Pith. sign in

REVIEW 2 major objections 4 minor 63 references

Tabular foundation models for the estimation of probabilistic quasar photometric redshifts in S-PLUS

T0 review · 2 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A frozen tabular foundation model, TabPFN 2.5, is the strongest overall probabilistic quasar redshift estimator in S-PLUS DR6.

desk verdict Solid, carefully run benchmark showing TabPFN 2.5 is the best default for S-PLUS quasar photo-z, with the caveat that the covariate-shift calibration claim rests on assumptions the data cannot test. read the letter →

arxiv 2608.10280 v1 pith:ITOF5RV5 submitted 2026-08-10 astro-ph.IM astro-ph.CO

Raquel R. Valença , Lilianne Nakazono , Rafael Izbicki , Marco Henrique de Almeida Inácio , Bruno Marcondes e Resende , Maycon Jorge Deláqua da Silva , Kiana Coimbra Buin Lins , Natanael M. Cardoso
show 1 more author
Claudia Mendes de Oliveira
This is my paper · ORCID
classification astro-ph.IMastro-ph.CO
keywords photometricredshiftsquasarstabularfoundationmodelsTabPFNconditionaldensityestimationcovariateshiftimportanceweightingS-PLUSsurvey
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper asks whether a frozen, off-the-shelf tabular foundation model can replace survey-specific, trained-from-scratch estimators for probabilistic quasar photometric redshifts in the 12-band S-PLUS DR6 survey. It claims that TabPFN 2.5, used without any fine-tuning, is the strongest overall such estimator: best or statistically tied for best on nearly every density and point metric, with its largest advantages at small training-set sizes and in difficult regimes such as bright sources, faint sources, and high redshift. The paper also claims that TabPFN 2.5 keeps near-nominal calibration when evaluation is reweighted from the spectroscopic test distribution to the photometric target population under a covariate-shift assumption. A sympathetic reader would care because quasars have genuinely multimodal redshift posteriors and spectroscopic samples are shifted relative to photometric populations, so a ready-made probabilistic estimator that stays calibrated under that shift would be a practical default for survey-scale redshift inference.

What carries the argument

The load-bearing object is TabPFN 2.5 used as a tabular in-context foundation model: a transformer pre-trained on a large corpus of synthetic tabular regression tasks, applied with frozen weights by supplying the labelled spectroscopic sample as in-context support and reading out a piecewise-constant "bar" distribution over redshift, which is interpolated onto a 200-point redshift grid and renormalised to give a conditional density estimate. The companion machinery is the covariate-shift evaluation: importance weights equal to the ratio of target to source feature densities, estimated by a classifier distinguishing spectroscopic from photometric quasar candidates, tempered with an exponent to control effective sample size, so that weighted metrics approximate performance on the photometric target population. These two pieces together let the paper separate how well a method fits the spectroscopic sample from how well it will behave when deployed on the photometric catalogue.

What would settle it

Obtain spectroscopic redshifts for a random, feature-unselected subsample of the photometric quasar-candidate catalogue and compare the conditional redshift distributions at fixed features with those of the spectroscopic training sample; if they differ measurably, or if removing spectroscopically confirmed non-quasars from the candidate catalogue changes the importance-weighted ranking, the paper's calibration claim for TabPFN 2.5 would be refuted.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a transformer pre-trained on synthetic tabular tasks and applied with frozen weights can estimate the conditional redshift density for quasars at least as well as the best task-specific conditional-density estimators, and better in the regimes that matter most. In the benchmark with 121,626 training quasars and 13,538 test quasars, TabPFN 2.5 is best or statistically tied for best on CDE loss, log-likelihood, CRPS, PIT-KS, 90% coverage, RMSE, NMAD, and the catastrophic-outlier fractions; the one exception is unweighted CDE loss, where the normalising flow Flow-Spline is statistically tied and attains the lower mean. Under importance-weighted evaluation toward the photometric candidate population, TabPFN 2.5 remains best or tied for best on essentially all metrics while competitors such as FlexZBoost show degraded coverage, and its PIT-KS stays close to its unweighted value. Its largest gains appear when training sets are small (at most 1,000 objects) and in the low- and high-redshift tails, where catastrophic photo-z failures are most common.

Load-bearing premise

The load-bearing premise is that, at fixed observed photometric features, the redshift distribution of quasars is the same in the spectroscopic sample and the photometric candidate catalogue, and that the candidate catalogue is a sufficiently pure draw from the deployment population; if targeting carries extra information beyond the 39 features, or the catalogue contains non-quasar contaminants, the importance-weighted performance claims are biased.

