{"id":"eae0b9fe-4ebc-470c-bebf-974bb590306c","arxiv_id":"2608.07962","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":5,"one_line_summary":"A two-level zero-point calibration for LAMOST MRS radial velocities, anchored to Gaia and APOGEE, cuts high-S/N scatter against APOGEE from about 1.1 to 0.5 km/s and ships an 11-million-spectrum catalogue.","lead":"This paper builds a two-step correction for the LAMOST medium-resolution survey's radial velocities, using Gaia and APOGEE measurements as external references, and releases a catalogue of corrected velocities for over 11 million spectra. It matters because the correction roughly halves the scatter in high-quality stellar velocity measurements, which improves searches for binary stars, star cluster studies, and Milky Way kinematic maps.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline ~2x precision gain is measured against APOGEE DR17, which is in the calibration; the independent DR19 test gives 0.64 km/s rather than 0.52 km/s, so the external gain is ~1.7x unless a hold-out test shows DR17 remains at 0.52.","rationale":"The reader's ACCEPT is largely sound: the method is clear, code and catalogue are public, repeat observations improve, and DR19 is an independent check. The main soft spot is the presentation of the in-sample DR17 number as the headline precision. The reader flagged the in-sample nature but still accepted; this stress-test sharpens it: the DR19 point estimate (0.64 km/s) is 23% larger than DR17 (0.52 km/s), so the true out-of-sample gain is closer to 1.7x than 2x. The difference could be overfitting or pipeline differences; a hold-out DR17 test would settle which. The paper already reports DR19 and discusses the larger scatter, so this is a presentation and claim-calibration issue, not a soundness failure. Hence CONDITIONAL rather than REJECT or UNCHANGED: require the abstract and conclusion to state the externally validated improvement (~1.7x from DR19) or to clearly qualify the ~2x as measured against the calibration reference.","tokens_in":13588,"tokens_out":11196,"duration_ms":128710,"concrete_test":"Hold out a random 20% of APOGEE DR17 matches from the calibration (by unit), recompute both correction levels without them, and measure the S/N=50-60 residual scatter for the held-out stars. If the held-out scatter is about 0.64 km/s, the DR17 0.52 km/s is partly overfit and the external gain is ~1.7x; if it stays near 0.52 km/s, the larger DR19 scatter is a pipeline difference and the 2x headline stands.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's strongest quantitative claim (abstract; Section 5.1) is that high-S/N scatter relative to APOGEE DR17 drops from ~1.1 to ~0.52 km/s, 'implying a ~2x improvement in RV precision.' But DR17 is one of the two references used to estimate the zero points (Sections 3.2-3.3), so this is an in-sample metric. The unit-level weighted-median estimator can reduce scatter on the calibration stars by construction, particularly for the ~117k spectrograph-exposure units and 55k fiber-time units with only 20-17 reference contributions. The independent APOGEE DR19 validation (Section 5.3) gives 0.64 km/s residual scatter at S/N=50-60, an improvement factor of 1.07/0.64 ~ 1.7, not 2.1. If the DR17 number is quoted as the achieved precision, the catalogue's true external precision may be overstated by about 20%. This does not invalidate the method, but the central '~2x' claim needs to be tied to the out-of-sample DR19 result or explicitly labeled as relative to the calibration reference.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents an empirical two-level zero-point calibration for LAMOST DR12 v1.1 MRS blue-arm radial velocities. The first level subtracts weighted-median residuals against Gaia DR3 magnitude-colour-corrected and APOGEE DR17 RVs within (lmjm, planid, spid) spectrograph-exposure units; the second level does the same within (spid, fiberid, time_tag) units after the first-level correction, with grouped fallbacks for units with too few references. The authors validate the method on same-night and cross-night repeats, on residuals against both external references, on an independently derived blue/red-arm cross-correlation pipeline, and on APOGEE DR19. They release a value-added catalogue of 11,129,477 corrected blue-arm spectra and public