The reviewed record of science sign in
Pith

arxiv: 2607.05393 · v1 · pith:WRE72BBX · submitted 2026-07-06 · astro-ph.IM · astro-ph.GA· astro-ph.HE· cs.AI· cs.LG

Interpretable Human-Label-Free Deep Learning for Real-Bogus Classification with Uncertainty Quantification

Rapha\"el Bonnet-Guerrini , Bruno Sanchez , Dominique Fouchez , Benjamin Racine , Maya Guy , Mariam Sabalbal , Manal Yassine , Vincenzo Piuri This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-07-07 12:21 UTCglm-5.2pith:WRE72BBXrecord.jsonopen to challenge →

classification astro-ph.IM astro-ph.GAastro-ph.HEcs.AIcs.LG
keywords real-bogus classificationweakly supervised learningco-teachinguncertainty quantificationtransient detectiondifference image analysislabel noisedeep learning
0
0 comments X

The pith

Train Real-Bogus Classifiers Without Human Labels

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

The paper claims that transient-bogus classification for time-domain astronomical surveys can be trained without any human-labeled data by combining simulated source injections (clean positive examples) with raw survey detections (a noisy negative class dominated by artifacts but containing some real transients). The central mechanism is Asymmetric Co-teaching: two neural networks trained simultaneously, each selecting low-loss samples from the other network for training, but with class-specific discard rates that account for the fact that label noise is concentrated in one class. This setup treats the problem as weakly supervised learning with asymmetric, class-dependent label noise. The paper shows the method remains stable up to 35% contamination of the bogus-labeled class by genuine transients, and proposes a hybrid uncertainty strategy that combines the two co-teaching networks with MC dropout at inference time, achieving calibration competitive with 50-member deep ensembles at a fraction of the training cost.

Core claim

Asym-Co-teaching enables a classifier to separate real transients from bogus detections using only injected simulated sources and unlabeled survey data, maintaining ROC AUC around 0.95 even when over a third of the bogus-labeled training class is secretly real transients. The hybrid uncertainty method (Ensemble-MC-Dropout for Co-teaching) achieves the best calibration among all tested methods (NLL 0.292, ECE 0.0375) while requiring only two trained networks plus dropout sampling at inference, instead of 50 independently trained models. When source-level predictions are grouped by sky position into light-curve-level object classifications, the method achieves 99.35% accuracy on a curated 306-

What carries the argument

Asymmetric Co-teaching: a dual-network training procedure where each network filters training samples for the other, but with separate forget rates per class. The injected-transient class has a low forget rate (preserving nearly all samples) while the survey-bogus class has a higher forget rate (discarding likely-mislabeled examples). This asymmetry matches the noise structure: injections have clean labels, survey detections have corrupted labels.

If this is right

  • Future surveys like LSST could deploy real-bogus classifiers by re-running an injection pipeline on their own data, eliminating the need for survey-specific human-labeled training sets that are expensive and slow to produce.
  • The hybrid uncertainty approach (2 networks + MC dropout) could replace large ensembles in resource-constrained survey pipelines, cutting training cost by roughly 25x while improving calibration.
  • Epoch-level uncertainty from single-image classification could be used to down-weight ambiguous measurements in downstream light-curve fitting, propagating classification confidence into photometric analysis.
  • The asymmetric co-teaching framework could be applied to other weakly supervised problems where one class has reliable labels (from simulation) and the other is a noisy real-world sample.

Where Pith is reading between the lines

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

  • If the transient fraction in raw survey detections rises substantially (as DIA pipelines improve and produce fewer artifacts), the asymmetric noise assumption weakens and the method may need explicit noise-rate estimation rather than fixed forget rates.
  • The finding that the two-network co-teaching pair plus dropout matches 50-member ensembles suggests that training-time coupling between networks may produce more useful functional diversity than independent initialization alone, which has implications for ensemble design beyond astronomy.
  • The latent-space visualization revealing structured subclasses within the bogus population (dipoles, negative residuals, low-SNR artifacts) suggests that the model learns physically meaningful artifact categories without being told they exist, which could inform targeted pipeline improvements.

Load-bearing premise

The method assumes that unlabeled survey detections are overwhelmingly bogus, so they can serve as a noisy negative class. If a large fraction of survey detections are actually real transients, the training labels become too corrupted and the model would learn to suppress genuine signals. The paper tests robustness up to 35% contamination but relies on real surveys having a much lower transient fraction.