Editorial extensions

If this is right

  • TabPFN 2.5 can serve as a strong default probabilistic photo-z estimator for S-PLUS-like quasar samples, especially when training data are limited or calibrated conditional densities matter.
  • The three foundation models are the only methods that stay competitive on catastrophic-outlier fractions under importance-weighted evaluation, so survey pipelines that prioritise outlier control should consider them.
  • Method rankings change under covariate-shift weighting: unweighted spectroscopic metrics overstate deployed performance for some baselines, for example FlexZBoost's 90% coverage drops from 0.883 unweighted to 0.794 under raw importance weights.
  • Foundation models lead at training sizes of 1,000 or fewer on nearly all metrics; at the full DR6 training size the gap narrows, but TabPFN 2.5 remains among the leading methods.
  • SHAP attributions identify WISE W1 and W2 as the strongest individual predictors in TabPFN 2.5, with UV and optical bands providing smaller collective refinements.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the covariate-shift assumption holds generally, the same frozen-model protocol could transfer to other narrow-band surveys and to galaxies, where multimodality and sample shift are also present; the paper explicitly leaves this extension untested.
  • Because TabPFN 2.5's advantage concentrates in sparsely sampled tails, an untested but plausible prediction is that it will also dominate in deeper surveys where the photometric population extends beyond the spectroscopic support.
  • The main deployment bottleneck is inference memory, so a practical route would be to distill the frozen model's conditional densities into a cheaper survey-specific emulator; the paper notes distillation as a possible acceleration but does not test it.
  • A purity-aware weighting that uses calibrated candidate probabilities instead of a raw candidate catalogue could strengthen or weaken the reported calibration gains; the paper suggests this as future work.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper presents a benchmark of three tabular foundation models (TabPFN 2.5, RealTabPFN 2.5, and TabICL) against eight task-specific baselines for probabilistic quasar photometric redshift estimation in S-PLUS DR6. The feature set has 39 dimensions (S-PLUS PSF magnitudes, colours, magnitude errors, WISE W1/W2, and GALEX FUV/NUV); training sizes range from 500 to 121,626 with five repetitions and a fixed test set of 13,538 objects. The evaluation covers density quality (CDE loss, log-likelihood, CRPS, PIT-KS, 90% coverage) and point predictions (RMSE, bias, NMAD, outlier fractions), in both an unweighted mode and an importance-weighted mode designed to approximate the photometric target population, with tempered weights and a sensitivity analysis for the weight estimator. The main finding is that TabPFN 2.5 is best or statistically tied for best on nearly all metrics, with the largest advantages at small training sizes and in difficult regimes; Flow-Spline ties it on unweighted CDE loss. The paper concludes that TabPFN 2.5 is a strong default for S-PLUS-like probabilistic quasar photo-z inference.

Significance. The unweighted benchmark is careful and reproducible: a fixed train/test split inherited from the S-PLUS pipeline, five independent repetitions, per-repetition averaging with test-set bootstrap standard errors, a documented one-SE bold rule, a grid-resolution sensitivity check, and a full re-estimation of importance weights with a different classifier in Appendix C. The result that a frozen in-context foundation model is at least competitive with tuned task-specific density estimators for quasar photo-z is of practical value, especially for small spectroscopic samples, and the paper gives proper credit to the earlier S-PLUS QuCatS pipeline. The fundamental conclusion that TabPFN 2.5 is a strong default in the unweighted, distribution-matched sense is well supported. The deployment-oriented weighted claims, however, inherit the covariate-shift and target-purity assumptions discussed below. This is a benchmark rather than a derivation, and I see no circularity in the reported comparisons.

