pith. sign in

arxiv: 2605.23136 · v1 · pith:MUAYZTBAnew · submitted 2026-05-22 · 🌌 astro-ph.IM

A Gaia-linked High-purity QSO Candidate Catalog in Selected Fields with Extinction-binned Calibration and Spectrum-informed Training

Zi-Huang Cao , Zhao-Xiang Qi , Juan-Juan Ren , Bo Zhang , Dongwei Fan , Shi-Long Liao , Yuzhou Wang , Yong-Heng Zhao
show 6 more authors
Yong Zhang Meng-Xin Wang Yihan Tao Gao-Yuan Zhang Yong Yu A-Li Luo
This is my paper

Pith reviewed 2026-05-25 03:32 UTC · model grok-4.3

classification 🌌 astro-ph.IM
keywords QSO candidatesGaia cataloghigh-purity selectionextinction calibrationspectrum teacher modelquasar follow-upphotometric classification
0
0 comments X

The pith

Spectrum-informed selector on Gaia sources reaches 0.9809 purity and 0.8869 completeness for QSO candidates, versus 0.4493 for the official Gaia probability at matched threshold.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper constructs a high-purity QSO candidate catalog for selected fields intended as input for fiber spectroscopy rather than an all-sky census. It combines Gaia astrometry and photometry with optical and infrared features, applies E(B-V)-binned threshold calibration, and uses spectra only inside a source-grouped teacher model during training. Evaluation on a frozen Gaia-linked benchmark, with validation and test sources held out of teacher fitting, shows the deployed selector meets its 0.98 validation purity target while recovering nearly twice the spectroscopic completeness of the official Gaia QSO probability. The released catalog supplies calibrated scores, field-layer flags, coverage metadata, and provenance for the core, application, and stress-test domains.

Core claim

At the recommended conservative operating point calibrated to a validation-set purity of 0.98, the P3 spectrum-informed catalog selector achieves a measured test-set purity of 0.9809 and a spectroscopic-label completeness of 0.8869 within the frozen Gaia-linked benchmark, whereas the Gaia official QSO probability yields a spectroscopic-label completeness of 0.4493 under the same threshold protocol.

What carries the argument

The P3 spectrum-informed catalog selector that trains on Gaia-linked sources via a source-grouped spectrum-teacher model but applies only astrometric, photometric, and catalog features at inference, together with E(B-V)-binned threshold calibration across layered field domains.

If this is right

  • The catalog supplies source identifiers, field assignments, input-coverage flags, calibrated scores, threshold flags, and validation metadata ready for fiber follow-up scheduling.
  • Performance metrics are reported separately for the four-field core domain, four application or stress-test fields, and the COSMOS extreme-deep case.
  • Relative to the earlier P2 teacher, P3 produces a modest mean completeness gain across seeds, most visible in higher-extinction and faint-source subsets, at a small cost in purity.
  • The released empirical selection-function product allows downstream users to apply the same thresholds and coverage cuts in the covered fields.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the held-out protocol continues to block leakage, the same teacher could be reused to generate catalogs in additional fields that share similar photometric coverage.
  • The purity-first design implies the catalog is most useful when telescope time is limited and the cost of observing non-QSOs is high.
  • In regions where the Gaia-linked parent sample is shallower than deeper photometric catalogs, the output is best treated as a prioritized target list rather than a statistically complete sample.

Load-bearing premise

Excluding downstream validation and test Gaia source IDs from teacher fitting and checkpoint selection, while using teacher probabilities only for training rows, prevents spectra from leaking into the final purity and completeness numbers.

What would settle it

An independent spectroscopic campaign on sources inside one of the application fields that counts how many catalog-selected objects are confirmed QSOs versus contaminants and how many known QSOs fall below the threshold.

Figures

Figures reproduced from arXiv: 2605.23136 by A-Li Luo, Bo Zhang, Dongwei Fan, Gao-Yuan Zhang, Juan-Juan Ren, Meng-Xin Wang, Shi-Long Liao, Yihan Tao, Yong-Heng Zhao, Yong Yu, Yong Zhang, Yuzhou Wang, Zhao-Xiang Qi, Zi-Huang Cao.

