arxiv: 2605.03719 · v1 · submitted 2026-05-05 · 🌌 astro-ph.EP · astro-ph.IM

Recognition: unknown

A public dataset of Ariel simulated observations for developing exoplanetary atmosphere data reduction pipelines

Lorenzo V. Mugnai , Kai Hou Yip , Andrea Bocchieri , Andreas Papageorgiou , Virginie Batista , Orph\'ee Faucoz , Ang\`ele Syty , Tara Tahseen

show 2 more authors

Enzo Pascale Ingo Waldmann

Authors on Pith no claims yet

Pith reviewed 2026-05-07 13:12 UTC · model grok-4.3

classification 🌌 astro-ph.EP astro-ph.IM

keywords exoplanet atmospheresAriel missiontransmission spectroscopydata detrendingsimulated datasetsmachine learning baselinesdataset shiftnoise systematics

0 comments

The pith

A public dataset of simulated Ariel mission observations provides ground-truth benchmarks for developing and testing exoplanet atmosphere data reduction pipelines, including a neural-network baseline that exposes risks from dataset shift.

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

The paper generates and releases one of the largest public collections of simulated Ariel observations, created with ExoSim2 and TauREx to match the mission's current instrument design. The resource supplies known atmospheric signals plus realistic noise and systematics so that researchers can train and compare detrending methods without waiting for flight data. A deep neural network is included as a baseline for time-series reduction, and tests with it demonstrate that machine-learning approaches can fail when the statistical properties of new observations differ from the training set. This addresses a core bottleneck in exoplanet spectroscopy: atmospheric features are often only tens to hundreds of parts per million, easily masked by correlated noise. By making the dataset available through the Ariel Data Challenge on Kaggle, the work supplies a concrete testbed for any new algorithm before it is applied to real observations.

Core claim

The authors produce a large-scale public dataset of Ariel-simulated transit observations that includes both the planetary atmospheric signals and the full range of instrumental and astrophysical noise expected from the mission. They pair this with a deep neural network baseline for detrending and show that the network's performance drops when the test data distribution diverges from the training distribution, illustrating the practical limits of purely data-driven methods on future Ariel spectra.

What carries the argument

The simulated dataset itself, generated by ExoSim2 for instrument effects and TauREx for atmospheric models, together with the provided deep neural network baseline for time-series reduction.

If this is right

Algorithms developed and validated on the dataset can be applied directly to Ariel survey data with greater confidence in their robustness.
The explicit demonstration of dataset-shift failure provides a concrete test that any new machine-learning detrending method must pass before deployment.
The resource enables direct head-to-head comparison of classical and data-driven detrending techniques on identical inputs with known ground truth.
Community-wide stress-testing on the dataset can identify which methods scale reliably to the full Ariel sample of roughly one thousand planets.

Where Pith is reading between the lines

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

The same simulation framework and shift-testing approach could be adapted to benchmark pipelines for JWST or future missions such as Habitable Worlds Observatory.
If real Ariel data exhibit stronger or different systematics than the current simulations, retraining the baseline network on a mixture of simulated and early flight data would be a natural next step.
The dataset could also serve as a controlled testbed for studying how stellar variability or spot-crossing events interact with atmospheric retrievals.

Load-bearing premise

The simulated noise, systematics, and signal properties accurately reproduce what will appear in real Ariel observations.

What would settle it

Once Ariel flight data are available, compare the statistical distributions and residual noise properties of the real spectra against the simulated dataset; large, systematic mismatches would show that the simulations do not capture the actual instrument behavior.

Figures

Figures reproduced from arXiv: 2605.03719 by Andrea Bocchieri, Andreas Papageorgiou, Ang\`ele Syty, Enzo Pascale, Ingo Waldmann, Kai Hou Yip, Lorenzo V. Mugnai, Orph\'ee Faucoz, Tara Tahseen, Virginie Batista.

**Figure 2.** Figure 2: The figure reports two of the PSFs used to simulate FGS1 (left) and AIRS-CH0 (right). The PSFs are sampled at the sub-pixel level and reported in decibels normalised to the maximum value in the field. The axes report the grid into pixel and micron units. The reader may note that the FGS1 PSF is highly aberrated and complex, so it is sampled on a grid of the same size as the detector (32 × 32), while the AI… view at source ↗

**Figure 3.** Figure 3: shows the selected regions on the JWST flat field calibration product. These regions were manually selected by visually comparing different calibration products to include a comprehensive mixture of possible detector effects. Note that this flat field is calculated as the variation from the median value, i.e., using the flat field divided by the median view at source ↗