major comments (2)
  1. [§3.2 (Eq. 1) and §2.6 (Table 4)] The weighted metrics and the abstract's claim of 'near-nominal calibration under covariate shift' rest on the assumptions that p_s(z|x)=p_t(z|x) and that the DR6 photometric quasar-candidate catalogue is a representative, sufficiently pure draw from the deployment population. Table 4 shows that the target sample is dramatically different from the spectroscopic sample: r is missing for 55.2% of targets, u-r for 94.1%, and W1-W2 for 92.8%, with two-sample KS distances of 0.388-0.655 for the valid measurements. The density-ratio classifier in §3.2.1 therefore separates the samples largely on missingness patterns. If the candidate catalogue contains appreciable non-QSO contamination, or if spectroscopic targeting uses information beyond the 39 features (variability, morphology, ancillary detections), then beta(x) reweights toward a population for which Eq. (1) is undefined or false, and the weighted calibration and outlier-rate conclusions are biased in an unknown direction. Section 5 acknowledges this, but the conclusion still states the weighted behaviour as a property of TabPFN 2.5. Please add a purity-aware analysis (for example, using calibrated candidate probabilities P(QSO|x) as target weights, or sweeping assumed contamination fractions) or explicitly downgrade the weighted claims to conditional-on-purity statements.
  2. [§3.2.2 (Eq. 5) and Table 2, panel (b)] The headline importance-weighted evaluation uses alpha* ≈ 0.36, chosen to force the effective sample size to 30% of the test set; this is not the target distribution p_t but a variance-reduced interpolation between the spectroscopic and photometric distributions. The alpha = 1 panel, which actually estimates performance on p_t, has ESS ≈ 216 objects (1.6%) and gives TabPFN 2.5 a weighted PIT-KS of 0.064 ± 0.023 and 90% coverage of 0.910 ± 0.015. These numbers do not by themselves establish near-nominal calibration on the target population with meaningful precision. The abstract and Section 5 should be reworded so that 'near-nominal calibration under covariate shift' refers to the tempered interpolation, with target-population calibration reported as a high-variance, assumption-dependent estimate.
minor comments (4)
  1. [Abstract and Appendix B] The phrase 'statistically tied' corresponds to the one-combined-SE bold rule defined in Appendix B; please add a footnote clarifying that this is a descriptive rule and not a formal equivalence test.
  2. [Appendix B] The reported standard errors are finite-test-set bootstrap errors after averaging per-object values across the five repetitions; they do not include uncertainty from training-set subsampling. This should be stated explicitly so that readers do not interpret the SEs as full procedure uncertainty.
  3. [Section 4.4] The SHAP analysis explains only 100 test objects with 83 coalitions per object; the global feature-importance ranking is therefore a rough diagnostic. Please state the sensitivity of the ranking to the number of explained objects and to the coalition budget.
  4. [Table 4] Because missingness itself is a strong discriminator (94.1% for u-r and 92.8% for W1-W2), the text should note explicitly that the importance-weighted evaluation is reweighting toward objects with incomplete photometry, not only toward fainter or redder objects.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the benchmark is a held-out evaluation against fixed spectroscopic redshifts, and the importance weighting is an explicit identity under a stated, untestable covariate-shift assumption rather than a constructed prediction.

full rationale

This paper is a controlled benchmark, not a derivation: TabPFN 2.5, RealTabPFN 2.5, and TabICL are evaluated with frozen weights on a fixed spectroscopic test set with true redshifts, and no metric is produced by fitting the model to the quantity it is claimed to predict. The importance-weighting chain in Section 3.2 is an identity: Eq. (2) follows from the covariate-shift assumption in Eq. (1), and Eq. (4) is the Bayes-rule estimator of beta(x); the weights depend only on the features x and the S=0/S=1 label, not on redshifts or on the photo-z models being ranked. The tempering exponent alpha in Eq. (5) is transparently an evaluation choice, and Appendix C shows that replacing the TabICL weight estimator with gradient boosting leaves the rankings essentially unchanged. Self-citations to Nakazono et al. (2024) for the train/test split and to Izbicki and Lee for FlexZBoost and the CDE loss are methodological anchors rather than load-bearing premises. The paper also flags its real limitation explicitly in Section 5: 'uncertainty in the weights and impurity of the photometric candidate catalogue remain' and 'the covariate-shift assumption may also fail'; Appendix A, Table 4, quantifies the large missingness differences between the spectroscopic and photometric samples. That limitation means the deployment-oriented weighted claim is conditional on an untestable assumption, but it is an external-validity caveat, not a circular reduction of the result to its inputs. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no ansatz is smuggled in via self-citation.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper introduces no physical entities and no closed-form derivation. Its central claim rests on standard probability, on the untestable covariate-shift assumption, and on evaluation choices (alpha, clipping, grid resolution) that are reported and partly tested. Baseline hyperparameters are selected by cross-validation, not by fitting the headline conclusion.