Figure 1
Figure 1. Figure 1: Catalog coverage by field and source product for the current field-layer design. Rows are grouped by the layer definitions in [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Catalog-construction workflow used to turn survey measurements into a deployable QSO-candidate score. The four blocks separate information by its role in the released catalog. Source inputs contain Gaia astrometry and photome￾try, optical/infrared matches, optional variability information, source identifiers, missingness indicators, and masks. Catalog representation converts those heterogeneous measurement… view at source ↗
Figure 3
Figure 3. Figure 3: Selected-field footprints of the high-purity P3 QSO candidate catalog in the four core domain-ladder fields. Points are plotted in Galactic coordinates and colored by the deployable P3 QSO score. The panels show the released target-list distribution after the conservative validation-calibrated threshold is applied; they are therefore descriptive catalog products rather than measurements of the intrinsic QS… view at source ↗
Figure 4
Figure 4. Figure 4: Parameter-space coverage of selected, spectroscopically confirmed QSOs in the four core domain-ladder fields for the P3 catalog student. This validation view uses only sources that are already confirmed as QSOs by the frozen spectroscopic labels, so redshift is a measured spectroscopic coordinate and is not assigned to new candidates. The four panels project the selected QSO locus into redshift–Gaia magnit… view at source ↗
Figure 5
Figure 5. Figure 5: Catalog-quantity footprint of threshold-selected candidates in the four core domain-ladder fields before new spec￾troscopy is obtained for the P3 catalog student. This candidate view uses only quantities available at target-selection time, so no redshift coordinate is plotted and neither spectroscopic nor photometric redshift is assigned to the candidates. The panels show Gaia color–magnitude space, Legacy… view at source ↗
Figure 6
Figure 6. Figure 6: Seed-to-seed robustness of the P3–P2 completeness difference at the conservative purity ≥ 0.98 operating point with E(B − V )-binned threshold calibration. Each blue bar is the mean completeness difference, P3 minus P2, over five downstream student seeds; black error bars show the seed-to-seed scatter. Values are plotted in percentage points, so 1% on the y-axis is an absolute completeness difference of 0.… view at source ↗
Figure 7
Figure 7. Figure 7: Bootstrap uncertainty on the P3–P2 completeness difference at the conservative purity ≥ 0.98 operating point. Each point is the bootstrap median P3–P2 completeness difference for one diagnostic slice, and the horizontal bar spans the 2.5–97.5 percentile interval from resampling the frozen test set. Values are shown in percentage points. Intervals entirely to the right of zero support a resolved positive P3… view at source ↗
Figure 8
Figure 8. Figure 8: is the main regional comparison against this external reference classifier. The bars compare fixed-purity selections, not fixed score thresholds: both Gaia and P3 are calibrated to the same high-purity operating point, and the ordinate shows how many spectroscopic QSOs are recovered under that constraint. This comparison is central to the catalog interpretation because Gaia provides an all-sky, highly cura… view at source ↗
Figure 9
Figure 9. Figure 9: External support channels for the 39 robust COSMOS candidates. The first bar gives the retained subset after the Extreme Deep diagnostic check, and the second bar counts candidates with at least one direct support channel among X-ray, radio, and spectroscopy. The remaining bars show the individual support-channel counts and the number of candidates with valid redshift measurements. Because support channels… view at source ↗
Figure 10
Figure 10. Figure 10: Best available redshift distribution for the COSMOS robust candidates with valid redshift information. The histogram includes 33 of the 39 robust candidates, and the dashed vertical line marks the median redshift, z = 1.729. The distribution describes the externally supported robust subset and is useful for follow-up planning, especially because it indicates the redshift range over which the priority list… view at source ↗
Figure 11
Figure 11. Figure 11: False-positive composition in sky-transfer diagnostics. Bars show the spectroscopic labels of false positives when models trained in one sky regime are evaluated in another; AC denotes the anti-center field and HL denotes the high-latitude field. The diagnostic isolates domain-transfer failures and should not be read as the final catalog contaminant mixture. Stellar contaminants dominate the transfer fail… view at source ↗
Figure 12
Figure 12. Figure 12: Held-out catalog-space overlap diagnostic for the frozen test set. The horizontal axis gives the local QSO-label fraction among the 50 nearest training-set neighbors in standardized inference-time catalog features. The shaded interval marks the mixed region between 0.20 and 0.80; values above 0.80 are QSO-like and values below 0.20 are non-QSO-like. Most spectroscopic QSOs and non-QSOs occupy opposite end… view at source ↗
Figure 13
Figure 13. Figure 13: Interpretation-domain envelope for the present catalog. The horizontal axis orders regimes by increasing distance from the frozen benchmark and by increasing selection-function mismatch. The vertical placement is schematic, not a measured performance axis; the parenthetical text under each regime names the most direct inference supported by the available evidence. Points near the upper left correspond to … view at source ↗
Figure 14
Figure 14. Figure 14: Auxiliary GALAXY and STAR performance at the validation-calibrated purity ge 098 operating point. The left panel gives measured test purity in percent, and the right panel gives spectroscopic-label completeness in percent. The dashed horizontal line marks 98% for visual reference. Gaia DR3 denotes the official Gaia class probability for the corresponding class, used only as an external reference baseline;… view at source ↗
Figure 15
Figure 15. Figure 15: Field-level spectroscopic-label completeness for the P3 auxiliary GALAXY and STAR students. Values are plotted in percent at the same E(B − V )-binned validation-calibrated operating point as [PITH_FULL_IMAGE:figures/full_fig_p043_15.png] view at source ↗
read the original abstract

