Pith. sign in

REVIEW 2 major objections 5 minor 300 references

Supernova subtype classification stays accurate down to spectral resolution Rλ=50 and SNR=5, and is only mildly hurt even at Rλ=25.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-12 01:48 UTC pith:W3KG4IFN

load-bearing objection Solid empirical map of SN subtype classification vs R and SNR; the R=50/SNR=5 thresholds are usable but rest on a custom, hand-curated line SNR that is the main soft spot. the 2 major comments →

arxiv 2607.03532 v1 pith:W3KG4IFN submitted 2026-07-03 astro-ph.IM astro-ph.HEastro-ph.SR

How Low Can We Go? Minimum Spectroscopic Requirements For Supernova Subtype Classification

Willow Fox Fortino , Federica B. Bianco , Maryam Modjaz , Thomas Matheson , Umer Zubair This is my paper
classification astro-ph.IM astro-ph.HEastro-ph.SR
keywords supernova classificationspectral resolutionsignal-to-noise ratioLSST follow-upstripped-envelope supernovaedeep learningspectroscopic requirementsABC-SN
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Upcoming sky surveys will find millions of supernovae, far more than existing spectrographs can follow up at high quality. This paper measures how far spectral resolution and signal-to-noise can be lowered before an automated classifier can no longer tell the main supernova subtypes apart. Using a curated library of spectra and a deep-learning model, the authors create homogeneous datasets across many combinations of resolution and noise, then retrain and score the classifier on each one. They find that a refined taxonomy (including subtypes of stripped-envelope events) remains fully usable down to Rλ=50 and SNR=5, with only modest losses at Rλ=25. The practical payoff is clear: observers can free high-resolution instruments for rare or high-value targets, and smaller facilities can still contribute useful classifications by accepting lower resolution or shorter exposures.

Core claim

Classification of supernova spectra into a refined ten-subtype taxonomy is possible at low resolution and low SNR with no loss of model performance down to Rλ=50 and SNR=5; performance is only minimally reduced even at Rλ=25, and degrades rapidly only below Rλ≈20.

What carries the argument

A subtype- and phase-specific SNR definition that measures signal from a single emblematic spectral feature (e.g., Si II λ6355 for early Ia, He I λ5876 for Ib/Ibn) relative to a local pseudo-continuum, then injects controlled Gaussian noise and Gaussian-convolves the spectrum to any target Rλ before retraining the ABC-SN attention classifier.

Load-bearing premise

The custom line-based SNR measure, which needed heavy manual smoothing choices and the removal of roughly a tenth of the spectra, is assumed to give a fair, homogeneous ranking of classification difficulty across every subtype and every resolution.

What would settle it

Retrain the same classifier on an independent library that already has published uncertainty arrays (so SNR can be measured without manual Gaussian smoothing) and check whether the macro-F1 still stays flat down to Rλ=50 and SNR=5.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Spectrographs can deliberately trade resolution or exposure time for classification work without losing refined subtype purity.
  • High-resolution instruments can be reserved for detailed follow-up of rare or scientifically critical events rather than routine typing.
  • Smaller telescopes and lower-cost spectrographs become viable partners for LSST-scale classification campaigns.
  • Instrument designers can target Rλ~50 as a practical floor for classification-mode modes rather than pushing for higher resolving power.
  • Survey planners can set exposure-time calculators knowing that SNR~5 is already sufficient under this taxonomy.

Where Pith is reading between the lines

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

  • The same floor may apply to other modern spectral classifiers, not only the attention model used here, because the information content of the lines themselves is what is being degraded.
  • Including rare classes such as IIn (narrow lines) would likely push the useful resolution floor higher, so the present numbers are best read as optimistic for the included taxonomy.
  • A production pipeline that trains on mixed real-world SNR and resolution rather than uniform synthetic grids may still inherit the same practical thresholds if the median of the training set sits near SNR~20.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper systematically measures how supernova subtype classification performance depends on spectral resolution R_λ and signal-to-noise ratio (SNR). Using a curated SNID-derived library of ten subtypes (Ia-norm, Ia-91T, Ia-91bg, Iax, Ib-norm, Ibn, IIb, Ic-norm, Ic-broad, IIP), the authors define a line-based, width-normalized SNR (Eq. 2 and Table 3), degrade the spectra to a grid of 16 SNR values and 14 resolutions, retrain the attention-based classifier ABC-SN on each of 476 datasets (2-fold CV that keeps all spectra of a given SN in one fold), and report macro-F1 heat-maps (Figs. 6–7). They conclude that refined subtype classification remains essentially undiminished down to R_λ = 50 and SNR = 5, and is only mildly degraded to R_λ = 25.