**Figure 4.** Figure 4: The figure shows the first two NDRs for every channel. NDR 0 is collected at the beginning of the photon collection ramp, while NDR 1 is collected at the end. The figures are reported in counts units as they have been converted from ADUs by inverting Equation 4 as 𝑆𝑚𝑒𝑎𝑠 = 𝑆𝑎𝑑𝑢 𝐴𝐷𝐶𝑔𝑎𝑖𝑛 + 𝐴𝐷𝐶𝑜 𝑓 𝑓 𝑠𝑒𝑡. The top row shows the photometer FGS1 images, while the bottom row shows the spectrometer AIRS-CH0 images. … view at source ↗

**Figure 6.** Figure 6: Visualisation in spectral space of the different training and test cases presented in view at source ↗

**Figure 7.** Figure 7: White light curve (i.e., the resultant light curve from using the spectrometer as a single wavelength channel) for one of the simulated observations with Ariel AIRS-CH0. The signal dip from the transit event is highlighted in the green box, with the depth of the transit also marked in purple. We highlighted other significant features: the spread of the measurements due to high-frequency and quantum noise… view at source ↗

**Figure 8.** Figure 8: Maximum signal detected out-of-transit for the photometer (FGS1) and spectrometer (AIRS-CH0). The colour code indicates the stellar targets, illustrating how differences in stellar spectral energy distributions influence the observed signals. 0.75 0 100 200 300 400 500 2.00 2.25 2.50 2.75 3.00 3.25 3.50 3.75 Wavelength [ m] Reference: N/S [ppm] HD149026 K11 HD17156 HD209458 view at source ↗

**Figure 9.** Figure 9: Relative noise (ppm) as a function of wavelength for the reference timelines. The colour coding represents different stellar targets, reflecting how their SEDs influence the noise levels across the spectrum. The large spikes come from detector imperfections such as bad pixels that are left uncorrected. be traced to noise sources that are not corrected or insufficiently corrected, e.g., from correlated and … view at source ↗

**Figure 12.** Figure 12: Comparison of transit spectra for selected examples from the subsample. Black: ground truth spectra; orange: reference data; blue: test data. Error bars are included. 0 5 10 Counts Test data ( , ± ) (-76, +116 123 ) 400 300 200 100 0 100 200 (Rp/Rs) 2 (Rp/Rs) 2 G. T. [ppm] 0 5 10 Counts Reference ( , ± ) (12, +29 27 ) view at source ↗

**Figure 13.** Figure 13: Histogram of residuals between median extracted transit depths and ground truth. The black line shows the mean and the red dash line shows the standard deviation. Top panel: test data, showing larger deviations and a negative bias. Bottom panel: reference data, with deviations within ±100 ppm. should be able to retrieve the signal without bias within the permitted noise level (above or close to the photon… view at source ↗

**Figure 14.** Figure 14: Stellar normalisation of the detector images. Top: raw flux before normalisation, showing the strongly wavelength-dependent stellar continuum concentrated along the spatial trace. Bottom: after dividing by the out-oftransit stellar spectrum, equalising flux across wavelengths and revealing the spatial PSF profile. Left: 2D detector images (spatial pixels × wavelength bins); right: corresponding 3D surfac… view at source ↗

**Figure 15.** Figure 15: illustrate the two different ML strategies deployed in this investigation. The two approaches uses the same initial step (fit with a 1D CNN top row) followed by two different architectures (3D-CNN and 2D-CNN) to tackle the problem using different inputs view at source ↗

**Figure 16.** Figure 16: Example of a datacube used as an input for the first baseline with the 3D-CNN. The transit depth is exaggerated for clarity and better visualisation view at source ↗

**Figure 17.** Figure 17: Normalised transmission spectra (Rp/Rs)² across wavelength bins for the 3D-CNN baseline. Green: training targets; red: validation true means. use Monte Carlo dropout (Gal & Ghahramani 2016) fixing at 1000 and 30 the number of instances for the first (1D-CNN) and second (3D-CNN) CNNs respectively. 4.4 Second baseline The second baseline fits the tiny fluctuations around the mean of the spectra to search fo… view at source ↗

**Figure 19.** Figure 19: Normalised transmission spectra (Rp/Rs)² across wavelength bins for the 2D-CNN baseline. Green: training targets; red: validation true means view at source ↗

**Figure 20.** Figure 20: Normalization of input data (2D images) for the second baseline 2D CNN, shown here for a given planet at a given time. Ghahramani 2016) fixing at 1000 and 30 the number of instances for the first and second CNNs respectively. 4.5 Evaluation Metric We evaluate the quality of the predicted spectra, (𝑥˜𝜆), and their predicted uncertainties at different wavelengths, (𝜎˜ )𝜆), S˜ = (𝑥˜𝜆, 𝜎˜ 𝜆) against the grou… view at source ↗