calibration code.","tokens_in":13844,"tokens_out":9895,"duration_ms":113100,"significance":"If the reported gains hold, the paper materially improves the scientific usability of LAMOST MRS velocities: high-S/N scatter against the calibration reference drops from about 1.1 to 0.52 km/s, cross-night repeatability improves from 0.60 to 0.48 km/s, and the public catalogue and reproducible pipeline lower the barrier for time-domain and kinematic studies. The method is internally consistent, with a clean separation between calibration and application samples, a weighted-median estimator that is robust to binaries and variables, and an explicit fallback for sparse units. The strongest independent evidence is the cross-night repeat test, which does not rely on the external references, together with the transfer of the framework to the cross-correlation pipeline. The main weakness is that the headline precision gain is measured against a calibration reference, so the out-of-sample DR19 result should be the basis for the catalogue's claimed external precision.","major_comments":[{"comment":"The headline claim of a roughly 2x improvement in RV precision is based on the APOGEE DR17 comparison (1.09 -> 0.52 km/s in the S/N=50-60 bin), but DR17 is one of the two references used to derive the zero points in Sections 3.2 and 3.3. Scatter on the calibration sample can be reduced by construction, especially for the 117,428 spectrograph-exposure and 55,451 fiber-time units with only 20 and 17 reference contributions. The out-of-sample DR19 test in Section 5.3 gives 1.07 -> 0.64 km/s, i.e., an improvement of about 1.7x, not 2.1x. The abstract, Section 5.1, and the conclusions should lead with the DR19-based gain or explicitly state that the 0.52 km/s figure is relative to the calibration reference.","section":"Abstract; §5.1; §7"},{"comment":"Equation (3) assigns empirical uncertainties from the scatter of corrected LAMOST RVs relative to APOGEE DR17, the same reference that defines the correction. Because the independent DR19 comparison shows a larger scatter (0.64 vs 0.52 km/s at high S/N), the rv_lasp2_err column in the released catalogue likely understates the external precision by roughly 20%. The uncertainty calibration should be repeated on the DR19 sample, or a reference-frame systematic term should be added and documented in the catalogue description.","section":"§5.4, Eq. (3)"},{"comment":"The DR19 test is a useful check that the correction is not a pure re-fit of the DR17 residuals, but the text overstates its independence. DR19 is an updated APOGEE/SDSS reduction and therefore shares the APOGEE instrument and reference frame with DR17; moreover, the Gaia DR3 corrected RVs used in Section 2.2 are themselves tied to the APOGEE system. The sentence claiming the DR19 sample is 'not overlapping with the APOGEE DR17 set' should be clarified or removed, and the paper should state that the absolute zero-point frame is not independently verified. A comparison to a genuinely independent reference would be needed to test for common-mode reference systematics.","section":"§5.3"}],"minor_comments":[{"comment":"The calibration sample is described as excluding likely variables, binaries, and other problematic targets, but the actual exclusion criteria and masks are not specified; please state them or explicitly defer to the sigma-clipping procedure.","section":"§2.1/§3.2"},{"comment":"The justification for not deduplicating the Gaia/APOGEE overlap is plausible, but a one-line sensitivity test without the overlapping residuals would make the claimed ~17% influence quantitative.","section":"§3.2"},{"comment":"The key tuning parameters (0.5 km/s LASP error floor, n>=20 and n>=17 thresholds, time-tag segmentation, and clipping parameters) are not subjected to a robustness test; a short sensitivity run would show whether the DR19 0.64 km/s result is stable under reasonable variations.","section":"§3"},{"comment":"The external-scatter comparisons for the cross-correlation pipeline are not independent of the reference set because the two-level correction is re-derived using the same Gaia/DR17 residuals; please state that only the cross-night repeat test is fully reference-free.","section":"§5.2"},{"comment":"The reference 'Collaboration et al. 2016' should be formatted as 'Gaia Collaboration et al. 