free parameters (3)
  • Tempering exponent alpha = alpha ≈ 0.36 (ESS target 30%)
    Chosen to hit an effective sample size of 30% of the test set; the headline shift-robustness numbers are reported at this alpha rather than at the raw alpha=1 weights.
  • Importance-weight clipping bounds = [0.01, 0.99]
    Prevents extreme density-ratio weights from dominating the weighted metrics; the choice affects all weighted tables and is only partially explored via the boosting-based weight estimator.
  • Density grid resolution = 200 redshift values (10% expansion)
    Density metrics, especially CDE loss, depend on grid spacing; Appendix C shows the TabPFN/Flow-Spline CDE ordering reverses when the grid is coarsened to 100 points.
assumptions (4)
  • domain assumption Covariate-shift assumption: p(z|x) is identical in the spectroscopic and photometric target populations (Eq. 1).
    Used to justify importance-weighted target-population metrics; cannot be tested from these data and could fail if selection uses features outside the 39-vector or if the candidate catalogue is impure.
  • domain assumption Spectroscopic redshifts are exact ground truth with no reported uncertainty and the 1-arcsec cross-match correctly associates counterparts.
    The full benchmark treats z_spec as labels; mismatches or redshift errors would bias all methods, especially in faint and high-redshift bins.
  • standard math Bayesian and probabilistic identities for density-ratio estimation and proper scoring rules are valid.
    The importance weights (Eq. 4) and metrics (Eqs. 6-8) rest on standard probability; not load-bearing beyond their usual assumptions.
  • domain assumption TabPFN's pretrained prior provides a valid inductive bias for tabular regression.
    The paper assumes the in-context posterior is a meaningful conditional density after interpolation and renormalization; this is an external model assumption, tested only empirically.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Tabular foundation models for the estimation of probabilistic quasar photometric redshifts in S-PLUS." pith.science (2026). https://pith.science/paper/ITOF5RV5

@misc{pith2026260810280,
  author       = {Pith},
  title        = {Pith review of: Tabular foundation models for the estimation of probabilistic quasar photometric redshifts in S-PLUS},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ITOF5RV5}},
  note         = {Machine review of arXiv:2608.10280}
}
read the original abstract

We assess whether tabular foundation models can be used as off-the-shelf probabilistic photometric-redshift estimators for quasars in the 12-band S-PLUS DR6 survey, where colour-redshift degeneracies produce multi-modal posteriors and spectroscopic training sets are shifted relative to the photometric population. TabPFN 2.5, RealTabPFN 2.5, and TabICL are benchmarked against eight task-specific baselines, including linear conditional Gaussians, FlexZBoost, mixture-density networks, normalising flows, random forests, and gradient-boosted trees, with training sets from 500 to 121,626 quasars, using both density and point-prediction metrics, together with importance-weighted scores that approximate deployment on the photometric target sample. TabPFN 2.5 is best or statistically tied for best on all metrics except the unweighted CDE loss, on which the normalising flow is statistically tied and attains the lowest mean value; its largest gains occur for small training sets and in difficult regimes (very bright and faint sources, high redshift), while retaining near-nominal calibration under covariate shift. Its main practical cost is inference: with frozen weights, large support and target catalogues require substantial GPU/accelerator memory, and full-catalogue deployment may need support-set subsampling or distillation. SHAP attributions identify WISE W1/W2 as the strongest individual predictors, with UV and optical bands offering non-negligible refinements. We conclude that TabPFN 2.5 is a strong default for probabilistic quasar photo-z estimation, particularly when training data are limited or when calibration under covariate shift is critical.

Figures

Figures reproduced from arXiv: 2608.10280 by the authors.