**Figure 21.** Figure 21: MSE distribution for validation and test data (second baseline). 35 test MSE higher than 1000 are not captured in this plot, reaching a maximum of 3440. Valid GLL score Valid MSE (ppm) Test GLL score Test MSE (ppm) 3D-CNN Baseline 0.4681 128 0.4232 544 2D-CNN Baseline 0.5020 86 0.4246 542 view at source ↗

**Figure 22.** Figure 22: Model predictions from the second baseline across 16 randomly selected validation targets. Black dots/line: predicted mean spectrum; grey band: predicted uncertainty envelope; red line: true target spectrum view at source ↗

**Figure 23.** Figure 23: Model predictions from the second baseline across 16 randomly selected test targets. Black dots/line: predicted mean spectrum; grey band: predicted uncertainty envelope; red line: true target spectrum. RASTI 000, 1–25 (2025) view at source ↗

**Figure 24.** Figure 24: Distribution of transit duration, mid-transit time and transit depth, for train and test sets, represented by shades of blue. The solid rectangle is centered on the mean value, and its height is defined by the standard deviation. The whiskers extend to the minimum and maximum values, representing the range of the data. Time indices are converted in minutes (see conversion in section 4.1). their power from… view at source ↗

**Figure 25.** Figure 25: Predictions of spectra mean values for validation (top) and test (bottom) data, with error bars (black dots). Ground Truth mean value are represented by red dots and residuals are shown in the lower part of each figure (blue points). features of the shifted domain. However, further additions show diminishing returns: increasing from 50 to 400 samples provides only incremental MSE reduction (271 to 185 pp… view at source ↗

**Figure 26.** Figure 26: Test set transmission spectra after subtracting from the population median, highlighting relative atmospheric features across the wavelength range. Original scores With 50 test data With 100 test data With 150 test data With 400 test data With 700 test data Valid Test Valid Test Valid Test Valid Test Valid Test Valid Test MSE mean values (in ppm) 79 530 97 254 137 259 117 171 145 175 127 140 MSE fluctuati… view at source ↗

read the original abstract

Detecting and characterising exoplanet atmospheres remains challenging because atmospheric signals can be comparable to residual noise and instrumental/astrophysical systematics. Spectral features span from a few ppm for small planets up to $\sim 10^3$ ppm for warm/hot giants, while high-quality JWST time-series spectroscopy typically reaches $\sim 10$--$50$ ppm (occasionally $\sim 100$--$200$ ppm in the presence of stellar variability or stronger systematics), making correlated noise across temporal and spectral dimensions a key limitation. With JWST delivering an increasing volume of high-precision transmission spectra, and Ariel set to extend this to a homogeneous survey of $\sim 10^3$ exoplanet atmospheres, robust benchmarking resources with known ground truth are essential to develop and validate data-driven (including ML-based) detrending approaches. As a major step towards this goal, we use ExoSim2 and TauREx to generate one of the most comprehensive public datasets based on the current payload design of the ESA Ariel mission, specifically intended to benchmark detrending algorithms. We also provide a deep neural network baseline for time-series reduction, and use it to highlight the limitations of ML based detrendng methods, i.e. the risks posed by dataset shift when observed distributions diverge from those of the training set, a scenario likely to arise in real observations. This dataset is featured in the Ariel Data Challenge 2024 on Kaggle and has been field-tested for robustness and simulation fidelity. By making these resources publicly available, we aim to support the community in developing, comparing, and stress-testing scalable and reliable methods for exoplanet transmission spectroscopy.

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.

Desk Editor's Note private letter to a colleague

A practical public dataset release for Ariel detrending benchmarks, with a DNN baseline that flags dataset shift risks inside the simulations.

read the letter

The main thing to know is that this paper puts out a large simulated dataset of Ariel observations generated with ExoSim2 and TauREx, based on the current payload design, and includes a deep neural network baseline for time-series reduction plus a controlled demonstration of how ML methods can fail under dataset shift. They tie it to the Ariel Data Challenge 2024 on Kaggle, which makes the resource immediately usable for testing detrending pipelines ahead of the mission. That is the core contribution and it is new in its scale and Ariel-specific tailoring. The work does well by being openly available with ground truth, which directly supports the community need for benchmarks as Ariel targets a large homogeneous survey. The shift example is a straightforward and relevant caution, since real observations will likely diverge from training distributions. The simulations aim to capture noise and systematics at the levels expected from JWST experience, and the authors note field-testing for robustness. The softer spots are modest. The fidelity claims rest on the simulation tools reproducing real Ariel behavior, and the shift test stays within the same framework rather than crossing to actual data or other instruments. Without the full validation figures and quantitative checks in the methods, it is hard to judge how close the dataset sits to future observations, though the abstract presents this as a data-release effort rather than a new physical result. This paper is for people building or evaluating data reduction pipelines for Ariel, transmission spectroscopy teams, and anyone testing ML approaches on exoplanet time series. It gives them a concrete benchmark to work with. I would send it to peer review as a data paper; the release itself is useful and the baseline adds a clear practical point even if the central value is the dataset.

