arXiv Science⌕ Search

arXiv · 2611.10228

FAIR Universe Weak Lensing ML Uncertainty Challenge: Dataset and Cosmological Inference Competition

Abstract

Weak gravitational lensing is a powerful cosmological probe that enables direct mapping of the matter distribution and precise constraints on the fundamental parameters governing structure formation in the universe. In recent years, machine learning has shown strong potential for extracting cosmological information from weak-lensing data beyond traditional summary statistics. However, progress has been limited by the lack of standardized benchmarks that capture the complexity of the dataset, simulator imperfections, and various systematic biases encountered in realistic analyses. To address this gap, we present the FAIR Universe Weak Lensing Machine Learning Uncertainty Challenge, one of the NeurIPS 2025 competitions, as a benchmark dataset and evaluation study. Our dataset reflects realistic weak-lensing observations, incorporating multiple systematic biases, and thereby providing a controlled testbed for robust cosmological inference under uncertainties. We summarize the main methodological advances developed by the winners of our challenge, including data augmentation with conditional optimal transport, an AI-agent-driven approach for parameter inference, ensemble learning, Vision Transformer-based architectures, and physics-informed feature integration. By unifying the benchmark dataset and the winning solutions, this work provides a community resource for developing and evaluating machine learning methods for weak lensing cosmology and, more broadly, for scientific inference under systematic uncertainties.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Biwei Dai, Po-Wen Chang, Erwan Allys, Wahid Bhimji, Boris Bolliet, Thomas Borrett, Paolo Calafiura, Ragansu Chakkappai, Yuan-Tang Chou, Noé Dia, Sascha Diefenbacher, Ibrahim Elsharkawy, Steven Farrell, Sacha Guerrini, Isabelle Guyon, Chris Harris, Shih-Chieh Hsu, Wassim Kablan, François Lanusse, Celia Lecat, Julia Linhart, Yuezhou Ma, Benjamin Nachman, Shubhojit Naskar, Andy Nilipour, Laurence Perreault-Levasseur, Sébastien Pierre, Benjamin Remy, David Rousseau, Uroš Seljak, Haonan Shangguan, Sammy Nasser Sharief, Yuanxu Sun, Andreas Tersenov, Ihsan Ullah, Huikun Weng, Licong Xu, Justine Zeghal, Yulei Zhang, Hang Zhou. 2026-11-06. FAIR Universe Weak Lensing ML Uncertainty Challenge: Dataset and Cosmological Inference Competition. https://arxiv.org/abs/2611.10228

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

An information-theoretic framework for quantifying coherence in the cosmic density field

We introduce an information-theoretic framework for quantifying the coherence retained in the cosmic density field across cosmic time. Using mutual information between density fields at different redshifts in cosmological $N$-body simulations, and subtracting an appropriate reference baseline, we isolate the excess statistical dependence generated by gravitational evolution. Unlike conventional clustering statistics, this approach is sensitive to nonlinear and higher-order correlations and directly measures the shared information between different cosmic epochs. The shared mutual information exhibits a strong dependence on spatial scale, increasing as the density field is progressively coarse-grained, reaching a maximum, and declining at larger scales. Through a series of tests involving different simulation volumes, discretization schemes, alternative baseline constructions, and subvolume analyses, we show that the location of this maximum is governed primarily by finite-sampling effects in the mutual information estimator rather than by a fundamental physical coherence scale. The observed peak therefore provides an empirical indication of the range over which the present estimator yields robust measurements. We further demonstrate that the mutual information is sensitive to cosmic evolution and cosmological parameters. Our analysis shows that filamentary environments retain more statistically accessible shared information between cosmic epochs than sheets and clusters. We also note that the shared information between the initial and present-day density fields is found to be positive but extremely small, reflecting the progressive redistribution of information during nonlinear structure formation. Our results establish mutual information as a powerful new tool for quantifying coherence, memory, and information retention in the large-scale structure of the Universe.

astro-ph.CO↗

An efficient model of cosmology dependence in the covariance matrix of the matter power spectrum

Covariance matrices are essential for any cosmological likelihood analysis. They can be estimated from large suites of cosmological simulations, but this is computationally intensive, so various approximations are often adopted to simplify the analysis, such as treating covariance as being parameter independent to avoid having to compute matrices for every variation in physical parameters. In this paper, we develop a new approach to model the parameter dependence of the covariance matrix of the matter power spectrum. We do this by Taylor expanding the `true covariance matrix' around a fiducial point in parameter space projected onto a theoretically known precision matrix. In the limit that the known model approaches the true matrix then the Taylor expansion truncates at the first term; for more realistic models, additional terms are required. We use the publicly available Quijote suite of simulations, giving us both a fiducial model and variations in the six parameters $\{Ω_{\rm m}, Ω_{\rm b}, n_{\rm s},σ_{8}, h,w_{0}\}$. Using the linear power spectra from these simulations, we generate model covariance matrices with and without super-sample covariance. Using different approximations, we develop three versions of our interpolated covariance matrix and validate the formalism by recreating all of our input cosmologies using all three forms, both with and without super-sample covariance corrections, showing that the model provides good recreations of the original results, with the different approximations being optimal in different regimes. Our method enables rapid calculation of parameter dependent covariance matrices, and consequently precision matrices, for deployment in likelihood modelling codes.

astro-ph.CO↗

Model-independent Constraints on Dark Energy Density from CMB-SPA, DESI DR2 BAO, and Updated Type Ia Supernova Compilations

Recent combinations of DESI BAO, CMB, and Type~Ia supernova data report a preference for evolving dark energy that is obtained within the $(w_0, w_a)$ parameterization and depends on the choice of SN~Ia compilation. We constrain the dark energy density ratio $X(z) \equiv ρ_{\rm DE}(z)/ρ_{\rm DE}(0)$ at five redshifts over $0 < z < 2.33$, without assuming a parametric form for the equation of state, using CMB data from SPT-3G, Planck~2018, and ACT~DR6 (hereafter CMB-SPA), DESI~DR2 BAO, and the updated Pantheon+, Union3.1, and DES-Dovekie SN~Ia compilations. For all three compilations, $X(z)$ rises above unity at low redshift, crosses unity near $z \approx 1$, and falls below unity at higher redshift. Depending on the SN~Ia compilation, the preference for a departure from a cosmological constant is $2.3$--$2.8σ$ with the five $X(z)$ knots, compared with $3.0$--$3.2σ$ in the $(w_0, w_a)$ parameterization. Because $X(z)$ is constrained locally in redshift, we can trace how the $Λ$CDM tension between each pair of probes propagates into its shape. The tension between BAO and the CMB in the $h r_d$--$Ω_m$ plane reproduces this "see-saw" shape, whereas the $Ω_m$ tension between SNe~Ia and the CMB only raises $X(z)$ slightly and nearly uniformly. The origin of the BAO--CMB tension can be tested by Euclid and Roman BAO measurements at $z \gtrsim 1$. We derive the distance prior $(R, \ell_A, ω_b)$ from CMB-SPA data and show that it retains essentially all the information on $X(z)$ in the CMB-SPA power spectra without the lensing reconstruction, and that adding the lensing reconstruction only marginally increases the information on dark energy. Our work demonstrates the importance of model independence in exploring the nature of dark energy.

astro-ph.CO↗