Figure 1
Figure 1. PIT histograms for TabPFN 2.5, FlexZBoost, Flow-Spline, and MDN-deep. Panels are ordered by PIT-KS. The dotted line at density 1 marks the target Uniform(0, 1) density; the grey band is the 95% pointwise binomial enve￾lope expected under exact uniformity at this test-set size, and per-bin error bars are ±1 SE. TabPFN 2.5 shows the smallest aggregate PIT-KS among these methods and no large structured departure from u… view at source ↗
Figure 2
Figure 2. Example photo-z conditional densities on the DR6 test set at ntrain = 121 626 for the three leading density estimators: TabPFN 2.5 (red, solid; foundation), FlexZBoost (grey, dashed; tree-based), and Flow-Spline (blue, dotted; neural). The 3 × 3 grid cross-tabulates three r-magnitude bins (rows) against three zspec bins (columns); within each cell the displayed object is chosen uniformly at random (fixed seed). The … view at source ↗
Figure 3
Figure 3. complements the redshift breakdown by re￾binning the same three methods against the S-PLUS broad-band r magnitude, with bin width ∆r = 0.25 and a minimum of 30 test objects per bin. Two metrics are shown: panel (a) the per-bin RMSE and panel (b) the outlier fraction. As expected, both metrics worsen overall toward fainter magnitudes, although the finite bin counts introduce bin-to-bin fluctuations. On RMSE, TabPFN 2… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Per-redshift-bin RMSE and η0.15 on the DR6 test set at ntrain = 121 626 for the three methods with the lowest overall unweighted RMSE in [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Performance versus training-set size under the unweighted spectroscopic test distribution. Each curve shows the mean over five repetitions and error bars indicate ±1 SE. Methods are colour-coded by family (linear, tree-based, neural, foundation). Foundation models (red…
Figure 6
Figure 6. Figure 6: Performance versus training-set size under importance-weighted evaluation with the tuned recalibration exponent (α ≃ 0.36, ESS target 30%). Conventions match [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: SHAP interpretation for TabPFN 2.5. Left: beeswarm summary over 100 explained test objects. Features are ordered by mean absolute SHAP value; each point is one object, the horizontal position gives the contribution to the predicted redshift, and colour encodes the feat…
Figure 8
Figure 8. Figure 8: Ranks of the across-repetition mean performance at ntrain = 121 626 across the reported metrics and evaluation regimes. Rank 1 is best within each column; coverage is ranked by distance from 0.90 and bias by absolute value. White cells denote metrics that do not apply …
Figure 9
Figure 9. Figure 9: PIT P–P curves at ntrain = 121 626 before and after importance weighting. The black, blue, and orange curves correspond to the unweighted, tempered (α ≃ 0.36), and raw (α = 1) evaluations, respectively. The dotted diagonal represents exact PIT uniformity, and the PIT–K…
Figure 10
Figure 10. Figure 10: Targeted difficult photo-z cases selected using the catastrophic point-prediction criterion ∆z = |zˆ − zspec|/(1 + zspec) > 0.15. Panel (a) shows a high-redshift object for which TabPFN 2.5 is non-catastrophic while FlexZBoost and Flow-Spline are catastrophic; panel (…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 21 canonical work pages

  1. [1]

    G., Aguado, D., et al

    Abdul Karim, M., Adame, A. G., Aguado, D., et al. 2026, Astronomical Journal, 171, doi: 10.3847/1538-3881/ae4c43

    doi:10.3847/1538-3881/ae4c43 2026

  2. [2]

    2016, Monthly Notices of the Royal Astronomical Society, 459, 1293

    Asorey, J., Carrasco Kind, M., Sevilla-Noarbe, I., Brunner, R., & Thaler, J. 2016, Monthly Notices of the Royal Astronomical Society, 459, 1293

    work page 2016

  3. [3]

    J., Stern, D., Kochanek, C

    Assef, R. J., Stern, D., Kochanek, C. S., et al. 2013, ApJ, 772, 26, doi: 10.1088/0004-637X/772/1/26

    doi:10.1088/0004-637x/772/1/26 2013

  4. [4]

    J., Myers, A., et al

    Ball, N., Brunner, R. J., Myers, A., et al. 2008, The Astrophysical Journal, 683, 12, doi: 10.1086/589646

    doi:10.1086/589646 2008

  5. [5]

    2010, in Proceedings of Machine Learning Research, Vol

    Ben-David, S., Lu, T., Luu, T., & P´ al, D. 2010, in Proceedings of Machine Learning Research, Vol. 9, Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, 129–136. https://proceedings.mlr.press/v9/david10a.html Ben’ıtez, N., Dupke, R., Moles, M., & et al. 2014, arXiv e-prints

    work page 2010

  6. [6]

    2017, ApJS, 230, 24, doi: 10.3847/1538-4365/aa7053

    Bianchi, L., et al. 2017, ApJS, 230, 24, doi: 10.3847/1538-4365/aa7053

    doi:10.3847/1538-4365/aa7053 2017

  7. [7]

    2009, Journal of Machine Learning Research, 10, 2137, doi: 10.5555/1577069.1755858

    Bickel, S., Br¨ uckner, M., & Scheffer, T. 2009, Journal of Machine Learning Research, 10, 2137, doi: 10.5555/1577069.1755858

    arXiv 2009

  8. [8]

    J., & Amara, A

    Bordoloi, R., Lilly, S. J., & Amara, A. 2010, Monthly Notices of the Royal Astronomical Society, 406, 881

    work page 2010

Show all 63 references
  1. [9]

    B., van Dokkum, P

    Brammer, G. B., van Dokkum, P. G., & Coppi, P. 2008, The Astrophysical Journal, 686, 1503, doi: 10.1086/591786

    2008 doi

  2. [10]

    2001, Machine Learning, 45, 5, doi: 10.1023/A:1010933404324

    Breiman, L. 2001, Machine Learning, 45, 5, doi: 10.1023/A:1010933404324

    2001 doi

  3. [11]

    2013, The Astrophysical Journal, 772, 140, doi: 10.1088/0004-637X/772/2/140

    Mercurio, A. 2013, The Astrophysical Journal, 772, 140, doi: 10.1088/0004-637X/772/2/140