Referee Report

0 major / 3 minor

Summary. The paper presents a public dataset of simulated Ariel exoplanet observations generated with ExoSim2 and TauREx based on the current mission payload design. The dataset is intended to support development and benchmarking of detrending and data reduction pipelines for transmission spectroscopy. The authors also supply a deep neural network baseline for time-series reduction and use it to illustrate the performance degradation caused by dataset shift when test distributions differ from the training set.

Significance. If the simulations faithfully reproduce Ariel's expected noise properties and systematics, the dataset will serve as a key community resource for validating algorithms ahead of the mission's survey of approximately 1000 atmospheres. The DNN baseline provides a concrete, reproducible demonstration of dataset-shift risks that is directly relevant to real observations, where training and test distributions are unlikely to match perfectly. Public release through the Ariel Data Challenge 2024 on Kaggle increases the work's immediate utility for comparative testing of detrending methods.

minor comments (3)

Abstract: 'detrendng' is a typographical error and should read 'detrending'.
Abstract and introduction: the statement that the dataset is 'one of the most comprehensive' would benefit from a brief quantitative comparison (e.g., number of planets, spectral channels, or total simulated hours) to previously released Ariel or JWST simulation suites.
The manuscript would be strengthened by an explicit statement of the dataset access URL, file formats, and any accompanying documentation or metadata files in the main text rather than only in supplementary material.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive assessment of the manuscript, recognition of the dataset's utility for the Ariel mission community, and recommendation to accept. We are pleased that the work's relevance for benchmarking detrending pipelines and illustrating dataset-shift risks in ML-based methods was acknowledged.

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

This is a data-release paper whose core contribution is the generation and public provision of simulated Ariel observations using the established external tools ExoSim2 and TauREx, together with a DNN baseline trained and evaluated inside the same simulation framework. No equations, fitted parameters, or predictions are presented that reduce to the inputs by construction. The dataset-shift demonstration is a controlled internal experiment within the simulated distributions and does not rely on self-definitional steps, load-bearing self-citations, or renaming of known results. All claims rest on the explicit provision of the dataset and code, which are externally verifiable and independent of any internal derivation chain.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a data-release paper using existing simulation software; no new free parameters, axioms, or invented entities are introduced by the central claim.

pith-pipeline@v0.9.0 · 5649 in / 1051 out tokens · 63332 ms · 2026-05-07T13:12:35.145649+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

66 extracted references · 55 canonical work pages

[1]

H., Glidden, A., et al

JWST-TST DREAMS: NIRSpec/PRISM Transmission Spectroscopy of the Habitable Zone Planet TRAPPIST-1 e. , keywords =. doi:10.3847/2041-8213/adf42e , archivePrefix =. 2509.05414 , primaryClass =

work page doi:10.3847/2041-8213/adf42e 2041
[2]

D., et al

JWST NIRISS transmission spectroscopy of the super-Earth GJ 357b, a favourable target for atmospheric retention. , keywords =. doi:10.1093/mnras/staf894 , archivePrefix =. 2505.24462 , primaryClass =

work page doi:10.1093/mnras/staf894
[3]

L., May, E

A benchmark JWST near-infrared spectrum for the exoplanet WASP-39 b. Nature Astronomy , keywords =. doi:10.1038/s41550-024-02292-x , archivePrefix =. 2407.13893 , primaryClass =

work page doi:10.1038/s41550-024-02292-x
[4]

2023, Monthly Notices of the Royal Astronomical Society, 524, 835, doi: 10.1093/mnras/stad1762

Awesome SOSS: transmission spectroscopy of WASP-96b with NIRISS/SOSS. , keywords =. doi:10.1093/mnras/stad1762 , archivePrefix =. 2305.17001 , primaryClass =

work page doi:10.1093/mnras/stad1762
[5]

C., Weiner Mansfield, M., Cubillos, P

Horizontal and vertical exoplanet thermal structure from a JWST spectroscopic eclipse map. Nature Astronomy , keywords =. doi:10.1038/s41550-025-02666-9 , archivePrefix =. 2510.24708 , primaryClass =

work page doi:10.1038/s41550-025-02666-9
[6]