We present an extinction-calibrated, Gaia-source-level QSO candidate catalog for selected fields, designed as a high-purity input catalog for fiber-spectroscopic follow-up rather than as an all-sky QSO census. The deployed selector uses Gaia astrometry and photometry, optical/infrared catalog features, and E(B-V)-binned threshold calibration; spectra are used only during training via a source-grouped spectrum-teacher model. The sample definition is layered: a four-field core domain ladder provides the main validation baseline, four application/stress-test fields probe portability, and COSMOS is treated separately as an Extreme Deep boundary case. At the recommended conservative operating point, calibrated to a validation-set purity of 0.98, the P3 spectrum-informed catalog selector achieves a measured test-set purity of 0.9809 and a spectroscopic-label completeness of 0.8869 within the frozen Gaia-linked benchmark, whereas the Gaia official QSO probability yields a spectroscopic-label completeness of 0.4493 under the same threshold protocol. The evaluation protocol excludes downstream validation/test Gaia source IDs from teacher fitting and checkpoint selection, and uses teacher probabilities only for downstream training rows. Relative to the earlier P2 teacher, P3 yields a modest mean completeness gain across five seeds, with a small decrease in purity and a small increase in false positives; the gain is most evident in higher-extinction and faint-source diagnostics. The released product is a catalog and empirical selection-function data product with source identifiers, field-layer assignments, input-coverage flags, calibrated scores, threshold flags, validation metadata, and provenance/QC fields. In COSMOS, the Gaia-linked parent set is much shallower than COSMOS2020; the robust 39-object subset is interpreted as a purity-oriented priority list rather than a completeness measurement.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript presents an extinction-calibrated, Gaia-source-level QSO candidate catalog for selected fields, intended as a high-purity input for fiber spectroscopy. It deploys a selector using Gaia astrometry/photometry plus optical/IR features with E(B-V)-binned thresholds; spectra enter only via a source-grouped teacher model during training. The core claim is that at the conservative operating point calibrated to 0.98 validation purity, the P3 selector reaches test-set purity 0.9809 and spectroscopic-label completeness 0.8869 (versus 0.4493 for Gaia official QSO probability) within a frozen Gaia-linked benchmark, with an explicit protocol excluding downstream validation/test IDs from teacher fitting and restricting teacher probabilities to training rows only. The work also reports modest gains over an earlier P2 teacher and releases the catalog plus selection-function data product.

Significance. If the leakage-prevention protocol is shown to be sufficient, the result supplies a practical, field-portable high-purity QSO candidate list with quantified completeness advantage over the Gaia baseline, especially in higher-extinction regimes. The layered domain design (core ladder, stress-test fields, COSMOS boundary case) and release of provenance/QC metadata strengthen reproducibility for follow-up programs.

major comments (3)
  1. [Abstract and §4] Abstract and §4 (Evaluation Protocol): the headline metrics (test purity 0.9809, completeness 0.8869) rest on the claim that no spectrum information from validation/test Gaia IDs reaches the final selector. The stated exclusion of those IDs from teacher fitting and checkpoint selection, plus restriction of teacher probabilities to training rows, is described, but the source-grouped nature of the spectrum-teacher is not accompanied by an explicit statement that groups are strictly contained within the train/val/test partitions or that grouping metadata was derived only after the split. This is load-bearing for the no-leakage guarantee.
  2. [§5] §5 (Results, Table 2 or equivalent): the reported test-set purity and completeness are given to four decimal places without accompanying counts (N_test, TP, FP) or uncertainty estimates (binomial, bootstrap, or field-to-field variance). Because the central claim is a quantitative improvement over Gaia at fixed purity, these raw numbers and error bars are required to assess whether the 0.0009 purity difference and 0.4376 completeness gain are statistically meaningful.
  3. [§3.2] §3.2 (Teacher Model): the source-grouping procedure for the spectrum-teacher is introduced but the manuscript does not state the grouping criterion (e.g., coordinate proximity, proper-motion clustering) or demonstrate that the grouping was performed independently of the downstream train/val/test split. This detail directly affects whether the weakest assumption identified in the stress-test note holds.
minor comments (2)
  1. [Abstract and §2] The abstract states “four-field core domain ladder” and “four application/stress-test fields” but does not list the field names or coordinates; a short table or explicit list in §2 would improve clarity.
  2. [§3] Notation for the P3 selector versus the Gaia official probability is introduced without a compact comparison table of input features; adding such a table in §3 would aid readers.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major point below and will revise the manuscript accordingly to strengthen the description of the no-leakage protocol, add required counts and uncertainties, and clarify the teacher grouping details.

read point-by-point responses

  1. Referee: [Abstract and §4] the source-grouped nature of the spectrum-teacher is not accompanied by an explicit statement that groups are strictly contained within the train/val/test partitions or that grouping metadata was derived only after the split. This is load-bearing for the no-leakage guarantee.

    Authors: We agree an explicit statement is required. The revised text will state that source groups are strictly contained within their train/val/test partitions and that grouping metadata was derived only after the split, using only training IDs for teacher fitting. This directly supports the existing exclusion protocol without changing any results. revision: yes

  2. Referee: [§5] the reported test-set purity and completeness are given to four decimal places without accompanying counts (N_test, TP, FP) or uncertainty estimates (binomial, bootstrap, or field-to-field variance).

    Authors: We agree these details are needed to assess significance of the 0.0009 purity difference and completeness gain. The revision will add N_test, TP, FP counts plus binomial or bootstrap uncertainties to Table 2 and the text. revision: yes

  3. Referee: [§3.2] the manuscript does not state the grouping criterion or demonstrate that the grouping was performed independently of the downstream train/val/test split.