    2013 doi

  4. [12]

    J., Mol´ es, M., Crist’obal-Hornillos, D., et al

    Cenarro, A. J., Mol´ es, M., Crist’obal-Hornillos, D., et al. 2019, Astronomy & Astrophysics, 622, A176, doi: 10.1051/0004-6361/201833036

    2019 doi

  5. [13]

    2016, in Knowledge Discovery and Data Mining, 785–794, doi: 10.1145/2939672.2939785

    Chen, T., & Guestrin, C. 2016, in Knowledge Discovery and Data Mining, 785–794, doi: 10.1145/2939672.2939785

    2016

  6. [14]

    E., Lima, M., Oyaizu, H., Frieman, J., & Lin, H

    Cunha, C. E., Lima, M., Oyaizu, H., Frieman, J., & Lin, H. 2009, Monthly Notices of the Royal Astronomical Society, 396, 2379

    2009

  7. [15]

    B., et al

    Dalmasso, N., Pospisil, T., Lee, A. B., et al. 2020, Astronomy and Computing, 30, 100362 27 DESI Collaboration, Aghamousa, A., Aguilar, J., et al. 2016, arXiv e-prints

    2020

  8. [16]

    H., et al

    Dey, B., Zhao, D., Andrews, B. H., et al. 2025, Machine Learning: Science and Technology, 6, 045058

    2025

  9. [17]

    A., Bovy, J., Myers, A

    DiPompeo, M. A., Bovy, J., Myers, A. D., & Lang, D. 2015, Monthly Notices of the Royal Astronomical Society, 452, 3124–3138, doi: 10.1093/mnras/stv1562

    2015 doi

  10. [18]

    2019, in Neural Information Processing Systems

    Durkan, C., Bekasov, A., Murray, I., & Papamakarios, G. 2019, in Neural Information Processing Systems. https://arxiv.org/abs/1906.04032

    2019 arXiv

  11. [19]

    2023, Annual Review of Astronomy and Astrophysics, 61, 373, doi: 10.1146/annurev-astro-052920-102455

    Fan, X., Ba˜ nados, E., & Simcoe, R. 2023, Annual Review of Astronomy and Astrophysics, 61, 373, doi: 10.1146/annurev-astro-052920-102455

    2023 doi

  12. [20]

    Freeman, P., Izbicki, R., & Lee, A. B. 2017, Monthly notices of the Royal Astronomical Society, 468, 4556, doi: 10.1093/mnras/stx764

    2017 doi

  13. [21]

    2025, arXiv.org, arXiv:2507.03971, doi: 10.48550/arXiv.2507.03971

    Garg, A., Ali, M., Hollmann, N., et al. 2025, arXiv.org, arXiv:2507.03971, doi: 10.48550/arXiv.2507.03971

    2025 doi arXiv

  14. [22]

    2007, Journal of the American Statistical Association, 102, 359, doi: 10.1198/016214506000001437

    Gneiting, T., & Raftery, A. 2007, Journal of the American Statistical Association, 102, 359, doi: 10.1198/016214506000001437

    2007 doi

  15. [23]

    2025, arXiv.org, arXiv:2511.08667, doi: 10.48550/arXiv.2511.08667

    Grinsztajn, L., Floge, K., Key, O., et al. 2025, arXiv.org, arXiv:2511.08667, doi: 10.48550/arXiv.2511.08667

    2025 doi arXiv

  16. [24]

    2022, in Proceedings of Machine Learning Research, Vol

    Gulrajani, I., & Hashimoto, T. 2022, in Proceedings of Machine Learning Research, Vol. 162, Proceedings of the 39th International Conference on Machine Learning, 7982–7997. https://proceedings.mlr.press/v162/gulrajani22a.html

    2022

  17. [25]

    2014, Annual Review of Astronomy and Astrophysics, 52, 589, doi: 10.1146/annurev-astro-081913-035722

    Heckman, T., & Best, P. 2014, Annual Review of Astronomy and Astrophysics, 52, 589, doi: 10.1146/annurev-astro-081913-035722

    2014 doi

  18. [26]

    Herpich, F., Almeida-Fernandes, F., Schwarz, G. B. O., et al. 2024, Astronomy & Astrophysics, 689, A249, doi: 10.1051/0004-6361/202449725

    2024 doi

  19. [27]

    2017, The Astrophysical Journal, 849, 53, doi: 10.3847/1538-4357/aa8c77

    Hickox, R., Myers, A., Greene, J., et al. 2017, The Astrophysical Journal, 849, 53, doi: 10.3847/1538-4357/aa8c77