2018, Experimental Astronomy, 46, 135, doi: 10.1007/s10686-018-9598-x

A chemical survey of exoplanets with ARIEL. Experimental Astronomy , keywords =. doi:10.1007/s10686-018-9598-x , adsurl =

work page doi:10.1007/s10686-018-9598-x
[7]

P., Line, M

Characterizing Transiting Exoplanet Atmospheres with JWST. , keywords =. doi:10.3847/0004-637X/817/1/17 , archivePrefix =. 1511.05528 , primaryClass =

work page doi:10.3847/0004-637x/817/1/17
[8]

MIRI-LRS Spectrum of a Cold Exoplanet around a White Dwarf: Water, Ammonia, and Methane Measurements

MIRI-LRS Spectrum of a Cold Exoplanet around a White Dwarf: Water, Ammonia, and Methane Measurements. , keywords =. doi:10.3847/2041-8213/adbd46 , archivePrefix =. 2503.04531 , primaryClass =

work page doi:10.3847/2041-8213/adbd46 2041
[9]

, keywords =

Integrating Light Curve and Atmospheric Modeling of Transiting Exoplanets. , keywords =. doi:10.3847/1538-3881/abaabc , archivePrefix =. 1811.04686 , primaryClass =

work page doi:10.3847/1538-3881/abaabc
[10]

, keywords =

Toward Atmospheric Retrievals of Panchromatic Light Curves: EXPLOR-ing Generalized Inversion Techniques for Transiting Exoplanets with JWST and Ariel. , keywords =. doi:10.3847/1538-3881/ad3032 , archivePrefix =. 2403.02244 , primaryClass =

work page doi:10.3847/1538-3881/ad3032
[11]

Empirical line lists for AlH and AlD and experimental emission spectroscopy of AlD in A ^ 1 (v = 0, 1, 2)

ExoMol line lists - LIV. Empirical line lists for AlH and AlD and experimental emission spectroscopy of AlD in A ^ 1 (v = 0, 1, 2). , keywords =. doi:10.1093/mnras/stad3802 , archivePrefix =. 2312.05958 , primaryClass =

work page doi:10.1093/mnras/stad3802
[12]

RAS Techniques and Instruments , keywords =

Simple semiclassical model of pressure-broadened infrared/microwave linewidths in the temperature range 200-3000 K. RAS Techniques and Instruments , keywords =. doi:10.1093/rasti/rzac004 , adsurl =

work page doi:10.1093/rasti/rzac004
[13]

S., Tennyson, J., Yurchenko, S

ExoMol molecular line lists - XIV. The rotation-vibration spectrum of hot SO _ 2. , keywords =. doi:10.1093/mnras/stw849 , archivePrefix =. 1603.04065 , primaryClass =

work page doi:10.1093/mnras/stw849
[14]

Experimental Astronomy , keywords =

De-jittering Ariel: An optimized algorithm. Experimental Astronomy , keywords =. doi:10.1007/s10686-025-09999-3 , archivePrefix =. 2504.12907 , primaryClass =

work page doi:10.1007/s10686-025-09999-3
[15]

, keywords =

Spectroscopy of planetary atmospheres in our Galaxy. , keywords =. doi:10.1007/s00159-013-0063-6 , adsurl =

work page doi:10.1007/s00159-013-0063-6
[16]

Journal of Astronomical Telescopes, Instruments, and Systems , keywords =

Modeling phase variations introduced by extreme broadband dichroics for astronomical photometry. Journal of Astronomical Telescopes, Instruments, and Systems , keywords =. doi:10.1117/1.JATIS.11.1.014003 , adsurl =

work page doi:10.1117/1.jatis.11.1.014003
[17]

and Bocchieri, Andrea and Pascale, Enzo and Lorenzani, Andrea and Papageorgiou, Andreas , title =

Mugnai, Lorenzo V. and Bocchieri, Andrea and Pascale, Enzo and Lorenzani, Andrea and Papageorgiou, Andreas , title =. Experimental Astronomy , year =. doi:10.1007/s10686-024-09976-2 , url =

work page doi:10.1007/s10686-024-09976-2
[18]

Advances in neural information processing systems , volume=

The marginal value of adaptive gradient methods in machine learning , author=. Advances in neural information processing systems , volume=
[19]

International conference on machine learning , pages=

Train faster, generalize better: Stability of stochastic gradient descent , author=. International conference on machine learning , pages=. 2016 , organization=

2016
[20]