What would settle it

If the transient fraction among raw DIA detections exceeds roughly 35%, the asymmetric noise assumption breaks down and the classifier would be trained to mislabel real transients as bogus.

Figures

Figures reproduced from arXiv: 2607.05393 by Benjamin Racine, Bruno Sanchez, Dominique Fouchez, Manal Yassine, Mariam Sabalbal, Maya Guy, Rapha\"el Bonnet-Guerrini, Vincenzo Piuri.

Figure 1
Figure 1. Figure 1: Example of difference imaging analysis (DIA) stamps. Each triplet shows the Science image (left), the Template image (middle), and the Difference (Science−Template=Difference) (right). Top: a real transient produces a compact, PSF-like pos￾itive residual in the subtraction. Bottom: a bogus detection ex￾hibiting a structured residual (dipole-like), characteristic of sub￾traction artifacts. 2.1. Difference I… view at source ↗
Figure 2
Figure 2. Figure 2: Schematic of the CNN architecture selected by hyperparameter optimization. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of sample-selection rules in Co-teaching vari [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Two-dimensional UMAP projection of the embeddings from the penultimate dense layer of the classifier. Each point corre [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The top panel shows the fraction of positive samples as [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Light-curve-level classification score distributions. Shown are the distributions of the fraction of source detections classified [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
read the original abstract

Time-domain surveys generate many transient candidates, making Real-Bogus classification a critical step in automated discovery pipelines. Reliable labels are costly, while community labels can be noisy and survey-dependent. We aim to develop a Real-Bogus classification framework that can be trained without human-labeled data using injected transients and bogus-dominated survey data, remains robust under strong class contamination, and provides calibrated uncertainty quantification. We combine simulated transient injections with a contaminated survey class and train a dual-network model using asymmetric co-teaching for classes with different label-noise levels. We evaluate performance on a benchmark subset and analyze the learned representation with latent-space visualization tools. For uncertainty quantification (UQ), we compare MC dropout and deep ensembles and propose a low-cost hybrid strategy that exploits the dual-network setting to improve calibration. We extend the evaluation to the light-curve domain to assess recovery of light-curve classes. The method achieves strong Real-Bogus performance on the labeled subset and remains stable under severe class contamination. It recovers transient light-curve classes with high fidelity, while single-source identification is limited by ambiguity in light-curve-derived labels. Our hybrid UQ approach achieves competitive calibration relative to more expensive ensemble baselines. Latent-space analyses indicate that uncertainty aligns with the decision boundary and reveal subclasses within the bogus population. Our results show that injection-driven, weakly supervised training can enable scalable and consistent Real-Bogus classification without human-labeled training data while providing calibrated uncertainties. The method is suited for transfer to forthcoming surveys by re-running the injection-based training pipeline.

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 / 8 minor

Summary. This paper presents a human-label-free approach to Real-Bogus classification for transient candidates, using injected supernova-like sources as a clean positive class and raw survey detections as a noisy negative class. The authors introduce Asym-Co-teaching, a class-dependent extension of co-teaching, to handle the asymmetric label noise inherent in this setup. They evaluate robustness under controlled contamination (up to 35%), compare uncertainty quantification methods (MC dropout, deep ensembles, repulsive ensembles, and a proposed hybrid), and provide latent-space visualizations via UMAP. The method is evaluated on a manually curated set of 306 objects (4,820 source detections) from HSC-UDEEP data, and extended to object-level light-curve classification.

Significance. The human-label-free training paradigm is a genuine and timely contribution for the LSST era, where labeled training data is a bottleneck. The asymmetric co-teaching adaptation is well-motivated by the class-dependent noise structure of the problem. The controlled noise experiments (Table 5) provide useful evidence of robustness. The authors provide reproducible code and interactive visualizations on Zenodo, which is commendable. The hybrid UQ strategy (Section 5.5) that reuses the co-teaching dual-network pair is a practical and computationally efficient idea. The latent-space analysis (Section 6, Fig. 4) provides qualitative insight into model behavior and bogus substructure.