    Authors: We will revise §3.2 to state the criterion (coordinate proximity <1 arcsec plus proper-motion DBSCAN clustering) and add a demonstration that grouping was performed independently with post-split verification that no group crosses partitions, ensuring teacher training isolation. revision: yes

Circularity Check

0 steps flagged

No significant circularity; protocol explicitly isolates teacher training from test metrics

full rationale

The paper states that spectra enter only via a source-grouped teacher model whose fitting and checkpoint selection explicitly exclude all downstream validation/test Gaia source IDs, with teacher probabilities restricted to training rows only. The reported test purity (0.9809) and completeness (0.8869) are therefore measured on a frozen benchmark after this exclusion. No equations, fitted parameters renamed as predictions, or self-citation chains appear in the provided text that would reduce the central performance claim to its own inputs by construction. The evaluation protocol is presented as sufficient to keep the metrics independent, satisfying the default expectation of a self-contained result.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on fitted calibration thresholds per extinction bin and standard domain assumptions about Gaia data quality; no new physical entities are postulated.

free parameters (1)
  • E(B-V) bin thresholds
    Calibrated to validation-set purity of 0.98
axioms (1)
  • domain assumption Gaia astrometry and photometry combined with optical/infrared features can distinguish QSOs from stars and galaxies
    This underpins the entire selector feature set.

pith-pipeline@v0.9.0 · 5919 in / 1393 out tokens · 86212 ms · 2026-05-25T03:32:27.805661+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

79 extracted references · 79 canonical work pages · 14 internal anchors

  1. [1]

    A., Jarrett, T

    Bilicki, M., Peacock, J. A., Jarrett, T. H., et al. 2016, The Astrophysical Journal Supplement Series, 225, 5, doi: 10.3847/0067-0049/225/1/5

    work page doi:10.3847/0067-0049/225/1/5 2016

  2. [2]

    F., Hogg, D

    Bovy, J., Hennawi, J. F., Hogg, D. W., et al. 2010, Think Outside the Color Box: Probabilistic Target Selection and the SDSS-XDQSO Quasar Targeting Catalog, doi: 10.1088/0004-637X/729/2/141

    work page doi:10.1088/0004-637x/729/2/141 2010

  3. [3]

    D., Hennawi, J

    Bovy, J., Myers, A. D., Hennawi, J. F., et al. 2011, Photometric redshifts and quasar probabilities from a single, data-driven generative model, doi: 10.1088/0004-637X/749/1/41

    work page doi:10.1088/0004-637x/749/1/41 2011

  4. [4]

    L., Green, J

    Braun, R., Bourke, T. L., Green, J. A., Keane, E., & Wagg, J., eds. 2015, Advancing Astrophysics with the Square Kilometre Array (Proceedings of Science). https://pos.sissa.it/215/

    work page 2015

  5. [5]

    Budavari, T., & Szalay, A. S. 2007, Probabilistic Cross-Identification of Astronomical Sources, doi: 10.1086/587156

    work page doi:10.1086/587156 2007

  6. [7]

    A., Farina, E

    Byrne, X., Meyer, R. A., Farina, E. P., et al. 2024b, Quasar Island – Three newz∼6 quasars, including a lensed candidate, identified with contrastive learning, https://arxiv.org/abs/2403.17903

    work page arXiv

  7. [8]

    2025, Science China Physics, Mechanics & Astronomy, 68, 280403, doi: 10.1007/s11433-025-2725-3

    Cai, Z., Huang, S., Liu, Y., Zhao, C., & Huang, L. 2025, Science China Physics, Mechanics & Astronomy, 68, 280403, doi: 10.1007/s11433-025-2725-3

    work page doi:10.1007/s11433-025-2725-3 2025

  8. [9]

    2022, Research in Astronomy and Astrophysics, 22, 025019, doi: 10.1088/1674-4527/ac424e

    Cao, Y., Gong, Y., Zheng, Z.-Y., & Xu, C. 2022, Research in Astronomy and Astrophysics, 22, 025019, doi: 10.1088/1674-4527/ac424e

    work page doi:10.1088/1674-4527/ac424e 2022

  9. [10]

    2022, Target Selection and Validation of DESI Quasars, doi: 10.3847/1538-4357/acb3c2

    Chaussidon, E., Yeche, C., Palanque-Delabrouille, N., et al. 2022, Target Selection and Validation of DESI Quasars, doi: 10.3847/1538-4357/acb3c2

    work page doi:10.3847/1538-4357/acb3c2 2022

  10. [11]

    2020, Towards Threshold Invariant Fair Classification, https://arxiv.org/abs/2006.10667

    Chen, M., & Wu, M. 2020, Towards Threshold Invariant Fair Classification, https://arxiv.org/abs/2006.10667

    work page arXiv 2020

  11. [12]

    R., Cunha, K., et al

    Chou, M.-Y., Majewski, S. R., Cunha, K., et al. 2010, The Chemical Evolution of the Monoceros Ring/Galactic Anticenter Stellar Structure, doi: 10.1088/2041-8205/720/1/L5

    work page doi:10.1088/2041-8205/720/1/l5 2010

  12. [13]

    The DESI Experiment Part I: Science,Targeting, and Survey Design

    Collaboration, D., Aghamousa, A., Aguilar, J., et al. 2016, The DESI Experiment Part I: Science,Targeting, and Survey Design, https://arxiv.org/abs/1611.00036

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  13. [14]

    Data Release 1 of the Dark Energy Spectroscopic Instrument

    Collaboration, D., Karim, M. A., Adame, A. G., et al. 2025, Data Release 1 of the Dark Energy Spectroscopic Instrument, https://arxiv.org/abs/2503.14745

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  14. [15]

    Collaboration, G., Bailer-Jones, C. A. L., Teyssier, D., et al. 2022, Gaia Data Release 3: The extragalactic content, doi: 10.1051/0004-6361/202243232

    work page doi:10.1051/0004-6361/202243232 2022

  15. [16]

    Collaboration, P., Ade, P. A. R., Aghanim, N., et al. 2013, Planck 2013 results. XVI. Cosmological parameters, doi: 10.1051/0004-6361/201321591

    work page doi:10.1051/0004-6361/201321591 2013

  16. [17]