Significance. If the thresholds hold under the authors’ SNR definition, the result is immediately useful for LSST-era follow-up strategy and for the design of low-cost classification spectrographs. The work is the first systematic R_λ–SNR grid for a refined SN taxonomy, is fully reproducible (public GitHub repository with notebooks that regenerate the figures), and employs careful leakage control and macro-F1 reporting that correctly handle class imbalance. These strengths make the paper a concrete reference for observers and instrument builders even if the precise numerical thresholds later shift under alternative SNR conventions.

major comments (2)
  1. The central numerical claim (no loss to R_λ=50 / SNR=5; only minimal impact to R_λ=25) is defined exclusively by the custom, subtype- and phase-specific line SNR of §3.2–3.3 and Table 3. S is the width-normalized area of one hand-chosen diagnostic feature after Gaussian smoothing whose σ was manually tuned per spectrum; ~10 % of the library (190+179 spectra) was discarded because the procedure failed or produced outliers. At low R the fixed shoulder wavelengths become unreliable (authors note this explicitly), yet the grid still reports “SNR=5”. Because every cell of Figs. 6–7 is generated from this definition, any systematic bias in how it ranks classification difficulty across subtypes or resolutions directly shifts the claimed thresholds. The paper should either (i) re-run a subset of the grid with at least one alternative SNR estimator (e.g., continuum rms in a fixed line-free window
  2. The noise model used to reach target SNR (§3.3) is additive white Gaussian noise scaled to the measured S and added to the extracted signal. Real SN spectra are dominated by Poisson statistics, wavelength-dependent sky, host-galaxy continuum, and residual tellurics. While the idealized model is a reasonable first step, the paper should quantify (or at least discuss with a small controlled experiment) whether the performance cliff moves when more realistic noise is injected. Without that check, the claim that “SNR=5 is sufficient” risks being optimistic for actual observing conditions.
minor comments (5)
  1. Abstract and §5 state “no loss … down to R_λ=50 and SNR=5” while Fig. 6 already shows a few-percent drop at R=50 for several SNR rows; “no statistically significant loss” or “within the scatter of the original-SNR row” would be more precise.
  2. Table 2 and the appendix tables list removed spectra, but the text never states the final number of unique SNe remaining after both culling steps; a single sentence would help readers assess residual class imbalance.
  3. Fig. 1 caption says spectra were normalized to [0,1] for display while ABC-SN trains on standardized data; a brief reminder in the main text would avoid confusion when comparing panels.
  4. The wavelength cut 4500–7000 Å excludes the O I 7774 and Ca NIR triplet that are often decisive for late-time SESNe; a short paragraph on how this restriction may affect the low-R thresholds would strengthen the discussion.
  5. Minor typos: “W e” in the title page, “dificult” throughout, and “for arbitrary \SNR{}” in the abstract.

Circularity Check

1 steps flagged

Empirical grid of retrained ABC-SN performance on independently degraded SNID spectra; thresholds are measured, not derived by construction from inputs.

specific steps
  1. self citation load bearing [§1.3, §2, §3.5, Abstract]
    "we tested the classification performance of a recently developed deep-learning classifier, ABC-SN, on each R_λ and SNR combination. ... This model has demonstrated state-of-the-art performance at R_λ = 100 , and we have complete control and understanding of its structure and performance since it was developed by our group. ... we have re-trained it on the dataset at the original SNR and at all 14 R_λ values. Its (unchanged) performance is included in Figure 6."

    ABC-SN is the authors' own prior model (F26). Its architecture and original performance are taken as the baseline classifier whose degradation response is measured. This is ordinary self-citation of a tool, not a load-bearing uniqueness claim or a reduction of the new thresholds to the prior paper; the thresholds themselves are new empirical F1 values on the degraded grid. Flagged only as the single mild self-reference.

full rationale

The paper's central claim (no performance loss down to R_λ=50 and SNR=5, minimal impact to R_λ=25) is an empirical measurement: spectra from an external SNID-derived library are degraded by Gaussian convolution (resolution) and additive Gaussian noise scaled to a measured S (SNR), then ABC-SN is fully retrained and evaluated via macro F1 on held-out folds for each of 476 (R, SNR) cells. Labels are the original external SNID classifications; the degradation operators do not embed subtype labels or the final F1 values. The custom line-based SNR definition (Table 3, §3.2–3.3) is a methodological choice that ranks difficulty and required manual culling, but it is not circular: S is computed from feature area relative to a pseudo-continuum before any classification, and the reported thresholds simply report the F1 surface under that definition. The only mild self-reference is reuse of the authors' own ABC-SN architecture (F26), which is re-validated on the original-SNR row and is not load-bearing for the new grid. No equation reduces a claimed prediction to a fitted input, no uniqueness theorem is imported, and no ansatz is smuggled. Score 1 reflects only the ordinary self-citation of the classifier; the derivation chain is self-contained and non-circular.

Axiom & Free-Parameter Ledger

3 free parameters · 3 axioms · 1 invented entities

The central claim rests on (1) the representativeness of the SNID-derived spectral library after aggressive cleaning, (2) the validity of the authors’ line-based SNR definition as a fair difficulty metric, and (3) the assumption that ABC-SN’s performance surface generalizes to other classifiers and to real mixed-SNR observing conditions. No free physical constants are fitted; the free choices are algorithmic (feature windows, smoothing ranges, outlier thresholds).