2016'.","section":"§2.2"}],"recommendation":"major_revision","confidential_remarks":"To the editor: This is a useful and well-executed calibration paper with public code and a public catalogue, and the core method is defensible. My recommendation of major revision is driven by the in-sample framing of the headline precision gain and the corresponding uncertainty calibration on the DR17 reference; both are fixable by foregrounding the DR19 result and adjusting Section 5.4. I do not see grounds for rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a genuinely useful calibration paper, and the strongest part is what it ships: a clear two-level zero-point scheme, a public pipeline, and an 11-million-spectrum catalogue. If LAMOST MRS multi-epoch kinematics are your science, this is the reference to beat.\n\nThe new bits are real: the explicit fiber–time correction layer beyond the exposure-level correction in Zhang et al. (2021), the extension of the hierarchical philosophy from low-res to medium-res spectra, and the scale of the value-added catalogue. The method is described with enough precision to reproduce, and the authors are careful to separate calibration from application samples, use clipped weighted medians rather than means, and provide a fallback for sparse units. They also go out of their way to validate: internal repeats split by time baseline, an independent cross-correlation pipeline on both arms, and an APOGEE DR19 comparison that was not used in the fit. The code and catalogue are public. That is real credit.\n\nThe main soft spot is the abstract's '~2x improvement' claim. The 1.09 to 0.52 km/s scatter reduction is measured against APOGEE DR17, which is one of the two references used to estimate the zero points. That is in-sample, and the weighted-median unit-level fit can reduce scatter on the calibration stars by construction. The independent DR19 test gives 0.64 km/s at the same S/N, an improvement factor of about 1.7, not 2.1. The paper does acknowledge the larger DR19 scatter in Section 5.3, so it is not hiding anything, but the headline number should be tied to DR19 or explicitly labeled as the calibration-reference comparison. This is a wording and emphasis problem, not a flaw in the method.\n\nTwo smaller caveats: the zero points are only as good as the Gaia/APOGEE reference frame, and the paper does not independently verify that frame's absolute zero point; and the fiber-time correction uses a coarse three-segment-per-year binning that leaves a residual ~0.2 km/s intermediate-timescale systematic. Both are stated openly, and neither undercuts the main deliverable.\n\nWho is this for? Anyone doing multi-epoch RV science with LAMOST MRS, or building calibration pipelines for fiber spectrographs. The paper deserves a serious referee. I would recommend publish after a revision that realigns the abstract and conclusion with the out-of-sample validation.","headline":"A solid, reusable calibration product for LAMOST MRS; the headline 2x precision gain is partly in-sample, but the independent DR19 check and public data make this well worth publishing.","tokens_in":14382,"tokens_out":2039,"would_cite":true,"duration_ms":20962,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"LAMOST MRS radial velocities can be made roughly twice as precise by subtracting two nested zero-point layers, and the paper releases a corrected catalogue of 11,129,477 blue-arm spectra.","keywords":["radial velocities","zero-point calibration","LAMOST medium-resolution survey","Gaia DR3","APOGEE DR17","value-added catalogue","multi-epoch kinematics","instrumental systematics"],"falsifier":"Compare the corrected LAMOST velocities for stars with S/N~50–60 to a high-precision reference catalogue built on an independent wavelength and technique, such as high-resolution optical echelle spectra calibrated without Gaia/APOGEE zero-point assumptions. If the reference-residual scatter does not drop to roughly 0.5 km/s, or if the residuals show magnitude- or colour-dependent offsets, then the two-level corrections would be absorbing shared reference systematics rather than instrument zero points.","tokens_in":13369,"feed_emoji":"🔭","tokens_out":7042,"duration_ms":72787,"temperature":0.7,"pith_summary":"This paper seeks to establish that the LAMOST medium-resolution radial velocities are