment":"'LAMOSR MRS spectra' should be 'LAMOST MRS spectra'.","section":"§4.1"},{"comment":"The captions contain 'Sc la p calibrated' and 'Sc lam calibrated'; both should be 'Sc lamp calibrated'.","section":"Figure 6 and Figure 7 captions"},{"comment":"'RVAE RVs are inconsistent' should be 'RAVE RVs are inconsistent'.","section":"§4.3"},{"comment":"The summary states the observations were collected 'till 31st Dec 2017', but Section 2.1 states the sample extends to 31st Dec 2018; please correct the date.","section":"§5"},{"comment":"The provenance of the quoted precision values 1.08 and 0.91 km/s is unclear; please specify whether these are from the blue band, the red band, or the mean of the two, and define the S/N binning used to read the values from Figure 13.","section":"§4.4, Fig. 13"}],"recommendation":"major_revision","confidential_remarks":"The catalog and the external comparisons are worth publishing, but the paper currently overstates its accuracy and contains several unreconciled numbers. I would ask that the authors revise the abstract and summary to separate calibration-sample consistency from independent accuracy, and address the binary-contamination issue in the precision estimate. If those changes are made, the paper should be acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a legitimate data paper — the first RV catalog for LAMOST-II medium-resolution spectra — and the catalog will be useful. The method is standard template matching and the zero-point calibration is conventional. The problem is the abstract's accuracy claim: the 0.0227 km/s (or 0.0277 in Table 2) is the residual on the same HY18 standards used to fit the zero-point offsets, so it measures internal consistency, not absolute accuracy. The paper's own independent comparisons against APOGEE, RAVE, and Gaia DR2 show mean offsets of 0.17–0.46 km/s and scatter of 1.7–4.1 km/s. That is a real gap between the headline and what the data support.\n\nWhat is new: the catalog itself, 1.59 million MRS spectra with calibrated RVs, RV errors, template parameters, flags, plus public release. The precision estimates from repeat observations (1.36/1.08/0.91 km/s at S/N 10/20/50) are plausible for R~7500 spectra. The paper also usefully documents the Sc vs Th-Ar lamp zero-point offsets — a practical detail that will matter to users.\n\nSoft spots: the precision estimate does not exclude binaries or variables, so those numbers could be upper bounds on true single-epoch error. Minor, because the effect is probably small for most stars, but worth stating. The abstract/table inconsistency (0.0227 vs 0.0277, summary says 0.03) is sloppy. The circular accuracy claim is the main issue; it should be reframed as an internal calibration consistency check, with the independent offsets as the external accuracy. None of this invalidates the catalog.\n\nWho this is for: anyone using LAMOST MRS data, especially for Galactic archaeology or time-domain work. It deserves a serious referee. My recommendation: accept after the accuracy claim is corrected and the abstract made consistent.","headline":"Useful first RV catalog for LAMOST MRS, but the headline accuracy claim is circular and needs reframing.","tokens_in":14376,"tokens_out":4101,"would_cite":true,"duration_ms":33682,"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":"LAMOST medium-resolution spectra yield radial velocities accurate to ~0.03 km/s across 1.6 million spectra.","keywords":["radial velocities","LAMOST","medium-resolution spectroscopy","template cross-correlation","zero-point calibration","RV standard stars","signal-to-noise precision","spectroscopic surveys"],"falsifier":"Take the paper's repeat-observation sample, discard any star whose RV varies by more than about 3 km/s between epochs or that is flagged as a binary or variable in public catalogs, and recompute the per-S/N rms; if the 1.36/1.08/0.91 km/s values drop noticeably, the quoted precision is contaminated by astrophysical variability.","tokens_in":13186,"feed_emoji":"🔭","tokens_out":7483,"duration_ms":65007,"temperature":0.7,"pith_summary":"This paper reports a uniform radial-velocity (RV) measurement for the 1,594,956 medium-resolution spectra of the LAMOST-II survey. The authors show that a two-step template cross-correlation over 2,194 synthetic spectra, followed by a zero-point correction derived from 983 RV standard stars, removes lamp- and spectrograph-dependent offsets and brings the RV zero point into agreement with external standard-star catalogs at the level of about 0.03 km/s. They further estimate that a single spectrum delivers an intrinsic RV precision of 1.36 km/s at signal-to-noise 10, 1.08 km/s at 20, and 0.91 km/s at 50, improving with S/N and slightly better in the blue band than the red. If these numbers hold, the accompanying catalog gives Galactic-kinematics and time-domain studies a large, consistently calibrated RV sample from an R~7500 survey.","feed_headline":"1.6 million spectra get radial velocities good to ~0.03 km/s","feed_subtitle":"Calibrated with standard stars, LAMOST-II spectra reach 1.36 km/s precision at S/N 10 and a zero point near 0.03 km/s.","key_machinery":"The load-bearing mechanism is a two-step cross-correlation search: first match each continuum-normalized spectrum against all 2,194 synthetic templates shifted in coarse 40 km/s steps, then refine in 1 km/s steps and fit a Gaussian to ten points around the correlation peak. The per-band RVs (blue and