major comments (3)
  1. §7.2, Table 6: The UQ comparison reports small margins between methods (e.g., NLL 0.292 vs 0.314 for Deep Ensemble; Brier 0.063 vs 0.065; ECE 0.0375 vs 0.0535) on a single evaluation set of 4,820 sources, with no confidence intervals, bootstrap analysis, or significance tests. The text in §7.2 states the method 'consistently outperforms' ensemble baselines, but without uncertainty estimates on the metrics themselves, this claim is not statistically supported. This is load-bearing for the UQ superiority claim and should be addressed with bootstrap confidence intervals or similar.
  2. §2.3, Table 1: The evaluation set is constructed with aggressive filtering (SNR>5, ≥6 nights, host-galaxy cuts, flux-ratio >1.4) that explicitly removes ambiguous cases. The paper acknowledges this defines 'a high-purity evaluation subset rather than a representative sample of the full alert stream,' but the 99.35% object-level accuracy reported in §7.4 and the UQ calibration results in Table 6 are all measured on this easy subset. Since the paper claims the method is 'suited for transfer to forthcoming surveys' (Abstract, §9), the absence of any evaluation on harder or more representative cases weakens the generalizability evidence. At minimum, the authors should explicitly scope their performance and UQ claims to high-purity SN-like events and discuss how the evaluation bias might affect the relative UQ comparison.
  3. §4.2, Appendix C: The forget rates for Asym-Co-teaching are set using prior knowledge of the noise levels (e.g., (0.05, 0.01) for the baseline). In the controlled-noise experiments (Table 5), this is reasonable since the true contamination is known. However, for the baseline dataset representing real survey data, the 5% noise estimate for the survey class is an assumption whose sensitivity is not tested. The paper acknowledges this in §8 ('noise-rate estimation is an interesting direction for future work'), but the practical applicability of the method to a new survey depends on choosing appropriate forget rates without ground truth. A brief sensitivity analysis showing how performance degrades if the forget rate is misspecified would strengthen the transfer claim.
minor comments (8)
  1. §3.3: The Bayesian optimization is performed on a 30% subset of the data. It would be useful to state whether the selected hyperparameters (Table 3) were validated on the full dataset or held-out data.
  2. Table 5: The 'B' column header is ambiguous — it presumably means 'Bogus specificity' but is not explicitly defined in the caption.
  3. §7.4: The 99.35% accuracy (304/306) is reported without noting that it corresponds to only 2 misclassifications. This should be stated explicitly to give readers proper context on the statistical fragility.
  4. Fig. 4 caption: The cutouts are described as 'approximate with respect to exact UMAP locations.' This is fine for illustration but should be noted more prominently in the figure itself, not just the caption.
  5. §5.1, Eq. (6): The Spearman correlation formula assumes no tied ranks. With discrete or binned data, ties are likely; the standard tie-corrected formula should be used or the assumption noted.
  6. Appendix C: The forget-rate values for Asym-Co-teaching use (forget_rate_0, forget_rate_1) but it is not immediately clear which rate corresponds to which class. This should be stated explicitly.
  7. §2.2: The injection magnitude prior m_inj ~ U(m_host-1, m_host+3) is stated without justification for these particular bounds. A brief motivation would help.
  8. The paper would benefit from a comparison table placing this method alongside existing Real-Bogus approaches (e.g., Reyes et al. 2018, Carrasco-Davis et al. 2021) in terms of training data requirements and performance, even if direct metric comparison is not possible due to different datasets.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for a careful and constructive report. The referee correctly identifies the core contributions of the paper and raises three major comments, all of which are substantive and well-targeted. We address each below. In brief: (1) we agree that bootstrap confidence intervals are needed for the UQ comparison and will add them; (2) we agree that the evaluation-set scoping needs to be made more explicit throughout the paper, including in the abstract and conclusions; (3) we agree that a sensitivity analysis on forget-rate misspecification would strengthen the transfer claim and will add one. We detail our planned revisions for each point.

read point-by-point responses

  1. Referee: §7.2, Table 6: The UQ comparison reports small margins between methods on a single evaluation set of 4,820 sources, with no confidence intervals, bootstrap analysis, or significance tests. The claim that the method 'consistently outperforms' ensemble baselines is not statistically supported.