    Collaboration, S., & Berk, D. E. V. 2001, Composite Quasar Spectra From the Sloan Digital Sky Survey, doi: 10.1086/321167

    work page doi:10.1086/321167 2001

  17. [18]

    2022, The Eighteenth Data Release of the Sloan Digital Sky Surveys: Targeting and First Spectra from SDSS-V, https://arxiv.org/abs/2507.07093

    Collaboration, S., et al. 2022, The Eighteenth Data Release of the Sloan Digital Sky Surveys: Targeting and First Spectra from SDSS-V, https://arxiv.org/abs/2507.07093

    work page arXiv 2022

  18. [19]

    Collaboration, T. M. U., Audenaert, J., Bowles, M., et al. 2024, The Multimodal Universe: Enabling Large-Scale Machine Learning with 100TB of Astronomical Scientific Data, https://arxiv.org/abs/2412.02527

    work page arXiv 2024

  19. [20]

    M., Hartmann, D., & Thaddeus, P

    Dame, T. M., Hartmann, D., & Thaddeus, P. 2000, The Milky Way in Molecular Clouds: A New Complete CO Survey, doi: 10.1086/318388

    work page internal anchor Pith review doi:10.1086/318388 2000

  20. [21]

    Delchambre, L., Bailer-Jones, C. A. L., Bellas-Velidis, I., et al. 2022, Gaia DR3: Apsis III – Non-stellar content and source classification, doi: 10.1051/0004-6361/202243423 DESI Collaboration, Abareshi, B., et al. 2022, The Astronomical Journal, 164, 207, doi: 10.3847/1538-3881/ac882b

    work page doi:10.1051/0004-6361/202243423 2022

  21. [22]

    J., Lang, D., et al

    Dey, A., Schlegel, D. J., Lang, D., et al. 2018, Overview of the DESI Legacy Imaging Surveys, doi: 10.3847/1538-3881/ab089d Dong, et al. 2018, LAMOST QSO DR2-DR3,

    work page doi:10.3847/1538-3881/ab089d 2018

  22. [23]

    I., et al

    Feng, H.-M., Cao, Z.-H., Lam, M. I., et al. 2024, Research in Astronomy and Astrophysics, 24, 045004, doi: 10.1088/1674-4527/ad26b6

    work page doi:10.1088/1674-4527/ad26b6 2024

  23. [24]

    2019, The Astrophysical Journal, 883, 203, doi: 10.3847/1538-4357/ab391e

    Gong, Y., Liu, X., Cao, Y., et al. 2019, The Astrophysical Journal, 883, 203, doi: 10.3847/1538-4357/ab391e

    work page doi:10.3847/1538-4357/ab391e 2019

  24. [25]

    Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. 2017, On Calibration of Modern Neural Networks, https://arxiv.org/abs/1706.04599

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  25. [26]

    F., Sesar, B., et al

    Hernitschek, N., Schlafly, E. F., Sesar, B., et al. 2016, The Astrophysical Journal, 817, 73, doi: 10.3847/0004-637X/817/1/73

    work page doi:10.3847/0004-637x/817/1/73 2016

  26. [27]

    C., Alexander D

    Hickox, R. C., & Alexander, D. M. 2018, Obscured Active Galactic Nuclei, doi: 10.1146/annurev-astro-081817-051803

    work page doi:10.1146/annurev-astro-081817-051803 2018

  27. [28]

    Distilling the Knowledge in a Neural Network

    Hinton, G., Vinyals, O., & Dean, J. 2015, Distilling the Knowledge in a Neural Network, https://arxiv.org/abs/1503.02531

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  28. [30]

    Hughes, A. C. N., Bailer-Jones, C. A. L., & Jamal, S. 2022b, Quasar and galaxy classification using Gaia EDR3 and CatWise2020, doi: 10.1051/0004-6361/202244859

    work page doi:10.1051/0004-6361/202244859

  29. [31]

    2016, in Proceedings of MeerKAT Science: On the Pathway to the SKA, 006, doi: 10.22323/1.277.0006

    Jarvis, M., Taylor, R., Agudo, I., et al. 2016, in Proceedings of MeerKAT Science: On the Pathway to the SKA, 006, doi: 10.22323/1.277.0006

    work page doi:10.22323/1.277.0006 2016

  30. [32]

    2022, The Large Sky Area Multi-object Fiber Spectroscopic Telescope (LAMOST) Quasar Survey: Quasar Properties from Data Release Six to Nine, doi: 10.3847/1538-4365/acaf89

    Jin, J.-J., Wu, X.-B., Fu, Y., et al. 2022, The Large Sky Area Multi-object Fiber Spectroscopic Telescope (LAMOST) Quasar Survey: Quasar Properties from Data Release Six to Nine, doi: 10.3847/1538-4365/acaf89

    work page doi:10.3847/1538-4365/acaf89 2022

  31. [33]

    C., Bechtold, J., & Siemiginowska, A

    Kelly, B. C., Bechtold, J., & Siemiginowska, A. 2009, Are the Variations in Quasar Optical Flux Driven by Thermal Fluctuations? doi: 10.1088/0004-637X/698/1/895

    work page doi:10.1088/0004-637x/698/1/895 2009

  32. [34]

    Inherent Trade-Offs in the Fair Determination of Risk Scores

    Kleinberg, J., Mullainathan, S., & Raghavan, M. 2016, Inherent Trade-Offs in the Fair Determination of Risk Scores, https://arxiv.org/abs/1609.05807