    2017 doi

  20. [28]

    2010, Astronomy & Astrophysics, 523, A31, doi: 10.1051/0004-6361/201014885

    Hildebrandt, H., Arnouts, S., Capak, P., et al. 2010, Astronomy & Astrophysics, 523, A31, doi: 10.1051/0004-6361/201014885

    2010 doi

  21. [29]

    G., Purucker, L., et al

    Hollmann, N., M¨ uller, S. G., Purucker, L., et al. 2025, Nature, 637, 319, doi: 10.1038/s41586-024-08328-6

    2025 doi

  22. [30]

    2025, Machine Learning Beyond Point Predictions: Uncertainty Quantification, 1st edn., 260

    Izbicki, R. 2025, Machine Learning Beyond Point Predictions: Uncertainty Quantification, 1st edn., 260

    2025

  23. [31]

    2014, in Artificial intelligence and statistics, PMLR, 420–429

    Izbicki, R., Lee, A., & Schafer, C. 2014, in Artificial intelligence and statistics, PMLR, 420–429

    2014

  24. [32]

    Izbicki, R., & Lee, A. B. 2016, Journal of Computational and Graphical Statistics, 25, 1297

    2016

  25. [33]

    Izbicki, R., & Lee, A. B. 2017, Electronic Journal of Statistics, 11, 2800, doi: 10.1214/17-EJS1302

    2017 doi

  26. [34]

    B., & Freeman, P

    Izbicki, R., Lee, A. B., & Freeman, P. 2017, Annals of Applied Statistics, 11, 698, doi: 10.1214/16-AOAS1013

    2017 doi

  27. [35]

    Izbicki, R., & Rodrigues, P. L. C. 2026, arXiv e-prints, arXiv:2603.26611, doi: 10.48550/arXiv.2603.26611

    2026 doi

  28. [36]

    2026, arXiv e-prints, arXiv:2603.08206, doi: 10.48550/arXiv.2603.08206

    Landsgesell, J., & Knoll, P. 2026, arXiv e-prints, arXiv:2603.08206, doi: 10.48550/arXiv.2603.08206

    2026 doi arXiv

  29. [37]

    2025, ErikVini/specz compilation: Southern Hemisphere Spectrocopic Redshift Compilation, 2025.03.27 Zenodo, doi: 10.5281/zenodo.15127060

    Lima, E. 2025, ErikVini/specz compilation: Southern Hemisphere Spectrocopic Redshift Compilation, 2025.03.27 Zenodo, doi: 10.5281/zenodo.15127060

    2025 doi

  30. [38]

    2001, ARA&A, 39, doi: 10.1146/annurev.astro.39.1.19

    Loeb, A., & Barkana, R. 2001, ARA&A, 39, doi: 10.1146/annurev.astro.39.1.19

    2001 doi

  31. [39]

    M., & Lee, S.-I

    Lundberg, S. M., & Lee, S.-I. 2017, in Neural Information Processing Systems, Vol. 30

    2017

  32. [40]

    2008, MNRAS, 386, 781, doi: 10.1111/j.1365-2966.2008.12947.x

    Mandelbaum, R., Seljak, U., Hirata, C., et al. 2008, MNRAS, 386, 781, doi: 10.1111/j.1365-2966.2008.12947.x

    2008

  33. [41]

    C., et al

    Martin, D. C., et al. 2005, ApJL, 619, L1, doi: 10.1086/426387 Mendes de Oliveira, C., Ribeiro, T., Schoenell, W., et al. 2019, Monthly Notices of the Royal Astronomical Society, 489, 241, doi: 10.1093/mnras/stz1985

    2005 doi

  34. [42]

    2007, ApJS, 173, 682, doi: 10.1086/520512

    Morrissey, P., et al. 2007, ApJS, 173, 682, doi: 10.1086/520512

    2007 doi

  35. [43]

    2016, in Astrophysics and Space Science

    Mortlock, D. 2016, in Astrophysics and Space Science

    2016

  36. [44]

    423, Understanding the Epoch of Cosmic Reionization: Challenges and Progress, ed

    Library, Vol. 423, Understanding the Epoch of Cosmic Reionization: Challenges and Progress, ed. A. Mesinger, 187–226, doi: 10.1007/978-3-319-21957-8 7

    doi

  37. [45]

    2024, in Neural Information Processing Systems, Vol

    Muschalik, M., Baniecki, H., Fumagalli, F., et al. 2024, in Neural Information Processing Systems, Vol. 37 (Neural Information Processing Systems Foundation, Inc. (NeurIPS)), 130324–130357, doi: 10.48550/arXiv.2410.01649