    Authors: The referee is correct. The margins in Table 6 are small (e.g., NLL 0.292 vs. 0.314 for Deep Ensemble; ECE 0.0375 vs. 0.0535), and without confidence intervals or significance tests, the claim of consistent outperformance is not statistically justified. We will revise the manuscript to include bootstrap confidence intervals (with at least 1,000 resamples) for all metrics in Table 6 (NLL, Brier score, ECE). We will also add paired bootstrap tests to assess whether the differences between our method and each baseline are statistically significant. The text in §7.2 will be revised accordingly: where the confidence intervals overlap, we will describe the result as 'competitive with' rather than 'consistently outperforms,' and we will explicitly state which differences are significant at a given level. We agree this is load-bearing for the UQ claim and the revision is necessary. We note that the ECE difference (0.0375 vs. 0.0535 for Deep Ensemble, a 29.9% relative reduction) is the largest margin and may survive a significance test, but we will let the data determine this rather than asserting it a priori. revision: yes

  2. Referee: §2.3, Table 1: The evaluation set is constructed with aggressive filtering that explicitly removes ambiguous cases. The 99.35% object-level accuracy and UQ calibration results are all measured on this easy subset. The paper claims the method is 'suited for transfer to forthcoming surveys' but the absence of evaluation on harder or more representative cases weakens the generalizability evidence.

    Authors: The referee raises a valid concern about evaluation bias. We agree that the current evaluation set, by construction, favors high-purity SN-like events and that the 99.35% object-level accuracy and the UQ calibration metrics should not be interpreted as representative of performance on the full alert stream. We will make the following revisions: (1) In §2.3, we will add an explicit statement that all quantitative results in the paper apply to the high-purity SN-like evaluation subset and that performance on a more representative or harder sample is not assessed. (2) In the Abstract and §9 (Conclusions), we will scope the transfer claim more carefully, replacing 'suited for transfer to forthcoming surveys' with language that acknowledges the method is transferable in principle (the training pipeline is survey-agnostic) but that performance characterization is limited to the curated high-purity subset. (3) In §7.4, we will add a caveat that the 99.35% accuracy is measured on an easy subset and should not be extrapolated. (4) We will add a paragraph in §8 discussing how the evaluation bias likely affects the relative UQ comparison: since all methods are evaluated on the same subset, the relative ranking may be less sensitive to the bias than the absolute values, but we cannot rule out that harder cases would shift the comparison. We acknowledge that we cannot, within the scope of the current revision, construct a new representative evaluation set, as this would require additional manual labeling or a different evaluation methodology. We will state this limitation explicitly. revision: partial

  3. Referee: §4.2, Appendix C: The forget rates for Asym-Co-teaching are set using prior knowledge of the noise levels. For the baseline dataset, the 5% noise estimate for the survey class is an assumption whose sensitivity is not tested. A sensitivity analysis showing how performance degrades if the forget rate is misspecified would strengthen the transfer claim.

    Authors: This is a fair and constructive suggestion. The practical applicability of Asym-Co-teaching to a new survey does depend on choosing appropriate forget rates without ground truth, and the current manuscript does not test sensitivity to misspecification. We will add a sensitivity analysis in which we train Asym-Co-teaching on the baseline dataset with forget rates that are systematically misspecified (e.g., setting the survey-class forget rate to 0%, 10%, 15%, 20% when the assumed true noise is ~5%, and similarly varying the injected-class forget rate). We will report the resulting accuracy, ROC AUC, and calibration metrics in a new table or figure. This will allow readers to assess how much performance degrades under misspecification and provide practical guidance for applying the method to a new survey. We expect that moderate overestimation of the forget rate will lead to some loss of training signal but should not catastrophically degrade performance, since co-teaching is designed to be robust to the discarded samples being clean rather than noisy; however, we will let the empirical results speak. We will also add a brief discussion in §8 on strategies for estimating forget rates in practice for a new survey (e.g., using training dynamics or a small validation set), while noting that automatic noise-rate estimation remains future work as already stated. revision: yes

Circularity Check

0 steps flagged

No significant circularity found; forget-rate tuning is acknowledged as prior-knowledge-based but is not a fitted-then-predicted circularity pattern.