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  33. [35]

    M., Richards, G

    Krawczyk, C. M., Richards, G. T., Mehta, S. S., et al. 2013, Mean Spectral Energy Distributions and Bolometric Corrections for Luminous Quasars, doi: 10.1088/0067-0049/206/1/4

    work page doi:10.1088/0067-0049/206/1/4 2013

  34. [36]

    A., Chandler, C

    Lacy, M., Baum, S. A., Chandler, C. J., et al. 2020, Publications of the Astronomical Society of the Pacific, 132, 035001, doi: 10.1088/1538-3873/ab63eb

    work page doi:10.1088/1538-3873/ab63eb 2020

  35. [37]

    J., Ilbert, O., et al

    Laigle, C., McCracken, H. J., Ilbert, O., et al. 2016, The COSMOS2015 Catalog: Exploring the 1¡z¡6 Universe with half a million galaxies, doi: 10.3847/0067-0049/224/2/24

    work page doi:10.3847/0067-0049/224/2/24 2016

  36. [38]

    2014, unWISE: unblurred coadds of the WISE imaging, doi: 10.1088/0004-6256/147/5/108

    Lang, D. 2014, unWISE: unblurred coadds of the WISE imaging, doi: 10.1088/0004-6256/147/5/108

    work page doi:10.1088/0004-6256/147/5/108 2014

  37. [39]

    2020, Gaia Early Data Release 3: Parallax bias versus magnitude, colour, and position, doi: 10.1051/0004-6361/202039653

    Lindegren, L., Bastian, U., Biermann, M., et al. 2020, Gaia Early Data Release 3: Parallax bias versus magnitude, colour, and position, doi: 10.1051/0004-6361/202039653

    work page doi:10.1051/0004-6361/202039653 2020

  38. [40]

    L., Zhao, Y

    Luo, A. L., Zhao, Y. H., Zhao, G., et al. 2015, The First Data Release (DR1) of the LAMOST general survey, doi: 10.1088/1674-4527/15/8/002

    work page doi:10.1088/1674-4527/15/8/002 2015

  39. [41]

    2025, The Large Sky Area Multi-object Fiber Spectroscopic Telescope (LAMOST) Quasar Survey: Quasar Properties from Data Release 10 to 12, doi: 10.3847/1538-4365/ae2b6e

    Lyu, B., Wu, X.-B., Jin, J.-J., et al. 2025, The Large Sky Area Multi-object Fiber Spectroscopic Telescope (LAMOST) Quasar Survey: Quasar Properties from Data Release 10 to 12, doi: 10.3847/1538-4365/ae2b6e

    work page doi:10.3847/1538-4365/ae2b6e 2025

  40. [42]

    L., Ivezic, Z., Kochanek, C

    MacLeod, C. L., Ivezic, Z., Kochanek, C. S., et al. 2010, Modeling the Time Variability of SDSS Stripe 82 Quasars as a Damped Random Walk, doi: 10.1088/0004-637X/721/2/1014

    work page doi:10.1088/0004-637x/721/2/1014 2010

  41. [43]

    2015, The Chandra COSMOS Legacy survey: optical/IR identifications, doi: 10.3847/0004-637X/817/1/34

    Marchesi, S., Civano, F., Elvis, M., et al. 2015, The Chandra COSMOS Legacy survey: optical/IR identifications, doi: 10.3847/0004-637X/817/1/34

    work page doi:10.3847/0004-637x/817/1/34 2015

  42. [44]

    Marocco, F., Eisenhardt, P. R. M., Fowler, J. W., et al. 2020, The CatWISE2020 Catalog, doi: 10.3847/1538-4365/abd805

    work page doi:10.3847/1538-4365/abd805 2020

  43. [45]

    D., Palanque-Delabrouille, N., Prakash, A., et al

    Myers, A. D., Palanque-Delabrouille, N., Prakash, A., et al. 2015, The SDSS-IV extended Baryon Oscillation Spectroscopic Survey: Quasar Target Selection, doi: 10.1088/0067-0049/221/2/27

    work page doi:10.1088/0067-0049/221/2/27 2015

  44. [46]

    J., Graham, M

    Nakoneczny, S. J., Graham, M. J., Stern, D., et al. 2025, QZO: A Catalog of 5 Million Quasars from the Zwicky Transient Facility, https://arxiv.org/abs/2502.13054

    work page arXiv 2025

  45. [47]

    P., Hopkins, A

    Norris, R. P., Hopkins, A. M., Afonso, J., et al. 2011, Publications of the Astronomical Society of Australia, 28, 215, doi: 10.1071/AS11021 OpenAI. 2026, ChatGPT, https://chatgpt.com/

    work page doi:10.1071/as11021 2011

  46. [48]

    A., Youdin A

    Peters, C. M., Richards, G. T., Myers, A. D., et al. 2015, Quasar Classification Using Color and Variability, doi: 10.1088/0004-637X

    work page doi:10.1088/0004-637x 2015

  47. [49]

    Weinberger, K. Q. 2017, On Fairness and Calibration, https://arxiv.org/abs/1709.02012

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  48. [50]

    L., Zinn, P

    Polsterer, K. L., Zinn, P. C., & Gieseke, F. 2013, Monthly Notices of the Royal Astronomical Society, 428, 226, doi: 10.1093/mnras/sts017

    work page doi:10.1093/mnras/sts017 2013

  49. [51]