free parameters (3)
  • Gaussian smoothing σ range for signal extraction
    Chosen per spectrum by manual review in the interval 10–50 Å; directly affects the measured S and therefore the entire SNR scale.
  • Feature wavelength windows and shoulder locations (Table 3)
    Subtype- and phase-specific line choices (Si II 6355, He I 5876, Hα, Na I D/S blend, Fe lines, Blend 5855) are selected by eye and literature precedent; different choices would rescale SNR.
  • 95th-percentile SNR outlier cut
    179 spectra removed after re-measurement at SNR_new=100; the percentile threshold is an ad-hoc cleaning parameter.
axioms (3)
  • domain assumption The SNID template library, after continuum removal and rebinning to R=738, constitutes a representative and correctly labeled training distribution for the ten subtypes.
    All labels and the bulk of the spectra originate from SNID; any systematic misclassification or selection bias in that library propagates directly into the performance maps.
  • ad hoc to paper Additive white Gaussian noise scaled to the measured S produces a realistic SNR degradation that preserves the relative difficulty ranking of subtypes.
    Real noise is wavelength-dependent and includes sky, host, and instrumental components; the paper’s pure Gaussian model is a simplifying assumption required for automation.
  • domain assumption ABC-SN’s macro-F1 surface is a faithful proxy for the classification power of any competent modern spectral classifier.
    Only one architecture is tested; the claim that the thresholds are instrument- and model-agnostic rests on this untested generalization.
invented entities (1)
  • Line-based, width-normalized SNR definition (Eq. 2 + Table 3) no independent evidence
    purpose: To create a homogeneous, subtype-aware SNR scale that permits fair comparison across the 476 degraded datasets.
    No community-standard SNR definition for SN spectra exists; the authors construct one from pseudo-continuum areas of hand-chosen features. Independent evidence is limited to internal consistency checks.

pith-pipeline@v1.1.0-grok45 · 31525 in / 2800 out tokens · 25757 ms · 2026-07-12T01:48:40.121488+00:00 · methodology

0 comments
read the original abstract

Millions of supernovae will be discovered with the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST). As a result, spectrographs around the world will have to make difficult decisions about which supernova candidates receive spectroscopic follow-ups. This work identifies the minimum spectral resolution, $R_{\lambda} = \frac{\lambda}{\Delta \lambda}$, as a function of signal-to-noise ratio (SNR) at which spectral classification of supernova subtypes becomes impossible. We include supernova types Ia, Ia-91T, Ia-91bg, Iax, Ib, Ic, broad-lined Ic, IIb, IIP, and Ibn in this work. We produce a definition of SNR based on specific lines for each SN subtype that allows us to generate homogeneous datasets at 16 different values of $R_{\lambda}$ and 14 different SNR's and we tested the classification performance of a recently developed deep-learning classifier, ABC-SN, on each $R_{\lambda}$ and SNR combination. We find that classification of supernova spectra into a refined taxonomy that separates, for example, between different subtypes of stripped envelope supernovae, is possible at low resolution and low SNR with no loss in model performance down to $R_{\lambda} = 50$ and $\text{SNR} = 5$. Classification performance is only minimally impacted even as low as $R_{\lambda} = 25$. We hope that astronomers using the LSST alert stream, as well as designers of future instruments and observatories, will benefit from knowing what spectral resolution is necessary to classify a supernova for arbitrary \SNR{}.

Figures

Figures reproduced from arXiv: 2607.03532 by Federica B. Bianco, Maryam Modjaz, Thomas Matheson, Umer Zubair, Willow Fox Fortino.

Figure 1
Figure 1. Figure 1: Type Ic Supernova 1990b at 4 days after peak brightness shown in four different spectral resolutions. Rλ = 738, the resolution of the data as we retrieved it from SNID templates. Rλ = 100, the resolution of the SEDM and the resolution that the original ABC-SN was trained at. Rλ = 50, the lowest resolution at which we measure no ABC-SN per￾formance loss for any SNR. Rλ = 25, below this resolution classifica… view at source ↗
Figure 2
Figure 2. Figure 2: An example spectrum is shown for each feature that we measure for each subtype. The extracted signal is plotted in orange on top of the spectrum in blue. The black dotted vertical line shows the extremum corresponding to the observed feature. The blue and red vertical lines correspond to the shoulders of the feature, and the blue and red shaded regions correspond to the wavelengths where the noise, N, is e… view at source ↗
Figure 3
Figure 3. Figure 3: A stacked histogram of spectrum SNR is shown for 3395 spectra. The median SNR is ≈ 20. Left: the spectra are stacked based on their main type: Ia, Ib, Ic, II. Right: the spectra are stacked based on their subtype. feature, we also penalize S by the overall width, such that the measured SNR in both cases does not differ substantially. To measure the noise N, we compute the standard deviation of the data poi… view at source ↗
Figure 4
Figure 4. Figure 4: Example spectra are plotted for each subtype. The left and right columns show spectra with an original SNR of ≈ 25 and ≈ 2.5, respectively [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Supernova 1990b (Type Ic-norm, same as in [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The macro F1−score of ABC-SN is plotted on the color-axis while the SNR and Rλ of the dataset it was trained on are shown on the vertical and horizontal axes. We perform a 2-fold cross-validation, so each cell corresponds to the average macro F1−score of two ABC-SN models. The top row shows ABC-SN performance when the SNR of the dataset isn’t changed at all. This represents a validation and extension of mo… view at source ↗
Figure 7
Figure 7. Figure 7: An abridged version of [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗

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

300 extracted references · 84 canonical work pages · 11 internal anchors

  1. [1]

    and Han, Zhanwen , month = jul, year =

    Liu, Zheng-Wei and Röpke, Friedrich K. and Han, Zhanwen , month = jul, year =. Type. Research in Astronomy and Astrophysics , publisher =. doi:10.1088/1674-4527/acd89e , abstract =

    doi:10.1088/1674-4527/acd89e

  2. [2]