full rationale

The paper's central claim — human-label-free Real-Bogus classification via injection-driven weakly supervised training — is not circular by construction. The training data (injected transients + survey detections) and the evaluation set (manually curated, used only for testing) are genuinely independent. The Asym-Co-teaching method (Section 4.2, Eqs. 1-2) is a well-defined extension of standard Co-teaching with class-specific forget rates; it is not defined in terms of its own outputs. The hybrid UQ method (Section 5.5, Eqs. 7-13) combines two independently trained networks with MC Dropout at inference — a heuristic mixture approximation, not a self-referential construction. The one potentially circular-adjacent element is the forget-rate setting (Appendix C): the paper states 'We select the forget-rate schedules using prior knowledge estimated from the label noise inferred by the standard training method... we use the observed false-positive and false-negative fractions on the training set as approximate indicators of mislabeling.' This means the forget rates are set from training-set error rates, not from evaluation-set performance, so it is not a fit-then-predict pattern. The paper transparently acknowledges this as prior-knowledge-based tuning rather than presenting it as a first-principles prediction. No self-citation chain is load-bearing: the methodological foundations (Co-teaching from Han et al. 2018, MC Dropout from Gal & Ghahramani 2016, Deep Ensembles from Lakshminarayanan et al. 2017) are all external citations. The evaluation metrics (NLL, Brier, ECE, Spearman correlation) are standard and not defined in terms of the method's own outputs. The derivation chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

7 free parameters · 4 axioms · 0 invented entities

The paper does not invent new physical entities or particles. It introduces a methodological variant (Asym-Co-teaching) and a hybrid UQ strategy, which are algorithmic contributions rather than postulated physical objects. The free parameters are standard ML hyperparameters and injection priors, fitted or chosen as described.

free parameters (7)
  • Base filters F = 64
    Determined by Bayesian hyperparameter optimization (Optuna) on a 30% subset.
  • Dense units = 64
    Determined by Bayesian hyperparameter optimization.
  • Base dropout rate B = 0.25
    Determined by Bayesian hyperparameter optimization.
  • Learning rate = 1.62e-4
    Determined by Bayesian hyperparameter optimization.
  • Forget rates (r_a, r_b) = e.g., (0.05, 0.01) for baseline
    Set using prior knowledge of label noise levels estimated from training set false-positive and false-negative fractions (Appendix C).
  • Injection magnitude prior = U(m_host-1, m_host+3)
    Chosen to span a range of source-to-host contrast ratios.
  • Injection offset prior = N(0, a^2)
    Variance set by host galaxy semi-major axis.
axioms (4)
  • domain assumption Unlabeled survey detections are dominated by bogus examples.
    Invoked in Section 2.2 and Section 4 to justify treating survey detections as a noisy bogus class during training.
  • standard math Small-loss samples are more likely to be correctly labeled than high-loss samples.
    Inherited from the Co-teaching framework (Han et al. 2018), invoked in Section 4.1.
  • standard math MC Dropout approximates Bayesian inference.
    Invoked in Section 5.4 to justify using dropout at inference for uncertainty estimation.
  • domain assumption The evaluation set is representative enough to assess model performance.
    The evaluation set is constructed with aggressive filtering for high-purity SN-like candidates (Section 2.3), which may not represent the full alert stream.

pith-pipeline@v1.1.0-glm · 29119 in / 2484 out tokens · 144028 ms · 2026-07-07T12:21:04.758235+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

57 extracted references · 57 canonical work pages · 2 internal anchors

  1. [1]

    A., Allison, J., Anderson, S

    Abell, P. A., Allison, J., Anderson, S. F., et al. 2009, LSST Science Book, Version 2.0, Tech. rep., LSST Science Collaboration, 596 pages. Also available at full resolution at http://www.lsst.org/lsst/scibook

    work page 2009

  2. [2]

    2019, Optuna: A Next- generation Hyperparameter Optimization Framework

    Akiba, T., Sano, S., Yanase, T., Ohta, T., & Koyama, M. 2019, Optuna: A Next- generation Hyperparameter Optimization Framework

    work page 2019

  3. [3]

    2000, A&AS, 144, 363

    Alard, C. 2000, A&AS, 144, 363

    work page 2000

  4. [4]

    & Lupton, R

    Alard, C. & Lupton, R. H. 1998, The Astrophysical Journal, 503, 325–331

    work page 1998

  5. [5]

    2017, in Proceedings of Machine Learning Research, V ol

    Arpit, D., Jastrz˛ ebski, S., Ballas, N., et al. 2017, in Proceedings of Machine Learning Research, V ol. 70, Proceedings of the 34th International Conference on Machine Learning, ed. D. Precup & Y . W. Teh (PMLR), 233–242

    work page 2017

  6. [6]

    2011, in Advances in Neural Information Processing Systems, ed

    Bergstra, J., Bardenet, R., Bengio, Y ., & Kégl, B. 2011, in Advances in Neural Information Processing Systems, ed. J. Shawe-Taylor, R. Zemel, P. Bartlett, F. Pereira, & K. Weinberger, V ol. 24 (Curran Associates, Inc.)