, keywords =

Comparing transit spectroscopy pipelines at the catalogue level: evidence for systematic differences. , keywords =. doi:10.1093/mnras/stae1073 , archivePrefix =. 2402.15803 , primaryClass =

work page doi:10.1093/mnras/stae1073
[21]

, keywords =

Exoplanet transit spectroscopy with JWST NIRSpec: diagnostics and homogeneous case study of WASP-39 b. , keywords =. doi:10.1093/mnras/stae1230 , archivePrefix =. 2405.06737 , primaryClass =

work page doi:10.1093/mnras/stae1230
[22]

B., Mansfield, M., et al

Early Release Science of the exoplanet WASP-39b with JWST NIRCam. , keywords =. doi:10.1038/s41586-022-05590-4 , archivePrefix =. 2211.10489 , primaryClass =

work page doi:10.1038/s41586-022-05590-4
[23]

L., Kyuberis, A

ExoMol molecular line lists XXX: a complete high-accuracy line list for water. , keywords =. doi:10.1093/mnras/sty1877 , archivePrefix =. 1807.04529 , primaryClass =

work page doi:10.1093/mnras/sty1877
[24]

, year = 2010, month = oct, volume =

HITEMP, the high-temperature molecular spectroscopic database. , year = 2010, month = oct, volume =. doi:10.1016/j.jqsrt.2010.05.001 , adsurl =

work page doi:10.1016/j.jqsrt.2010.05.001 2010
[25]

and Amundsen, David S

A hybrid line list for CH _ 4 and hot methane continuum. , keywords =. doi:10.1051/0004-6361/201731026 , archivePrefix =. 1706.05724 , primaryClass =

work page doi:10.1051/0004-6361/201731026
[26]

A., Yurchenko, S

ExoMol molecular line lists - XXXV. A rotation-vibration line list for hot ammonia. , keywords =. doi:10.1093/mnras/stz2778 , archivePrefix =. 1911.10369 , primaryClass =

work page doi:10.1093/mnras/stz2778 1911
[27]

L., Tennyson, J., & Yurchenko, S

ExoMol molecular line lists - XXXVII. Spectra of acetylene. , keywords =. doi:10.1093/mnras/staa229 , archivePrefix =. 2001.04550 , primaryClass =

work page doi:10.1093/mnras/staa229 2001
[28]

” Rothman et al. (2010) Polyansky et al. (2018) “

ExoMol line lists - III. An improved hot rotation-vibration line list for HCN and HNC. , keywords =. doi:10.1093/mnras/stt2011 , archivePrefix =. 1311.1328 , primaryClass =

work page doi:10.1093/mnras/stt2011
[29]

E., Rothman, L

Rovibrational Line Lists for Nine Isotopologues of the CO Molecule in the X ^ 1 ^ + Ground Electronic State. , keywords =. doi:10.1088/0067-0049/216/1/15 , adsurl =

work page doi:10.1088/0067-0049/216/1/15
[30]

Azzam, Ala'a A. A. and Tennyson, Jonathan and Yurchenko, Sergei N. and Naumenko, Olga V. , title = ". Monthly Notices of the Royal Astronomical Society , volume =. 2016 , month =. doi:10.1093/mnras/stw1133 , url =

work page doi:10.1093/mnras/stw1133 2016
[31]

K., Mukherjee, S., et al

Early Release Science of the exoplanet WASP-39b with JWST NIRSpec PRISM. , keywords =. doi:10.1038/s41586-022-05677-y , archivePrefix =. 2211.10487 , primaryClass =

work page doi:10.1038/s41586-022-05677-y
[32]

R., Alam, M

Early Release Science of the exoplanet WASP-39b with JWST NIRSpec G395H. , keywords =. doi:10.1038/s41586-022-05591-3 , archivePrefix =. 2211.10488 , primaryClass =

work page doi:10.1038/s41586-022-05591-3
[33]

D., Radica, M., Welbanks, L., et al

Early Release Science of the exoplanet WASP-39b with JWST NIRISS. , keywords =. doi:10.1038/s41586-022-05674-1 , archivePrefix =. 2211.10493 , primaryClass =

work page doi:10.1038/s41586-022-05674-1
[34]

2023, Monthly Notices of the Royal Astronomical Society, 524, 377, doi: 10.1093/mnras/stad1580

Exoplanet spectroscopy with JWST NIRISS: diagnostics and case studies. , keywords =. doi:10.1093/mnras/stad1580 , archivePrefix =. 2306.04676 , primaryClass =

work page doi:10.1093/mnras/stad1580
[35]

, keywords =

Detection of an Atmosphere on a Rocky Exoplanet. , keywords =. doi:10.3847/1538-3881/abe879 , archivePrefix =. 2103.05657 , primaryClass =

work page doi:10.3847/1538-3881/abe879
[36]