limited less by random measurement noise than by a hierarchy of instrument-imprinted zero-point offsets that vary with each exposure, spectrograph, fiber, and observing season. The authors estimate and subtract those offsets in two nested levels — first per spectrograph–exposure, then per fiber–time — using externally anchored reference residuals. With the full correction, high-signal-to-noise scatter relative to APOGEE DR17 falls from about 1.1 km/s to 0.52 km/s, a roughly twofold precision gain, while the released catalogue supplies corrected velocities for 11,129,477 blue-arm spectra. A sympathetic reader would care because these velocities sit at the foundation of Galactic-kinematics, binary, and time-domain studies that need stable zero points across epochs.","feed_headline":"Two-level correction halves LAMOST radial-velocity errors","feed_subtitle":"Gaia and APOGEE anchors drop high-S/N scatter from ~1.1 to ~0.52 km/s across 11M spectra","key_machinery":"The load-bearing mechanism is a two-level zero-point calibration applied to residuals defined as RVZP = RV_LASP0 − RV_reference. At level one each unit (exposure, plan, spectrograph) collects residuals relative to both external catalogues, assigns inverse-variance weights with a 0.5 km/s floor on the LASP uncertainty, clips outliers, and takes the weighted median to define the spectrograph–exposure zero point; the same recipe is repeated on the first-level-corrected velocities inside (spectrograph, fiber, time-tag) units at level two. Low-evidence units inherit a grouped baseline (year×spectrograph for level one; time-part×25-fiber bundle for level two), so every spectrum receives a correction while the fit remains stable. What makes the hierarchy necessary is the observed structure: the zero-point maps vary coherently with season and spectrograph at the km/s level, then show residual fiber-scale patches and bundle discontinuities after the first level is removed.","core_discovery":"The paper's central claim is that the MRS radial-velocity zero point is a two-level empirical object rather than a single constant: a spectrograph–exposure term in (lmjm, planid, spid) units captures offsets shared within an exposure, and a fiber–time term in (spid, fiberid, time_tag) units captures the residual fiber-dependent, slowly varying structure. Both are estimated by weighted-median residuals against Gaia DR3 magnitude–colour-corrected RVs and APOGEE DR17 RVs, with outlier clipping and a grouped fallback for sparsely sampled units. After both corrections, internal cross-night single-observation precision at S/N 50–60 improves from 0.60 to 0.48 km/s, scatter against APOGEE DR17 falls from 1.09 to 0.52 km/s, and the improvement grows with the time between visits, reaching 0.63 km/s for baselines longer than a year versus 0.86 km/s before. An out-of-sample check against APOGEE DR19 reduces the same high-S/N scatter to 0.64 km/s, indicating the correction generalizes beyond its training references.","pith_inferences":["Because both external references are anchored to a common system, a future Gaia RV re-release with corrected zero points could require re-running the same hierarchy rather than reusing the published catalogue unchanged; this is a testable recalibration path, not a claim the paper makes.","The unresolved ~0.2 km/s gap between same-night and cross-night precision suggests that intermediate-timescale systematics (possibly wavelength-dependent residuals or the coarse three-segment-per-year time tagging) survive the two levels; a finer temporal grid or a wavelength-resolved layer would be a natural next experiment.","The same hierarchical recipe could be transplanted to other multi-fiber, multi-epoch surveys with access to high-precision external RVs, with the same two-level unit structure adapted to their observing geometry.","Because the correction is derived per measurement unit rather than per star, it should also apply to non-stellar or low-S/N spectra in the application sample after the same reference anchors are matched, though the gain there will be smaller because noise dominates."],"forward_implications":["Corrected LAMOST MRS RVs become roughly twice as precise for high-S/N spectra and are consistent across exposures, spectrographs, fibers, and epochs, enabling cleaner multi-epoch kinematics and time-domain variability