    work page 2011

  7. [7]

    2016, Layer- wise Relevance Propagation for Neural Networks with Local Renormaliza- tion Layers

    Binder, A., Montavon, G., Bach, S., Müller, K.-R., & Samek, W. 2016, Layer- wise Relevance Propagation for Neural Networks with Local Renormaliza- tion Layers

    work page 2016

  8. [8]

    L., Rykoff, E

    Burke, D. L., Rykoff, E. S., Allam, S., et al. 2017, The Astronomical Journal, 155, 41

    work page 2017

  9. [9]

    2023, Domain Adaptation via Minimax Entropy for Real/Bogus Classification of Astronomical Alerts

    Cabrera-Vives, G., Bolivar, C., Förster, F., et al. 2023, Domain Adaptation via Minimax Entropy for Real/Bogus Classification of Astronomical Alerts

    work page 2023

  10. [10]

    2020, Distill, https://distill.pub/2020/circuits

    Cammarata, N., Carter, S., Goh, G., et al. 2020, Distill, https://distill.pub/2020/circuits

    work page 2020

  11. [11]

    2021, The Astronomical Journal, 162, 231

    Carrasco-Davis, R., Reyes, E., Valenzuela, C., et al. 2021, The Astronomical Journal, 162, 231

    work page 2021

  12. [12]

    2023, TransientViT: A novel CNN - Vision Transformer hybrid real/bogus transient classifier for the Kilodegree Auto- matic Transient Survey D’Angelo, F

    Chen, Z., Zhou, W., Sun, G., et al. 2023, TransientViT: A novel CNN - Vision Transformer hybrid real/bogus transient classifier for the Kilodegree Auto- matic Transient Survey D’Angelo, F. & Fortuin, V . 2021, in Proceedings of the 35th International Con- ference on Neural Information Processing Systems, NIPS ’21 (Red Hook, NY , USA: Curran Associates Inc...

    work page 2023

  13. [13]

    2026, Opportunities in AI/ML for the Rubin LSST Dark Energy Science Collaboration

    DESC, L., Aubourg, E., Avestruz, C., et al. 2026, Opportunities in AI/ML for the Rubin LSST Dark Energy Science Collaboration

    work page 2026

  14. [14]

    W., & Dambre, J

    Dieleman, S., Willett, K. W., & Dambre, J. 2015, Monthly Notices of the Royal Astronomical Society, 450, 1441–1459

    work page 2015

  15. [15]

    2022, Toy Models of Superposition

    Elhage, N., Hume, T., Olsson, C., et al. 2022, Toy Models of Superposition

    work page 2022

  16. [16]

    2021, Transformer Circuits Thread, https://transformer-circuits.pub/2021/framework/index.html

    Elhage, N., Nanda, N., Olsson, C., et al. 2021, Transformer Circuits Thread, https://transformer-circuits.pub/2021/framework/index.html

    work page 2021

  17. [17]

    2020, Deep Ensembles: A Loss Land- scape Perspective

    Fort, S., Hu, H., & Lakshminarayanan, B. 2020, Deep Ensembles: A Loss Land- scape Perspective

    work page 2020

  18. [18]

    & Ghahramani, Z

    Gal, Y . & Ghahramani, Z. 2016, Dropout as a Bayesian Approximation: Repre- senting Model Uncertainty in Deep Learning

    work page 2016

  19. [19]

    2022, Monthly Notices of the Royal As- tronomical Society, 513, 1742

    Goode, S., Cooke, J., Zhang, J., et al. 2022, Monthly Notices of the Royal As- tronomical Society, 513, 1742

    work page 2022

  20. [20]

    L., Bellm, E

    Graham, M. L., Bellm, E. C., Guy, L. P., et al. 2024, Rubin Observatory Technical Note DMTN-102

    work page 2024

  21. [21]

    Guo, C., Pleiss, G., Sun, Y ., & Weinberger, K. Q. 2017, On Calibration of Mod- ern Neural Networks

    work page 2017

  22. [22]

    2018, Co-teaching: Robust Training of Deep Neural Networks with Extremely Noisy Labels

    Han, B., Yao, Q., Yu, X., et al. 2018, Co-teaching: Robust Training of Deep Neural Networks with Extremely Noisy Labels

    work page 2018

  23. [23]