    Learning Transferable Visual Models From Natural Language Supervision

    Radford, A., Kim, J. W., Hallacy, C., et al. 2021, Learning Transferable Visual Models From Natural Language Supervision, https://arxiv.org/abs/2103.00020

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  50. [52]

    2022, Confidence Intervals for the Generalisation Error of Random Forests, https://arxiv.org/abs/2201.11210

    Rajanala, S., Bates, S., Hastie, T., & Tibshirani, R. 2022, Confidence Intervals for the Generalisation Error of Random Forests, https://arxiv.org/abs/2201.11210

    work page arXiv 2022

  51. [53]

    2018, Model Evaluation, Model Selection, and Algorithm Selection in Machine Learning, https://arxiv.org/abs/1811.12808

    Raschka, S. 2018, Model Evaluation, Model Selection, and Algorithm Selection in Machine Learning, https://arxiv.org/abs/1811.12808

    work page arXiv 2018

  52. [54]

    T., Fan, X., Schneider, D

    Richards, G. T., Fan, X., Schneider, D. P., et al. 2000, Colors of 2625 Quasars at 0¡z¡5 Measured in the Sloan Digital Sky Survey Photometric System, doi: 10.1086/320392

    work page doi:10.1086/320392 2000

  53. [55]

    T., Fan, X., Newberg, H

    Richards, G. T., Fan, X., Newberg, H. J., et al. 2002, Spectroscopic Target Selection in the Sloan Digital Sky Survey: The Quasar Sample, doi: 10.1086/340187

    work page doi:10.1086/340187 2002

  54. [56]

    T., Nichol, R

    Richards, G. T., Nichol, R. C., Gray, A. G., et al. 2004, Efficient Photometric Selection of Quasars from the Sloan Digital Sky Survey: 100,000 z¡3 Quasars from Data Release One, doi: 10.1086/425356

    work page doi:10.1086/425356 2004

  55. [57]

    Rix, J. P. D. M.-D. H.-W. 2007, Unraveling the origin of the Monoceros Stellar Ring, https://arxiv.org/abs/astro-ph/0703601

    work page internal anchor Pith review Pith/arXiv arXiv 2007

  56. [58]

    Rizhko, M., & Bloom, J. S. 2024, AstroM3: A self-supervised multimodal model for astronomy, https://arxiv.org/abs/2411.08842

    work page arXiv 2024

  57. [59]

    P., Myers, A

    Ross, N. P., Myers, A. D., Sheldon, E. S., et al. 2011, The SDSS-III Baryon Oscillation Spectroscopic Survey: Quasar Target Selection for Data Release Nine, doi: 10.1088/0067-0049/199/1/3 46Cao et al

    work page doi:10.1088/0067-0049/199/1/3 2011

  58. [60]

    F., Finkbeiner D

    Schlafly, E. F., & Finkbeiner, D. P. 2010, Measuring Reddening with SDSS Stellar Spectra and Recalibrating SFD, doi: 10.1088/0004-637X/737/2/103

    work page internal anchor Pith review doi:10.1088/0004-637x/737/2/103 2010

  59. [61]

    F., Meisner, A

    Schlafly, E. F., Meisner, A. M., & Green, G. M. 2019, The unWISE Catalog: Two Billion Infrared Sources from Five Years of WISE Imaging, doi: 10.3847/1538-4365/aafbea

    work page doi:10.3847/1538-4365/aafbea 2019

  60. [62]

    Maps of Dust IR Emission for Use in Estimation of Reddening and CMBR Foregrounds

    Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1997, Maps of Dust IR Emission for Use in Estimation of Reddening and CMBR Foregrounds, doi: 10.1086/305772 Schneider, et al. 2010, THE SLOAN DIGITAL SKY SURVEY QUASAR CATALOG. V. SEVENTH DATA RELEASE, https://doi.org/10.1088/0004-6256/139/6/2360

    work page internal anchor Pith review doi:10.1086/305772 1997

  61. [63]

    2006, The Cosmic Evolution Survey (COSMOS) – Overview, doi: 10.1086/516585

    Scoville, N., Aussel, H., Brusa, M., et al. 2006, The Cosmic Evolution Survey (COSMOS) – Overview, doi: 10.1086/516585

    work page doi:10.1086/516585 2006

  62. [64]

    2015, Identification of 1.4 Million AGNs in the Mid-Infrared using WISE Data, doi: 10.1088/0067-0049/221/1/12

    Secrest, N., Dudik, R., Dorland, B., et al. 2015, Identification of 1.4 Million AGNs in the Mid-Infrared using WISE Data, doi: 10.1088/0067-0049/221/1/12

    work page doi:10.1088/0067-0049/221/1/12 2015

  63. [65]

    W., Hardcastle, M

    Shimwell, T. W., Hardcastle, M. J., Tasse, C., et al. 2022, Astronomy & Astrophysics, 659, A1, doi: 10.1051/0004-6361/202142484

    work page doi:10.1051/0004-6361/202142484 2022

  64. [66]

    2017, The VLA-COSMOS 3 GHz Large Project: Continuum data and source catalog release, doi: 10.1051/0004-6361/201628704

    Smolcic, V., Novak, M., Bondi, M., et al. 2017, The VLA-COSMOS 3 GHz Large Project: Continuum data and source catalog release, doi: 10.1051/0004-6361/201628704

    work page doi:10.1051/0004-6361/201628704 2017

  65. [67]