    Modeling

    Williamson, Marc and Kerzendorf, Wolfgang and Modjaz, Maryam , month = feb, year =. Modeling. The Astrophysical Journal , publisher =. doi:10.3847/1538-4357/abd244 , abstract =

    doi:10.3847/1538-4357/abd244

  3. [3]

    and Inserra, C

    Chen, T.-W. and Inserra, C. and Fraser, M. and Moriya, T. J. and Schady, P. and Schweyer, T. and Filippenko, A. V. and Perley, D. A. and Ruiter, A. J. and Seitenzahl, I. and Sollerman, J. and Taddia, F. and Anderson, J. P. and Foley, R. J. and Jerkstrand, A. and Ngeow, C.-C. and Pan, Y.-C. and Pastorello, A. and Points, S. and Smartt, S. J. and Smith, K. ...

    doi:10.3847/2041-8213/aaeb2e 2041

  4. [4]

    Multiwavelength

    Chandra, Poonam , month = nov, year =. Multiwavelength. Universe , publisher =. doi:10.3390/universe11110363 , abstract =

    work page doi:10.3390/universe11110363

  5. [5]

    Annual Review of Astronomy and Astrophysics , author =

    The. Annual Review of Astronomy and Astrophysics , author =. 2006 , note =. doi:10.1146/annurev.astro.43.072103.150558 , abstract =

    Pith/arXiv arXiv doi:10.1146/annurev.astro.43.072103.150558 2006

  6. [6]

    and Danziger, I

    Iwamoto, Koichi and Nakamura, Takayoshi and Nomoto, Ken’ichi and Mazzali, Paolo A. and Danziger, I. John and Garnavich, Peter and Kirshner, Robert and Jha, Saurabh and Balam, David and Thorstensen, John , month = may, year =. The. The Astrophysical Journal , publisher =. doi:10.1086/308761 , abstract =

    doi:10.1086/308761

  7. [7]

    Generative

    Taback, Nathan , month = apr, year =. Generative. doi:10.48550/arXiv.2604.02238 , abstract =

    work page doi:10.48550/arxiv.2604.02238

  8. [8]

    Computer

    Beale, Russell , month = jul, year =. Computer. doi:10.48550/arXiv.2507.02183 , abstract =

    Pith/arXiv arXiv doi:10.48550/arxiv.2507.02183

  9. [9]

    curse of knowledge

    Froyd, Jeff and Layne, Jean , month = oct, year =. Faculty development strategies for overcoming the “curse of knowledge” , issn =. 2008 38th. doi:10.1109/FIE.2008.4720529 , abstract =

    work page doi:10.1109/fie.2008.4720529 2008

  10. [10]

    Cureus , author =

    Heutagogy:. Cureus , author =. doi:10.7759/cureus.89731 , abstract =

    work page doi:10.7759/cureus.89731

  11. [11]

    Teaching Statistics , author =

    The curse of knowledge when teaching statistics , volume =. Teaching Statistics , author =. 2023 , note =. doi:10.1111/test.12320 , abstract =

    work page doi:10.1111/test.12320 2023

  12. [12]

    Journal of Career and Technical Education , author =

    Heutagogy in. Journal of Career and Technical Education , author =. doi:10.21061/jcte.500 , abstract =

    work page doi:10.21061/jcte.500

  13. [13]

    American Journal of Physics , author =

    Self‐directed learning:. American Journal of Physics , author =. 1995 , pages =. doi:10.1119/1.18080 , abstract =

    work page doi:10.1119/1.18080 1995

  14. [14]

    Cooperative

    Mesghina, Almaz and Hong, Guanglei and Durrell, Adelle , month = oct, year =. Cooperative. Journal of Statistics and Data Science Education , publisher =. doi:10.1080/26939169.2024.2302175 , abstract =

    doi:10.1080/26939169.2024.2302175 2024

  15. [15]

    Journal of Applied Research in Higher Education , author =

    Exploring independent learning (. Journal of Applied Research in Higher Education , author =. 2023 , pages =. doi:10.1108/JARHE-06-2023-0253 , abstract =

    work page doi:10.1108/jarhe-06-2023-0253 2023

  16. [16]

    Evaluation of a self-instructional self-regulated learning material in mathematics , volume =

    Balan, Andreia and Jönsson, Anders , month = mar, year =. Evaluation of a self-instructional self-regulated learning material in mathematics , volume =. Frontiers in Education , publisher =. doi:10.3389/feduc.2025.1507803 , abstract =

    work page doi:10.3389/feduc.2025.1507803 2025

  17. [17]

    Scientechno: Journal of Science and Technology , author =

    Self-. Scientechno: Journal of Science and Technology , author =. 2024 , pages =. doi:10.55849/Scientechno.v3i1.739 , abstract =

    work page doi:10.55849/scientechno.v3i1.739 2024

  18. [18]

    International Journal of Education in Mathematics, Science and Technology , author =

    Self-. International Journal of Education in Mathematics, Science and Technology , author =. 2025 , keywords =. doi:10.46328/ijemst.4770 , abstract =

    work page doi:10.46328/ijemst.4770 2025

  19. [19]

    and Heckler, Andrew F

    Mikula, Brendon D. and Heckler, Andrew F. , month = jun, year =. Framework and implementation for improving physics essential skills via computer-based practice:. Physical Review Physics Education Research , publisher =. doi:10.1103/PhysRevPhysEducRes.13.010122 , abstract =

    work page doi:10.1103/physrevphyseducres.13.010122

  20. [20]