studies.","The released catalogue supplies rv_lasp2 for 11,129,477 blue-arm spectra (8,158,271 single-exposure and 2,971,206 coadded), giving a uniform velocity baseline where none existed before.","Long-baseline repeated observations gain the most: the single-observation precision for visits separated by more than a year improves from about 0.86 to 0.63 km/s.","The correction transfers to an independent blue- and red-arm cross-correlation pipeline without re-tuning, and at S/N≥10 produces the same qualitative gains, implying the zero-point structure is instrumental rather than a quirk of one velocity pipeline.","An independent APOGEE DR19 validation shows the corrected scatter remains the smallest in every S/N bin, confirming the gain is not just a re-fit to the calibration references."],"supporting_citations":[{"why":"Supplies the Gaia DR3 magnitude- and colour-corrected radial velocities that define the primary external reference frame.","marker":"Katz, D. et al. 2023"},{"why":"Documents Gaia DR3 RV zero-point systematics that justify using the corrected Gaia velocities rather than raw DR3 values.","marker":"Blomme, R. et al. 2023"},{"why":"Defines the APOGEE survey whose DR17 RVs serve as the second, more precise external reference in each calibration unit.","marker":"Majewski et al. 2017"},{"why":"Establishes the early LAMOST MRS RV baseline and template-matching products that motivate a zero-point correction hierarchy.","marker":"Wang et al. 2019"},{"why":"Introduces spectrograph-exposure-level time-dependent RV zero-point corrections for MRS, which this work extends with a fiber-time level.","marker":"Zhang et al. 2021"},{"why":"Provides the hierarchical calibration philosophy for LAMOST low-resolution spectra that the paper adapts to MRS.","marker":"Zhang, J., Yuan, H., & Tian, Z. 2026"},{"why":"Supplies the APOGEE DR19 catalogue used as an independent out-of-sample validation set not involved in the calibration.","marker":"Pallathadka et al. 2025"}],"fun_headline_variants":["Two-level RV zero-point calibration halves LAMOST errors","Gaia-APOGEE anchored correction doubles LAMOST RV precision","11M LAMOST spectra get two-level RV zero-point correction","LAMOST MRS RV scatter drops from 1.1 to 0.52 km/s","Two-level calibration: LAMOST RV precision improved 2x"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The correction assumes the external references themselves are zero-point-clean: if Gaia's magnitude–colour-corrected RVs and APOGEE DR17 share a common zero-point error or a magnitude/colour/time-dependent residual, that error is copied into every LAMOST velocity.","fun_headline_variants_meta":{"raw":{"variants":["Two-level RV zero-point calibration halves LAMOST errors","Gaia-APOGEE anchored correction doubles LAMOST RV precision","11M LAMOST spectra get two-level RV zero-point correction","LAMOST MRS RV scatter drops from 1.1 to 0.52 km/s","Two-level calibration: LAMOST RV precision improved 2x"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000441,"raw_usage":{"total_tokens":2301,"prompt_tokens":1078,"completion_tokens":1223,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":694,"completion_tokens_details":{"reasoning_tokens":1126}},"tokens_in":694,"tokens_out":1223,"duration_ms":11848,"temperature":1.0,"reasoning_tokens":1126,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T00:36:31.582440+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compare the corrected LAMOST velocities for stars with S/N~50–60 to a high-precision reference catalogue built on an independent wavelength and technique, such as high-resolution optical echelle spectra calibrated without Gaia/APOGEE zero-point assumptions. If the reference-residual scatter does not drop to roughly 0.5 km/s, or if the residuals show magnitude- or colour-dependent offsets, then the two-level corrections would be absorbing shared reference systematics rather than instrument zero points.","supporting_citations":[{"cited_title":"A General Framework for Radial Velocity Calibration in Low-Resolution Spectroscopic Surveys: Correcting Wavelength-Dependent and Global Systematics with Application to LAMOST DR9","cited_arxiv_id":"2604.19119","evidence_quote":"Provides the hierarchical calibration philosophy for LAMOST low-resolution spectra that the paper adapts to MRS."}],"review_version":1}