COSMIO: A Benchmark for Cross-Survey Modality Imputation

Abstract

Large-scale astronomical surveys define a heterogeneous, partially observed multi-survey learning setting: each object is observed by only a subset of surveys, while survey observations differ in resolutions, wavelength coverage, noise characteristics, and footprints. This setting gives rise to a natural machine learning problem: given observations of an object in some surveys, how to infer its appearance in other surveys. We formalize this as the problem of cross-survey modality imputation and introduce COSMIO, a multi-survey benchmark of aligned observations obtained by forward-rendered shared astrophysical scenes across LSST, Euclid, and Roman under survey-specific instrument characteristics. The benchmark enables systematic evaluation of imputation across source-target survey combinations and missing-survey patterns. We evaluate a broad range of machine learning methods for imputation, which reveal a clear domain gap: methods developed for existing imputation settings transfer poorly to the survey setting. We identify rare-event imputation as a central challenge. This regime arises in the early stages of new surveys, when rare but scientifically valuable phenomena may be available only from prior surveys. We use strong gravitational lensing, where light from a background galaxy is gravitationally deflected by a foreground galaxy, as a representative rare-event setting. We evaluate model performance on zero-shot imputation of gravitational lenses and on downstream scientific tasks using the imputed lens observations. On our constructed evaluation set of galaxy-galaxy strong lenses, performance degrades under rare-event distribution shift, and this degradation further propagates to the downstream lens-detection task. Code: https://github.com/fw-ic/COSMIO.

The Dyn-Atmo Survey: JWST/NIRSpec spectroscopy of dynamical benchmark GJ 758 B
Alexander Madurowicz, Emily Rickman, Daniella Bardalez Gagliuffi +22
arXiv10 Sep 2026PreprintarXiv licenceFull textInstrumentation & methods Planetary science & exoplanetsarXiv
The XMM Cluster Survey: Cross-Calibration of eROSITA and XMM using Dark Energy Survey Y3 Galaxy Clusters
J E Pilling, D J Turner, P A Giles +9
Monthly Notices of the Royal Astronomical Society16 Oct 2026Peer-reviewedCC BY 4.0Full textGalaxiesCrossref
The XMM Cluster Survey: Cross-Calibration of eROSITA and XMM using Dark Energy Survey Y3 Galaxy Clusters
J. E. Pilling, D. J. Turner, P. A. Giles +9
arXiv12 Oct 2026PreprintCC BY 4.0Full textCosmology General physicsarXivHAL
Cosmology with the angular cross-correlation of gravitational-wave and galaxy catalogs: Forecasts for next-generation interferometers and the Euclid survey
Alessandro Pedrotti, Michele Mancarella, J. Bel +2
Astronomy and Astrophysics19 Jun 2026Peer-reviewedCC BY19 citedCosmologyOpenAlex
Gaia FGK Benchmark Stars: Impact of Spectral Resolution on Stellar Abundances
Ivanna Hernández-Araya, Claudia Aguilera-Gómez, Paula Jofré +5
HAL21 Oct 2026PreprintLicence not statedFull textSolar & stellar physicsHAL

Cite

BibTeX RIS

BibTeX

@misc{zhang2026cosmio,
  title = {{COSMIO: A Benchmark for Cross-Survey Modality Imputation}},
  author = {Dichang Zhang and Yixuan Shao and Jiaqi Deng and Jiali Cui and Haotian Yin and Yuanpeng Liu and Xiang Gao and Zhiqiang Lao and Heather Yu and Simon Birrer and Dimitris Samaras},
  howpublished = {arXiv},
  year = {2026},
  month = nov,
  doi = {10.48550/arxiv.2611.04139},
  eprint = {2611.04139},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.04139},
  note = {CC BY-SA 4.0},
}

RIS

TY  - UNPB
TI  - COSMIO: A Benchmark for Cross-Survey Modality Imputation
AU  - Dichang Zhang
AU  - Yixuan Shao
AU  - Jiaqi Deng
AU  - Jiali Cui
AU  - Haotian Yin
AU  - Yuanpeng Liu
AU  - Xiang Gao
AU  - Zhiqiang Lao
AU  - Heather Yu
AU  - Simon Birrer
AU  - Dimitris Samaras
PY  - 2026
DA  - 2026/11/03
PB  - arXiv
DO  - 10.48550/arxiv.2611.04139
UR  - https://arxiv.org/abs/2611.04139
AB  - Large-scale astronomical surveys define a heterogeneous, partially observed multi-survey learning setting: each object is observed by only a subset of surveys, while survey observations differ in resolutions, wavelength coverage, noise characteristics, and footprints. This setting gives rise to a natural machine learning problem: given observations of an object in some surveys, how to infer its appearance in other surveys. We formalize this as the problem of cross-survey modality imputation and introduce COSMIO, a multi-survey benchmark of aligned observations obtained by forward-rendered shared astrophysical scenes across LSST, Euclid, and Roman under survey-specific instrument characteristics. The benchmark enables systematic evaluation of imputation across source-target survey combinations and missing-survey patterns. We evaluate a broad range of machine learning methods for imputation, which reveal a clear domain gap: methods developed for existing imputation settings transfer poorly to the survey setting. We identify rare-event imputation as a central challenge. This regime arises in the early stages of new surveys, when rare but scientifically valuable phenomena may be available only from prior surveys. We use strong gravitational lensing, where light from a background galaxy is gravitationally deflected by a foreground galaxy, as a representative rare-event setting. We evaluate model performance on zero-shot imputation of gravitational lenses and on downstream scientific tasks using the imputed lens observations. On our constructed evaluation set of galaxy-galaxy strong lenses, performance degrades under rare-event distribution shift, and this degradation further propagates to the downstream lens-detection task. Code: https://github.com/fw-ic/COSMIO.
ER  -

Abstract

Related

Cite