    , month = mar, year =

    Liu, Qiaoyi and Moni Prakash, Harish and Heckler, Andrew F. , month = mar, year =. Algebra and other relevant physics skills:. Physical Review Physics Education Research , publisher =. doi:10.1103/PhysRevPhysEducRes.21.010121 , abstract =

    work page doi:10.1103/physrevphyseducres.21.010121

  21. [21]

    and Bouck, Emily C

    Powell, Sarah R. and Bouck, Emily C. and Sutherland, Marah and Clarke, Ben and Arsenault, Tessa L. and Freeman-Green, Shaqwana , month = sep, year =. Essential. TEACHING Exceptional Children , publisher =. doi:10.1177/00400599221125892 , language =

    work page doi:10.1177/00400599221125892

  22. [22]

    and Woolford, Douglas G

    Dickson, Brandon A. and Woolford, Douglas G. and Samuels, Boba and Kotsopoulos, Donna , month = apr, year =. Active. Journal of Statistics and Data Science Education , publisher =. doi:10.1080/26939169.2025.2539237 , abstract =

    doi:10.1080/26939169.2025.2539237 2025

  23. [23]

    and Trumble, Ilana M

    Alzen, Jessica L. and Trumble, Ilana M. and Cho, Kimberly J. and Vance, Eric A. , month = jan, year =. Training. Journal of Statistics and Data Science Education , publisher =. doi:10.1080/26939169.2023.2191666 , abstract =

    work page doi:10.1080/26939169.2023.2191666 2023

  24. [24]

    Business & Information Systems Engineering , author =

    Beyond the. Business & Information Systems Engineering , author =. 2025 , keywords =. doi:10.1007/s12599-025-00954-2 , abstract =

    work page doi:10.1007/s12599-025-00954-2 2025

  25. [25]

    Dong, Yixiao and Baral, Deodatta and Baral, Kushmakar , month = jan, year =. Are. Frontiers in Education , publisher =. doi:10.3389/feduc.2025.1657651 , abstract =

    work page doi:10.3389/feduc.2025.1657651 2025

  26. [26]

    Zarefard, Motahareh and Marsden, Nicola , month = jun, year =. The. Human. doi:10.1007/978-3-031-60125-5_27 , abstract =

    work page doi:10.1007/978-3-031-60125-5_27

  27. [27]

    Bringing

    Bednarowska-Michaiel, Zofia and Uprichard, Emma , month = jan, year =. Bringing. Journal of Statistics and Data Science Education , publisher =. doi:10.1080/26939169.2025.2507366 , abstract =

    work page doi:10.1080/26939169.2025.2507366 2025

  28. [28]

    Instructional Science , author =

    Cultivating science data literacy in. Instructional Science , author =. 2026 , keywords =. doi:10.1007/s11251-026-09803-5 , abstract =

    work page doi:10.1007/s11251-026-09803-5 2026

  29. [29]

    Data Driven Discovery in Astrophysics

    Longo, G. and Brescia, M. and Djorgovski, S. G. and Cavuoti, S. and Donalek, C. , month = nov, year =. Data. doi:10.48550/arXiv.1410.5631 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1410.5631

  30. [30]

    and Hoyt, Robert E

    Hersh, William R. and Hoyt, Robert E. and Chamberlin, Steven and Ancker, Jessica S. and Gupta, Aditi and Borlawsky-Payne, Tara B. , month = jul, year =. Beyond mathematics, statistics, and programming: data science, machine learning, and artificial intelligence competencies and curricula for clinicians, informaticians, science journalists, and researchers...

    work page doi:10.1080/20476965.2023.2237745 2023

  31. [31]

    Journal of Data Science , publisher =

    Fu, Valerie , month = jan, year =. Journal of Data Science , publisher =. doi:10.6339/25-JDS1214 , abstract =

    work page doi:10.6339/25-jds1214

  32. [32]

    Data science skills for the next generation of statisticians , volume =

    Antonucci, Laura and Balzanella, Antonio and Bruno, Elvira and Crocetta, Crocetta and Zio, Simone Di and Fontanella, Lara and Sanarico, Maurizio and Scarpa, Bruno and Verde, Rosanna and Vittadini, Giorgio , month = nov, year =. Data science skills for the next generation of statisticians , volume =. Statistical Journal of the IAOS , publisher =. doi:10.32...

    work page doi:10.3233/sji-230060

  33. [33]

    and Elofsson, Arne and Jarlebring, Elias and Kjellström, Hedvig and Markidis, Stefano and Marlevi, David and García-Martínez, Javier and Brunton, Steven L

    Vinuesa, Ricardo and Cinnella, Paola and Rabault, Jean and Azizpour, Hossein and Bauer, Stefan and Brunton, Bingni W. and Elofsson, Arne and Jarlebring, Elias and Kjellström, Hedvig and Markidis, Stefano and Marlevi, David and García-Martínez, Javier and Brunton, Steven L. , month = may, year =. Decoding complexity through machine learning is redefining s...

    work page doi:10.1038/s42005-026-02676-7

  34. [34]

    Porcu, Emilio and Moukari, Roy El and Najman, Laurent and Herrera, Francisco and Simon, Horst , month = jan, year =. Data. doi:10.48550/arXiv.2506.11010 , abstract =

    work page doi:10.48550/arxiv.2506.11010

  35. [35]