    2012, Advances in Neural Information Processing Systems, 1097

    Hinton, G., Krizhevsky, A., Sutskever, I., & Rachmad, Y . 2012, Advances in Neural Information Processing Systems, 1097

    work page 2012

  24. [24]

    2021, Experimental Astronomy, 51, 319–344

    Hosenie, Z., Bloemen, S., Groot, P., et al. 2021, Experimental Astronomy, 51, 319–344

    work page 2021

  25. [25]

    2022, Photometric calibration methods for wide-field photometric surveys Ivezi´c, Ž., Kahn, S

    Huang, B., Xiao, K., & Yuan, H. 2022, Photometric calibration methods for wide-field photometric surveys Ivezi´c, Ž., Kahn, S. M., Tyson, J. A., et al. 2019, ApJ, 873, 111

    work page 2022

  26. [26]

    2015, The Astronomical Journal, 150, 172

    Kessler, R., Marriner, J., Childress, M., et al. 2015, The Astronomical Journal, 150, 172

    work page 2015

  27. [27]

    L., Lyman, J., Steeghs, D., et al

    Killestein, T. L., Lyman, J., Steeghs, D., et al. 2021, Monthly Notices of the Royal Astronomical Society, 503, 4838–4854

    work page 2021

  28. [28]

    2018, Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCA V)

    Kim, B., Wattenberg, M., Gilmer, J., et al. 2018, Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCA V)

    work page 2018

  29. [29]

    Kulkarni, S. R. & Kasliwal, M. M. 2009, Transients in the Local Universe

    work page 2009

  30. [30]

    2017, Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles

    Lakshminarayanan, B., Pritzel, A., & Blundell, C. 2017, Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles

    work page 2017

  31. [31]

    2017, Monthly Notices of the Royal Astronom- ical Society, 473, 3895

    Lanusse, F., Ma, Q., Li, N., et al. 2017, Monthly Notices of the Royal Astronom- ical Society, 473, 3895

    work page 2017

  32. [32]

    S., et al

    LeCun, Y ., Boser, B., Denker, J. S., et al. 1989, Neural computation, 1, 541

    work page 1989

  33. [33]

    2010, Monthly Notices of the Royal Astronomical Society, 410, 166–178

    Lintott, C., Schawinski, K., Bamford, S., et al. 2010, Monthly Notices of the Royal Astronomical Society, 410, 166–178

    work page 2010

  34. [34]

    Lipton, Z. C. 2016, CoRR, abs/1606.03490 [1606.03490]

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  35. [35]

    M., Armstrong, R., et al

    Liu, S., Wood-Vasey, W. M., Armstrong, R., et al. 2024, The Astrophysical Jour- nal, 967, 10

    work page 2024

  36. [36]

    LSST Science Book, Version 2.0

    Liu, Y ., Fan, L., Hu, L., et al. 2025, Astronomy & Astrophysics, 693, A105 LSST Science Collaboration, Abell, P. A., Allison, J., et al. 2009, arXiv e-prints, arXiv:0912.0201

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  37. [37]

    & Lee, S.-I

    Lundberg, S. & Lee, S.-I. 2017, A Unified Approach to Interpreting Model Pre- dictions

    work page 2017

  38. [38]

    MacKay, D. J. 1992, Neural Computation, 4, 448

    work page 1992

  39. [39]

    2022, Astronomy & Astrophysics, 664, A81

    Makhlouf, K., Turpin, D., Corre, D., et al. 2022, Astronomy & Astrophysics, 664, A81

    work page 2022

  40. [40]

    2013, Spectroscopy in the Era of LSST

    Matheson, T., Fan, X., Green, R., et al. 2013, Spectroscopy in the Era of LSST

    work page 2013

  41. [41]

    2020, UMAP: Uniform Manifold Approx- imation and Projection for Dimension Reduction

    McInnes, L., Healy, J., & Melville, J. 2020, UMAP: Uniform Manifold Approx- imation and Projection for Dimension Reduction

    work page 2020

  42. [42]

    2017, Publications of the Astronomical Society of Japan, 70, S1

    Miyazaki, S., Komiyama, Y ., Kawanomoto, S., et al. 2017, Publications of the Astronomical Society of Japan, 70, S1

    work page 2017

  43. [43]