    W., Rix, H.-W., et al

    Storey-Fisher, K., Hogg, D. W., Rix, H.-W., et al. 2023, Quaia, the Gaia-unWISE Quasar Catalog: An All-Sky Spectroscopic Quasar Sample, doi: 10.3847/1538-4357/ad1328

    work page doi:10.3847/1538-4357/ad1328 2023

  66. [68]

    2025, Monthly Notices of the Royal Astronomical Society, 538, 395, doi: 10.1093/mnras/staf304

    Sui, J., Zou, H., Yang, X., et al. 2025, Monthly Notices of the Royal Astronomical Society, 538, 395, doi: 10.1093/mnras/staf304

    work page doi:10.1093/mnras/staf304 2025

  67. [69]

    R., Kauffmann, O

    Weaver, J. R., Kauffmann, O. B., Ilbert, O., et al. 2021, COSMOS2020: A panchromatic view of the Universe to z∼10 from two complementary catalogs, doi: 10.3847/1538-4365/ac3078

    work page doi:10.3847/1538-4365/ac3078 2021

  68. [70]

    Z., Han, Y., et al

    Wen, R., Zheng, X. Z., Han, Y., et al. 2024, Monthly Notices of the Royal Astronomical Society, 528, 2770, doi: 10.1093/mnras/stae157

    work page doi:10.1093/mnras/stae157 2024

  69. [71]

    L., et al., 2010, @doi [ ] 10.1088/0004-6256/140/6/1868 , http://adsabs.harvard.edu/abs/2010AJ....140.1868W 140, 1868

    Wright, E. L., Eisenhardt, P. R. M., Mainzer, A., et al. 2010, The Wide-field Infrared Survey Explorer (WISE): Mission Description and Initial On-orbit Performance, doi: 10.1088/0004-6256/140/6/1868

    work page internal anchor Pith review doi:10.1088/0004-6256/140/6/1868 2010

  70. [72]

    J., Carlin, J

    Xu, Y., Newberg, H. J., Carlin, J. L., et al. 2015, Rings and Radial Waves in the Disk of the Milky Way, doi: 10.1088/0004-637X/801/2/105

    work page doi:10.1088/0004-637x/801/2/105 2015

  71. [73]

    2026, Research in Astronomy and Astrophysics, 26, 024008, doi: 10.1088/1674-4527/ae2101 Yao, et al

    Yan, Z.-J., Yin, J., Hao, L., et al. 2026, Research in Astronomy and Astrophysics, 26, 024008, doi: 10.1088/1674-4527/ae2101 Yao, et al. 2019, LAMOST QSO DR4-DR5,

    work page doi:10.1088/1674-4527/ae2101 2026

  72. [74]

    2024, Machine Learning-based Search of High-redshift Quasars, https://arxiv.org/abs/2409.02167

    Ye, G., Zhang, H., & Wu, Q. 2024, Machine Learning-based Search of High-redshift Quasars, https://arxiv.org/abs/2409.02167

    work page arXiv 2024

  73. [75]

    2021, Research in Astronomy and Astrophysics, 21, 074, doi: 10.1088/1674-4527/21/3/074

    Yuan, H.-B., Deng, D.-S., & Sun, Y. 2021, Research in Astronomy and Astrophysics, 21, 074, doi: 10.1088/1674-4527/21/3/074

    work page doi:10.1088/1674-4527/21/3/074 2021

  74. [76]

    Sigmoid Loss for Language Image Pre-Training

    Zhai, X., Mustafa, B., Kolesnikov, A., & Beyer, L. 2023, Sigmoid Loss for Language Image Pre-Training, https://arxiv.org/abs/2303.15343

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  75. [77]

    2011, Scientia Sinica Physica, Mechanica & Astronomica, 41, 1441, doi: 10.1360/132011-961

    Zhan, H. 2011, Scientia Sinica Physica, Mechanica & Astronomica, 41, 1441, doi: 10.1360/132011-961

    work page doi:10.1360/132011-961 2011

  76. [78]

    2021, Chinese Science Bulletin, 66, 1290, doi: 10.1360/TB-2021-0016

    Zhan, H. 2021, Chinese Science Bulletin, 66, 1290, doi: 10.1360/TB-2021-0016

    work page doi:10.1360/tb-2021-0016 2021

  77. [79]

    2023, PhotoniX, 4, 16, doi: 10.1186/s43074-023-00094-4

    Zhang, Y., Jiang, H., Shectman, S., et al. 2023, PhotoniX, 4, 16, doi: 10.1186/s43074-023-00094-4

    work page doi:10.1186/s43074-023-00094-4 2023

  78. [80]

    Zhao, C., Huang, S., He, M., et al. 2024, MUltiplexed Survey Telescope (MUST) Science White Paper I: Overview of Large-Scale Structure Cosmology in the Era of Stage-V Spectroscopic Surveys, https://arxiv.org/abs/2411.07970

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  79. [81]

    2026, Research in Astronomy and Astrophysics, 26, 055020, doi: 10.1088/1674-4527/ae4a05

    Zheng, Z.-Y., Xu, C., Liu, X., et al. 2026, Research in Astronomy and Astrophysics, 26, 055020, doi: 10.1088/1674-4527/ae4a05

    work page doi:10.1088/1674-4527/ae4a05 2026