    Hey, Tony and Tansley, Stewart and Tolle, Kristin and Gray, Jim , month = oct, year =. The

  36. [36]

    Scientific

    Borne, Kirk , month = nov, year =. Scientific. doi:10.48550/arXiv.0911.0505 , abstract =

    Pith/arXiv arXiv doi:10.48550/arxiv.0911.0505

  37. [37]

    and Gutiérrez, C

    Csörnyei, G. and Gutiérrez, C. P. , month = apr, year =. Between plateaus and slopes:. Astronomy & Astrophysics , publisher =. doi:10.1051/0004-6361/202556773 , abstract =

    doi:10.1051/0004-6361/202556773

  38. [38]

    and Ishida, E

    Sasdelli, M. and Ishida, E. E. O. and Vilalta, R. and Aguena, M. and Busti, V. C. and Camacho, H. and Trindade, A. M. M. and Gieseke, F. and de Souza, R. S. and Fantaye, Y. T. and Mazzali, P. A. , month = sep, year =. Exploring the spectroscopic diversity of. Monthly Notices of the Royal Astronomical Society , publisher =. doi:10.1093/mnras/stw1228 , abstract =

    work page doi:10.1093/mnras/stw1228

  39. [39]

    Guillochon, James and Parrent, Jerod and Kelley, Luke Zoltan and Margutti, Raffaella , month = jan, year =. An. The Astrophysical Journal , publisher =. doi:10.3847/1538-4357/835/1/64 , abstract =

    doi:10.3847/1538-4357/835/1/64

  40. [40]

    , month = jan, year =

    Gal-Yam, A. , month = jan, year =. The

  41. [41]

    and Nugent, Peter E

    Silverman, Jeffrey M. and Nugent, Peter E. and Gal-Yam, Avishay and Sullivan, Mark and Howell, D. Andrew and Filippenko, Alexei V. and Arcavi, Iair and Ben-Ami, Sagi and Bloom, Joshua S. and Cenko, S. Bradley and Cao, Yi and Chornock, Ryan and Clubb, Kelsey I. and Coil, Alison L. and Foley, Ryan J. and Graham, Melissa L. and Griffith, Christopher V. and H...

    doi:10.1088/0067-0049/207/1/3

  42. [42]

    and Challis, P

    Foley, Ryan J. and Challis, P. J. and Chornock, R. and Ganeshalingam, M. and Li, W. and Marion, G. H. and Morrell, N. I. and Pignata, G. and Stritzinger, M. D. and Silverman, J. M. and Wang, X. and Anderson, J. P. and Filippenko, A. V. and Freedman, W. L. and Hamuy, M. and Jha, S. W. and Kirshner, R. P. and McCully, C. and Persson, S. E. and Phillips, M. ...

    doi:10.1088/0004-637x/767/1/57

  43. [43]

    and Gebru, Timnit and McMillan-Major, Angelina and Shmitchell, Shmargaret , month = mar, year =

    Bender, Emily M. and Gebru, Timnit and McMillan-Major, Angelina and Shmitchell, Shmargaret , month = mar, year =. On the. Proceedings of the 2021. doi:10.1145/3442188.3445922 , abstract =

    doi:10.1145/3442188.3445922 2021

  44. [44]

    2019 , note =

    The Astrophysical Journal , author =. 2019 , note =. doi:10.3847/1538-4357/ab042c , abstract =

    doi:10.3847/1538-4357/ab042c 2019

  45. [45]

    and Corbett, Hank and Galliher, Nathan W

    Law, Nicholas M. and Corbett, Hank and Galliher, Nathan W. and Gonzalez, Ramses and Vasquez, Alan and Walters, Glenn and Machia, Lawrence and Ratzloff, Jeff and Ackley, Kendall and Bizon, Chris and Clemens, Christopher and Cox, Steven and Eikenberry, Steven and Howard, Ward S. and Glazier, Amy and Mann, Andrew W. and Quimby, Robert and Reichart, Daniel an...

    doi:10.1088/1538-3873/ac4811

  46. [46]

    Deng, Yuyang and Hong, Junyuan and Zhou, Jiayu and Mahdavi, Mehrdad , month = mar, year =. On the. doi:10.48550/arXiv.2403.06871 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2403.06871

  47. [47]

    Erhan, Dumitru and Courville, Aaron and Bengio, Yoshua and Vincent, Pascal , month = mar, year =. Why. Proceedings of the

  48. [48]

    Neural Networks , author =

    Multilayer feedforward networks are universal approximators , volume =. Neural Networks , author =. 1989 , keywords =. doi:10.1016/0893-6080(89)90020-8 , abstract =

    doi:10.1016/0893-6080(89)90020-8 1989

  49. [49]

    , month = feb, year =

    Qi, Tao and Yin, Jinhua and Cai, Dongqi and Xie, Yueqi and Wang, Huili and Hu, Zhiyang and Yang, Peiru and Nan, Guoshun and Zhou, Zhili and Wu, Chuhan and Lyu, Lingjuan and Wang, Shangguang and Huang, Yongfeng and Lane, Nicholas D. , month = feb, year =. Auditing unauthorized training data from. Nature Communications , publisher =. doi:10.1038/s41467-026-...