    L., et al

    Mong, Y .-L., Ackley, K., Killestein, T. L., et al. 2022, Monthly Notices of the Royal Astronomical Society, 518, 752–762 Möller, A., Peloton, J., Ishida, E. E. O., et al. 2020, Monthly Notices of the Royal Astronomical Society, 501, 3272–3288

    work page 2022

  44. [44]

    P., et al

    Najita, J., Willman, B., Finkbeiner, D. P., et al. 2016, Maximizing Science in the Era of LSST: A Community-Based Study of Needed US Capabilities

    work page 2016

  45. [45]

    D., et al

    Narayan, G., Zaidi, T., Soraisam, M. D., et al. 2018, The Astrophysical Journal Supplement Series, 236, 9

    work page 2018

  46. [46]

    & Caruana, R

    Niculescu-Mizil, A. & Caruana, R. 2005, in Proceedings of the 22nd Interna- tional Conference on Machine Learning, ICML ’05 (New York, NY , USA: Association for Computing Machinery), 625–632

    work page 2005

  47. [47]

    2017, Distill, https://distill.pub/2017/feature-visualization

    Olah, C., Mordvintsev, A., & Schubert, L. 2017, Distill, https://distill.pub/2017/feature-visualization

    work page 2017

  48. [48]

    J., Finkbeiner, D

    Padmanabhan, N., Schlegel, D. J., Finkbeiner, D. P., et al. 2008, The Astrophys- ical Journal, 674, 1217–1233

    work page 2008

  49. [49]

    A., Reyes, I., et al

    Reyes, E., Estevez, P. A., Reyes, I., et al. 2018, in 2018 International Joint Con- ference on Neural Networks (IJCNN) (IEEE), 1–8

    work page 2018

  50. [50]

    Why Should I Trust You?

    Ribeiro, M. T., Singh, S., & Guestrin, C. 2016, "Why Should I Trust You?": Explaining the Predictions of Any Classifier

    work page 2016

  51. [51]

    2017, in Advances in Neural Information Processing Systems Article number, page 14 Bonnet-Guerrini et al.: Interpretable Human-Label-Free Real-Bogus with UQ

    Rolnick, D., Veit, A., Belongie, S., & Shavit, N. 2017, in Advances in Neural Information Processing Systems Article number, page 14 Bonnet-Guerrini et al.: Interpretable Human-Label-Free Real-Bogus with UQ

    work page 2017

  52. [52]

    & Mahabal, A

    Sedaghat, N. & Mahabal, A. 2018, Monthly Notices of the Royal Astronomical Society, 476, 5365–5376

    work page 2018

  53. [53]

    2025, Open Problems in Mechanistic Interpretability

    Sharkey, L., Chughtai, B., Batson, J., et al. 2025, Open Problems in Mechanistic Interpretability

    work page 2025

  54. [54]

    2017, Not Just a Black Box: Learning Important Features Through Propagating Activation Differences Sánchez, B

    Shrikumar, A., Greenside, P., Shcherbina, A., & Kundaje, A. 2017, Not Just a Black Box: Learning Important Features Through Propagating Activation Differences Sánchez, B. O., Kessler, R., Scolnic, D., et al. 2022, The Astrophysical Journal, 934, 96 van der Maaten, L. & Hinton, G. 2008, Journal of Machine Learning Research, 9, 2579 Vujˇci´c, V ., Sre´ckovi...

    work page 2017

  55. [55]

    1987, Chemometrics and Intelligent Lab- oratory Systems, 2, 37, proceedings of the Multivariate Statistical Workshop for Geologists and Geochemists

    Wold, S., Esbensen, K., & Geladi, P. 1987, Chemometrics and Intelligent Lab- oratory Systems, 2, 37, proceedings of the Multivariate Statistical Workshop for Geologists and Geochemists

    work page 1987

  56. [56]

    Wu, J. F. & Walmsley, M. 2025, Re-envisioning Euclid Galaxy Morphology: Identifying and Interpreting Features with Sparse Autoencoders

    work page 2025

  57. [57]

    prediction

    Zeiler, M. D. & Fergus, R. 2013, Visualizing and Understanding Convolutional Networks Article number, page 15 A&A proofs:manuscript no. main Appendix A: Injection magnitude We compare the magnitude distributions of injected sources, re- covered injections, and real detections with positive flux. As ex- pected, the recovered injections are systematically b...

    work page doi:10.5281/zenodo.18434676 2013