red) are computed independently, and applying per-exposure, per-spectrograph, per-band zero-point offsets from standard-star observations removes lamp-dependent shifts. Precision is estimated from repeated epochs using the rms scatter of each star's RV measurements about its mean.","core_discovery":"The paper's central claim is that a coarse-to-fine template cross-correlation over a grid of 2,194 synthetic spectra, applied independently to the blue (4900–5400 Å) and red (6300–6800 Å) bands, produces radial velocities for LAMOST-II medium-resolution spectra (R≈7500) whose zero point, after per-exposure, per-spectrograph, per-band correction using 983 standard stars, agrees with the HY18 RV-standard catalog to 0.0227 km/s (0.0277 km/s in the detailed residual table). The same measurement, assessed from repeated observations of the same stars, reaches an intrinsic precision of 1.36 km/s at S/N 10, 1.08 km/s at S/N 20, and 0.91 km/s at S/N 50. The paper also establishes that these properties hold across a catalog of 1,594,956 spectra and that blue-band RVs are systematically more precise than red-band RVs.","pith_inferences":["A direct test the paper leaves open: recompute the repeat-observation precision after removing known binaries and variable stars; if the scatter shrinks, the quoted 1.36/1.08/0.91 km/s values are upper limits rather than pure measurement error.","Because the blue and red bands are measured independently and disagree at the 9.44 km/s level overall, a combined-band RV with flag-aware weighting would likely beat the single-band precision at low S/N; the paper stops at recommending the blue value.","The per-spectrograph zero points were derived under two arc-lamp regimes; a reader might expect a separate comparison of Sc-lamp-only and Th-Ar-only subsets to reveal whether the lamp transition itself contributes residual offsets, which the paper does not quantify separately."],"forward_implications":["The released 1,594,956-spectrum catalog provides a calibrated RV per spectrum for both blue and red bands, enabling kinematic studies of the Milky Way at R≈7500 resolution.","Zero-point agreement with RV standards at about 0.03 km/s means time-series users can combine epochs across the lamp change without introducing large velocity jumps, provided they apply the per-spectrograph offsets.","The S/N-dependent precision curve gives a ready-made error model: 1.36 km/s at S/N 10, 1.08 at 20, and 0.91 at 50, with blue-band measurements more precise than red-band ones.","The full-grid two-step matching is computationally feasible at survey scale, taking about ten days on a 15-PC Spark cluster for the roughly 3.2 million single-band spectra."],"supporting_citations":[{"why":"Supplies the 983 RV standard stars used for zero-point calibration and accuracy assessment.","marker":"Huang et al. 2018 (HY18)"},{"why":"Provides the 2,194 synthetic template spectra used in the cross-correlation matching.","marker":"Castelli & Kurucz 2004"},{"why":"Provides the Gaia RV-STD catalogue used as an independent accuracy reference set.","marker":"Soubiran et al. 2018"},{"why":"Provides APOGEE DR14 radial velocities used as a high-resolution reference set.","marker":"Holtzman et al. 2018"},{"why":"Provides RAVE DR5 radial velocities, a same-resolution reference set.","marker":"Kunder et al. 2017"},{"why":"Provides Gaia DR2 radial velocities used for the large-sample accuracy comparison.","marker":"Gaia Collaboration et al. 2018"},{"why":"Describes the LAMOST reduction and stellar parameter pipeline used to process the spectra.","marker":"Luo et al. 2015"},{"why":"Supplies the continuum-normalization method applied to observed and synthetic spectra before matching.","marker":"Lee et al. 2008"}],"fun_headline_variants":["LAMOST-II RVs: 1.6M spectra, 0.023 km/s zero point","Radial velocities from 1.6M LAMOST-II spectra calibrated","Precision RVs: 0.91 km/s at S/N 50 from 1.6M spectra","1.6M stellar RVs with 0.02 km/s zero point from LAMOST-II","LAMOST-II: 1.6M spectra, 0.0227 km/s zero-point RV accuracy"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The precision estimates assume that repeated observations of the same star differ only because of measurement noise, so stars that are actually binaries or pulsators would inflate the quoted scatter.","fun_headline_variants_meta":{"raw":{"variants":["LAMOST-II RVs: 1.6M spectra, 0.023 km/s zero point","Radial velocities from 1.6M LAMOST-II spectra calibrated","Precision RVs: 0.91 km/s at S/N 50 from 1.6M spectra","1.6M stellar RVs with 0.02 km/s zero point from LAMOST-II","LAMOST-II: 1.6M spectra, 0.0227 km/s zero-point RV accuracy"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000587,"raw_usage":{"total_tokens":2770,"prompt_tokens":971,"completion_tokens":1799,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":587,"completion_tokens_details":{"reasoning_tokens":1669}},"tokens_in":587,"tokens_out":1799,"duration_ms":12672,"temperature":1.0,"reasoning_tokens":1669,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:33:06.693347+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the paper's repeat-observation sample, discard any star whose RV varies by more than about 3 km/s between epochs or that is flagged as a binary or variable in public catalogs, and recompute the per-S/N rms; if the 1.36/1.08/0.91 km/s values drop noticeably, the quoted precision is contaminated by astrophysical variability.","supporting_citations":[],"review_version":1}