    work page doi:10.1038/s41467-026-68862-x

  50. [50]

    Proceedings of the Association for Information Science and Technology , author =

    Library. Proceedings of the Association for Information Science and Technology , author =. 2025 , pages =. doi:10.1002/pra2.1340 , abstract =

    work page doi:10.1002/pra2.1340 2025

  51. [51]

    and Steinhardt, Jacob and Flynn, Carrick and hÉigeartaigh, Seán Ó and Beard, S

    Brundage, Miles and Avin, Shahar and Clark, Jack and Toner, Helen and Eckersley, Peter and Garfinkel, Ben and Dafoe, Allan and Scharre, Paul and Zeitzoff, Thomas and Filar, Bobby and Anderson, Hyrum and Roff, Heather and Allen, Gregory C. and Steinhardt, Jacob and Flynn, Carrick and hÉigeartaigh, Seán Ó and Beard, S. J. and Belfield, Haydn and Farquhar, S...

    Pith/arXiv arXiv doi:10.48550/arxiv.1802.07228

  52. [52]

    Proceedings of the AAAI Conference on Artificial Intelligence , author =

    Can. Proceedings of the AAAI Conference on Artificial Intelligence , author =. 2026 , pages =. doi:10.1609/aaai.v40i36.40269 , abstract =

    work page doi:10.1609/aaai.v40i36.40269 2026

  53. [53]

    Exploiting Instruction-Following Retrievers for Malicious Information Retrieval

    BehnamGhader, Parishad and Meade, Nicholas and Reddy, Siva , month = mar, year =. Exploiting. doi:10.48550/arXiv.2503.08644 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2503.08644

  54. [54]

    Loyola of Los Angeles Law Review , author =

    Hacking. Loyola of Los Angeles Law Review , author =. 2025 , pages =

    2025

  55. [55]

    Proceedings of the AAAI Conference on Artificial Intelligence , author =

    Response. Proceedings of the AAAI Conference on Artificial Intelligence , author =. 2026 , pages =. doi:10.1609/aaai.v40i41.40836 , abstract =

    work page doi:10.1609/aaai.v40i41.40836 2026

  56. [56]

    Wang, Xinlei and Ming, Ruibo and Qiu, Jing and Zhao, Junhua and Gu, Jinjin , month = may, year =. Hugging. doi:10.48550/arXiv.2605.01549 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2605.01549

  57. [57]

    Holistically

    Morrison, Jacob and Na, Clara and Fernandez, Jared and Dettmers, Tim and Strubell, Emma and Dodge, Jesse , month = mar, year =. Holistically. doi:10.48550/arXiv.2503.05804 , abstract =

    Pith/arXiv arXiv doi:10.48550/arxiv.2503.05804

  58. [58]

    Energy and

    Strubell, Emma and Ganesh, Ananya and McCallum, Andrew , editor =. Energy and. Proceedings of the 57th. 2019 , pages =. doi:10.18653/v1/P19-1355 , abstract =

    doi:10.18653/v1/p19-1355 2019

  59. [59]

    CarbonScaling: Extending Neural Scaling Laws for Carbon Footprint in Large Language Models

    Jiang, Lei and Chen, Fan , month = may, year =. doi:10.48550/arXiv.2508.06524 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2508.06524

  60. [60]

    The Hidden Cost of Thinking: Energy Use and Environmental Impact of LMs Beyond Pretraining

    Morrison, Jacob and Smith, Noah A. and Strubell, Emma , month = may, year =. The. doi:10.48550/arXiv.2605.01158 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2605.01158

  61. [61]

    Exposing

    Mochizuki, Riku and Komatsu, Shusuke and Noguchi, Souta and Ataka, Kazuto , month = mar, year =. Exposing. doi:10.48550/arXiv.2510.06823 , abstract =

    doi:10.48550/arxiv.2510.06823

  62. [62]

    Authority, Truth, and Citation Bias: A Large-Scale Multi-Domain Benchmark for Studying Epistemic Susceptibility in Large Language Models

    Khurana, Aryan and RN, Aravind Ramana and Kumar, Dhruv , month = jun, year =. Authority,. doi:10.48550/arXiv.2606.13104 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2606.13104

  63. [63]

    How unique are hallucinated citations offered by generative Artificial Intelligence models?

    Spennemann, Dirk HR , month = mar, year =. How unique are hallucinated citations offered by generative. doi:10.48550/arXiv.2604.16407 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2604.16407

  64. [64]

    Disinformation

    Vykopal, Ivan and Pikuliak, Matúš and Srba, Ivan and Moro, Robert and Macko, Dominik and Bielikova, Maria , editor =. Disinformation. Proceedings of the 62nd. 2024 , pages =. doi:10.18653/v1/2024.acl-long.793 , abstract =

    doi:10.18653/v1/2024.acl-long.793 2024

  65. [65]

    2026 , pages =

    Misinformation in the age of generative. 2026 , pages =. doi:10.1063/5.0329005 , author =

    work page doi:10.1063/5.0329005 2026

  66. [66]

    Artificial Intelligence Review , author =

    Lies, damned lies, and language statistics: a comprehensive review of risks from manipulation, persuasion, and deception with large language models , volume =. Artificial Intelligence Review , author =. 2026 , keywords =. doi:10.1007/s10462-026-11517-6 , abstract =

    doi:10.1007/s10462-026-11517-6 2026

  67. [67]