, keywords =

The Featureless HST/WFC3 Transmission Spectrum of the Rocky Exoplanet GJ 1132b: No Evidence for a Cloud-free Primordial Atmosphere and Constraints on Starspot Contamination. , keywords =. doi:10.3847/1538-3881/ac75de , archivePrefix =. 2105.10487 , primaryClass =

work page doi:10.3847/1538-3881/ac75de
[37]

ARES. V. No Evidence For Molecular Absorption in the HST WFC3 Spectrum of GJ 1132 b. , keywords =. doi:10.3847/1538-3881/abf3c3 , archivePrefix =. 2104.01873 , primaryClass =

work page doi:10.3847/1538-3881/abf3c3
[38]

NeurIPS 2024 Competition Track , year=

Ariel Data Challenge 2024: Extracting exoplanetary signals from the Ariel Space Telescope , author=. NeurIPS 2024 Competition Track , year=

2024
[39]

, keywords =

The Impact of Cosmic Rays on the Sensitivity of JWST/NIRSpec. , keywords =. doi:10.1088/1538-3873/ab2fd6 , archivePrefix =. 1907.04051 , primaryClass =

work page doi:10.1088/1538-3873/ab2fd6 1907
[40]

New evolutionary models for pre-main sequence and main sequence low-mass stars down to the hydrogen-burning limit , volume =

Baraffe, Isabelle and Homeier, Derek and Allard, France and Chabrier, Gilles , doi =. New evolutionary models for pre-main sequence and main sequence low-mass stars down to the hydrogen-burning limit , volume =. Astronomy
[41]

V., Pascale, E., Edwards, B., Papageorgiou, A., & Sarkar, S

ArielRad: the Ariel radiometric model. Experimental Astronomy , keywords =. doi:10.1007/s10686-020-09676-7 , archivePrefix =. 2009.07824 , primaryClass =

work page doi:10.1007/s10686-020-09676-7 2009
[42]

, keywords =

TauREx 3: A Fast, Dynamic, and Extendable Framework for Retrievals. , keywords =. doi:10.3847/1538-4357/ac0252 , archivePrefix =. 1912.07759 , primaryClass =

work page doi:10.3847/1538-4357/ac0252 1912
[43]

Mugnai and Enzo Pascale , title =

Andrea Bocchieri and Lorenzo V. Mugnai and Enzo Pascale , title =. Space Telescopes and Instrumentation 2024: Optical, Infrared, and Millimeter Wave , editor =. 2024 , doi =

2024
[44]

arXiv e-prints , keywords =

Observational Techniques With Transiting Exoplanetary Atmospheres. arXiv e-prints , keywords =. doi:10.48550/arXiv.1804.07357 , archivePrefix =. 1804.07357 , primaryClass =

work page doi:10.48550/arxiv.1804.07357
[45]

Subpixel Response Measurement of Near‐Infrared Detectors

Subpixel Response Measurement of Near-Infrared Detectors. , keywords =. doi:10.1086/517620 , archivePrefix =. astro-ph/0611339 , primaryClass =

work page doi:10.1086/517620
[46]

EPSC-DPS Joint Meeting 2019 , year = 2019, volume =

An Updated Study of Potential Targets for Ariel. EPSC-DPS Joint Meeting 2019 , year = 2019, volume =

2019
[47]

SIAM Journal on Imaging Sciences , volume=

Image denoising: The deep learning revolution and beyond—a survey paper , author=. SIAM Journal on Imaging Sciences , volume=. 2023 , publisher=

2023
[48]

Neurocomputing , volume=

A survey of deep learning approaches to image restoration , author=. Neurocomputing , volume=. 2022 , publisher=

2022
[49]

arXiv preprint arXiv:2309.00810 , year=

RenAIssance: A Survey into AI Text-to-Image Generation in the Era of Large Model , author=. arXiv preprint arXiv:2309.00810 , year=

work page arXiv
[50]

IEEE Transactions on Artificial Intelligence , year=

Transformer-based generative adversarial networks in computer vision: A comprehensive survey , author=. IEEE Transactions on Artificial Intelligence , year=
[51]

arXiv preprint arXiv:2303.07909 , year=

Text-to-image diffusion model in generative ai: A survey , author=. arXiv preprint arXiv:2303.07909 , year=

work page arXiv
[52]

, keywords =

Learning to denoise astronomical images with U-nets. , keywords =. doi:10.1093/mnras/staa3567 , archivePrefix =. 2011.07002 , primaryClass =

work page doi:10.1093/mnras/staa3567 2011
[53]