    2024 doi arXiv

  38. [46]

    2009, Monthly Notices of the Royal Astronomical Society, 399, 2279, doi: 10.1111/j.1365-2966.2009.15432.x

    Myers, A., White, M., & Ball, N. 2009, Monthly Notices of the Royal Astronomical Society, 399, 2279, doi: 10.1111/j.1365-2966.2009.15432.x

    2009

  39. [47]

    Nakazono, L., Oliveira, C. M. d., Hirata, N., et al. 2021, Monthly notices of the Royal Astronomical Society, 507, 5847, doi: 10.1093/mnras/stab1835

    2021 doi

  40. [48]

    R., Soares, G., et al

    Nakazono, L., Valencca, R. R., Soares, G., et al. 2024, Monthly notices of the Royal Astronomical Society, 531, 327, doi: 10.1093/mnras/stae971

    2024 doi

  41. [49]

    2016, arXiv e-prints

    Polsterer, K., D’Isanto, A., & Gieseke, F. 2016, arXiv e-prints

    2016

  42. [50]

    Qu, J., Holzm¨ uller, D., Varoquaux, G., & Morvan, M. L. 2025, International Conference on Machine Learning, arXiv:2502.05564, doi: 10.48550/arXiv.2502.05564

    2025 doi arXiv

  43. [51]

    2023, Monthly notices of the Royal Astronomical Society, 524, doi: 10.1093/mnras/stad1962

    Rau, M., Dalal, R., Zhang, T., et al. 2023, Monthly notices of the Royal Astronomical Society, 524, doi: 10.1093/mnras/stad1962

    2023 doi

  44. [52]

    T., Fan, X., Schneider, D

    Richards, G. T., Fan, X., Schneider, D. P., et al. 2001, The Astronomical Journal, 121, 2308 28

    2001

  45. [53]

    2024, in World Conference on Explainable Artificial Intelligence, Springer, 465–476

    Rundel, D., Kobialka, J., von Crailsheim, C., et al. 2024, in World Conference on Explainable Artificial Intelligence, Springer, 465–476

    2024

  46. [54]

    2019, Nature Astronomy, 3, 212, doi: 10.1038/s41550-018-0478-0

    Salvato, M., Ilbert, O., & Hoyle, B. 2019, Nature Astronomy, 3, 212, doi: 10.1038/s41550-018-0478-0

    2019 doi

  47. [55]

    F., et al

    Schlafly, E. F., et al. 2019, ApJS, 240, 30, doi: 10.3847/1538-4365/aafbea

    2019 doi

  48. [56]

    2020, Monthly notices of the Royal Astronomical Society, 499, 1587, doi: 10.1093/mnras/staa2799

    Schmidt, S., Malz, A., Soo, J., et al. 2020, Monthly notices of the Royal Astronomical Society, 499, 1587, doi: 10.1093/mnras/staa2799

    2020 doi

  49. [57]

    O., Herpich, F., Almeida-Fernandes, F., et al

    Schwarz, G. O., Herpich, F., Almeida-Fernandes, F., et al. 2025, Astronomy and Computing, 51, 100899, doi: 10.1016/j.ascom.2024.100899

    2025

  50. [58]

    2000, Journal of Statistical Planning and Inference, 90, 227, doi: 10.1016/S0378-3758(00)00115-4

    Shimodaira, H. 2000, Journal of Statistical Planning and Inference, 90, 227, doi: 10.1016/S0378-3758(00)00115-4

    2000 doi

  51. [59]

    J., Benford, D

    Stern, D., Assef, R. J., Benford, D. J., et al. 2012, ApJ, 753, 30, doi: 10.1088/0004-637X/753/1/30

    2012 doi

  52. [60]

    A., Richards, G

    Weinstein, M. A., Richards, G. T., Schneider, D. P., et al. 2004, ApJS, 155, 243, doi: 10.1086/425355

    2004 doi

  53. [61]

    L., et al

    Wright, E. L., et al. 2010, AJ, 140, 1868, doi: 10.1088/0004-6256/140/6/1868

    2010 doi

  54. [62]

    K., Anderson, J., et al

    York, D., Adelman, J. K., Anderson, J., et al. 2000, The Astronomical Journal, 120, 1579, doi: 10.1086/301513

    2000 doi

  55. [63]

    Zhao, D., Dalmasso, N., Izbicki, R., & Lee, A. B. 2021, in Uncertainty in Artificial Intelligence, PMLR, 1830–1840

    2021

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.