    Proceedings of the AAAI Conference on Artificial Intelligence , author =

    The. Proceedings of the AAAI Conference on Artificial Intelligence , author =. 2026 , pages =. doi:10.1609/aaai.v40i42.40856 , abstract =

    work page doi:10.1609/aaai.v40i42.40856 2026

  68. [68]

    Proceedings of the AAAI Conference on Artificial Intelligence , author =

    How. Proceedings of the AAAI Conference on Artificial Intelligence , author =. 2026 , pages =. doi:10.1609/aaai.v40i45.41181 , abstract =

    work page doi:10.1609/aaai.v40i45.41181 2026

  69. [69]

    Venkit, Pranav Narayanan and Li, Jiayi and Zhou, Yingfan and Rajtmajer, Sarah and Wilson, Shomir , month = may, year =. A. doi:10.48550/arXiv.2505.07850 , abstract =

    Pith/arXiv arXiv doi:10.48550/arxiv.2505.07850

  70. [70]

    Silenced

    Himelstein, Rom and LeVi, Amit and Youngmann, Brit and Nemcovsky, Yaniv and Mendelson, Avi , month = mar, year =. Silenced. doi:10.48550/arXiv.2511.03369 , abstract =

    work page doi:10.48550/arxiv.2511.03369

  71. [71]

    and Monroe-White, Thema , month = jan, year =

    Shieh, Evan and Vassel, Faye-Marie and Sugimoto, Cassidy R. and Monroe-White, Thema , month = jan, year =. Intersectional biases in narratives produced by open-ended prompting of generative language models , volume =. Nature Communications , publisher =. doi:10.1038/s41467-025-68004-9 , abstract =

    doi:10.1038/s41467-025-68004-9

  72. [72]

    Ethical and social risks of harm from

    Weidinger, Laura and Mellor, John and Rauh, Maribeth and Griffin, Conor and Uesato, Jonathan and Huang, Po-Sen and Cheng, Myra and Glaese, Mia and Balle, Borja and Kasirzadeh, Atoosa and Kenton, Zac and Brown, Sasha and Hawkins, Will and Stepleton, Tom and Biles, Courtney and Birhane, Abeba and Haas, Julia and Rimell, Laura and Hendricks, Lisa Anne and Is...

    Pith/arXiv arXiv doi:10.48550/arxiv.2112.04359

  73. [73]

    OpenAI and Achiam, Josh and Adler, Steven and Agarwal, Sandhini and Ahmad, Lama and Akkaya, Ilge and Aleman, Florencia Leoni and Almeida, Diogo and Altenschmidt, Janko and Altman, Sam and Anadkat, Shyamal and Avila, Red and Babuschkin, Igor and Balaji, Suchir and Balcom, Valerie and Baltescu, Paul and Bao, Haiming and Bavarian, Mohammad and Belgum, Jeff a...

    Pith/arXiv arXiv doi:10.48550/arxiv.2303.08774

  74. [74]

    Language

    Radford, Alec and Wu, Jeffrey and Child, Rewon and Luan, David and Amodei, Dario and Sutskever, Ilya , year =. Language

  75. [75]

    Improving

    Radford, Alec and Narasimhan, Karthik and Salimans, Tim and Sutskever, Ilya , year =. Improving

  76. [76]

    Brown, Tom B. and Mann, Benjamin and Ryder, Nick and Subbiah, Melanie and Kaplan, Jared and Dhariwal, Prafulla and Neelakantan, Arvind and Shyam, Pranav and Sastry, Girish and Askell, Amanda and Agarwal, Sandhini and Herbert-Voss, Ariel and Krueger, Gretchen and Henighan, Tom and Child, Rewon and Ramesh, Aditya and Ziegler, Daniel M. and Wu, Jeffrey and W...

    Pith/arXiv arXiv doi:10.48550/arxiv.2005.14165 2005

  77. [77]

    Instruction

    Cheng, Daixuan and Gu, Yuxian and Huang, Shaohan and Bi, Junyu and Huang, Minlie and Wei, Furu , month = nov, year =. Instruction. doi:10.48550/arXiv.2406.14491 , abstract =

    Pith/arXiv arXiv doi:10.48550/arxiv.2406.14491

  78. [78]

    and Chess, Benjamin and Child, Rewon and Gray, Scott and Radford, Alec and Wu, Jeffrey and Amodei, Dario , month = jan, year =

    Kaplan, Jared and McCandlish, Sam and Henighan, Tom and Brown, Tom B. and Chess, Benjamin and Child, Rewon and Gray, Scott and Radford, Alec and Wu, Jeffrey and Amodei, Dario , month = jan, year =. Scaling. doi:10.48550/arXiv.2001.08361 , abstract =

    Pith/arXiv arXiv doi:10.48550/arxiv.2001.08361 2001

  79. [79]

    Neural Computation , author =

    Long. Neural Computation , author =. 1997 , pages =. doi:10.1162/neco.1997.9.8.1735 , abstract =

    doi:10.1162/neco.1997.9.8.1735 1997

  80. [80]

    Proceedings of the IEEE , author =

    Backpropagation through time: what it does and how to do it , volume =. Proceedings of the IEEE , author =. 1990 , pages =. doi:10.1109/5.58337 , abstract =

    doi:10.1109/5.58337 1990

Showing first 80 references.