F., Valtchanov, I., Lieu, M., et al

Deep learning-based super-resolution and de-noising for XMM-newton images. , keywords =. doi:10.1093/mnras/stac2437 , archivePrefix =. 2205.01152 , primaryClass =

work page doi:10.1093/mnras/stac2437
[54]

arXiv e-prints , keywords =

Deeper, Sharper, Faster: Application of Efficient Transformer to Galaxy Image Restoration. arXiv e-prints , keywords =. doi:10.48550/arXiv.2404.00102 , archivePrefix =. 2404.00102 , primaryClass =

work page doi:10.48550/arxiv.2404.00102
[55]

The Astronomical Journal , keywords =

Random Forests Applied to High-precision Photometry Analysis with Spitzer IRAC. The Astronomical Journal , keywords =. doi:10.3847/1538-3881/aba11f , archivePrefix =. 2006.14095 , primaryClass =

work page doi:10.3847/1538-3881/aba11f 2006
[56]

The Astronomical Journal , keywords =

Repeatability and Accuracy of Exoplanet Eclipse Depths Measured with Post-cryogenic Spitzer. The Astronomical Journal , keywords =. doi:10.3847/0004-6256/152/2/44 , archivePrefix =. 1601.05101 , primaryClass =

work page doi:10.3847/0004-6256/152/2/44
[57]

RAS Techniques and Instruments , volume=

Lessons learned from the 1st Ariel Machine Learning Challenge: Correcting transiting exoplanet light curves for stellar spots , author=. RAS Techniques and Instruments , volume=. 2023 , publisher=

2023
[58]

The Astronomical Journal , keywords =

Detrending Exoplanetary Transit Light Curves with Long Short-term Memory Networks. The Astronomical Journal , keywords =. doi:10.3847/1538-3881/ab6aa7 , archivePrefix =. 2001.03370 , primaryClass =

work page doi:10.3847/1538-3881/ab6aa7 2001
[59]

Machine Learning for Astrophysics , year = 2022, month = jul, eid =

Don't Pay Attention to the Noise: Learning Self-supervised Representations of Light Curves with a Denoising Time Series Transformer. Machine Learning for Astrophysics , year = 2022, month = jul, eid =. doi:10.48550/arXiv.2207.02777 , archivePrefix =. 2207.02777 , primaryClass =

work page doi:10.48550/arxiv.2207.02777 2022
[60]

2016 , eprint=

Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning , author=. 2016 , eprint=

2016
[61]

and Changeat, Quentin and Morvan, Mario and Al-Refaie, Ahmed F

Yip, Kai Hou and Waldmann, Ingo P. and Changeat, Quentin and Morvan, Mario and Al-Refaie, Ahmed F. and Edwards, Billy and Nikolaou, Nikolaos and Tsiaras, Angelos and de Oliveira, Catarina Alves and Lagage, Pierre-Olivier and Jenner, Clare and Cho, James Y-K. and Thiyagalingam, Jeyan and Tinetti, Giovanna , year =. doi:10.48550/ARXIV.2206.14642 , abstract =

work page doi:10.48550/arxiv.2206.14642
[62]

2023 , pages =

RAS Techniques and Instruments , author =. 2023 , pages =. doi:10.1093/rasti/rzad001 , abstract =

work page doi:10.1093/rasti/rzad001 2023
[63]

P., MacDonald, R

A Comprehensive Reanalysis of K2-18 b's JWST NIRISS+NIRSpec Transmission Spectrum. , keywords =. doi:10.3847/1538-3881/ae019a , archivePrefix =. 2501.18477 , primaryClass =

work page doi:10.3847/1538-3881/ae019a
[64]

Research Notes of the American Astronomical Society , keywords =

Are There Spectral Features in the MIRI/LRS Transmission Spectrum of K2-18b?. Research Notes of the American Astronomical Society , keywords =. doi:10.3847/2515-5172/add881 , archivePrefix =. 2504.15916 , primaryClass =

work page doi:10.3847/2515-5172/add881
[65]

and McGill, Peter and Tilke, Lana J

Challenges in the detection of gases in exoplanet atmospheres. Nature Astronomy , keywords =. doi:10.1038/s41550-025-02730-4 , archivePrefix =. 2504.21788 , primaryClass =

work page doi:10.1038/s41550-025-02730-4
[66]

G., Collins, K

Accurate Empirical Radii and Masses of Planets and Their Host Stars with Gaia Parallaxes. , keywords =. doi:10.3847/1538-3881/aa5df3 , archivePrefix =. 1609.04389 , primaryClass =

work page doi:10.3847/1538-3881/aa5df3