https://xin-yuxiao.github.io/Awesome-AI-for-Astronomy/today.htmlNew arXiv papers that apply machine learning to astronomy, with one-line TL;DR summaries.https://arxiv.org/abs/2611.13118https://arxiv.org/abs/2611.13118Wed, 11 Nov 2026 06:00:00 +0000TL;DR Language model agents scan 156,006 arXiv astro-ph papers to extract 44,332 discussions and quotes for 13,616 galaxies, building GalaxyWiki.https://arxiv.org/abs/2611.10228https://arxiv.org/abs/2611.10228Tue, 10 Nov 2026 06:00:00 +0000TL;DR Presents the FAIR Universe weak-lensing ML uncertainty challenge, with a realistic benchmark dataset containing systematics and the winning solutions, including vision transformers, ensembles and AI agents.https://arxiv.org/abs/2611.10861https://arxiv.org/abs/2611.10861Tue, 10 Nov 2026 06:00:00 +0000TL;DR Introduces SWXAI, an open-source library unifying explainable-AI workflows for images, time series and tabular data in space weather and heliophysics forecasting models.https://arxiv.org/abs/2611.11922https://arxiv.org/abs/2611.11922Tue, 10 Nov 2026 06:00:00 +0000TL;DR Combines neural-network-guided parameter-space localization with particle swarm optimization and MCMC to recover EMRI posteriors from simulated six-month signals without prior knowledge of true parameters.https://arxiv.org/abs/2611.07174https://arxiv.org/abs/2611.07174Tue, 10 Nov 2026 06:00:00 +0000TL;DR An LLM scores daily arXiv submissions against researcher portfolios and publishes relevance heatmaps, placing about 70% of blind-rated top papers in its top decile.https://arxiv.org/abs/2611.10132https://arxiv.org/abs/2611.10132Tue, 10 Nov 2026 06:00:00 +0000TL;DR An artificial neural network trained on non-LTE spectra derives Cepheid atmospheric parameters and 17 elemental abundances, producing Milky Way radial abundance gradients for 4MOST.https://arxiv.org/abs/2611.10133https://arxiv.org/abs/2611.10133Tue, 10 Nov 2026 06:00:00 +0000TL;DR CNN on short-time Fourier transforms detects binary neutron-star mergers hours before merger in Einstein Telescope simulations, matching matched filtering at four orders lower computational cost.https://arxiv.org/abs/2611.12787https://arxiv.org/abs/2611.12787Tue, 10 Nov 2026 06:00:00 +0000TL;DR A learned Martian atmosphere model alternates two-hour forecasts with bounded Kalman observation updates, matching reanalyses near the surface and improving surface pressures on withheld observations.https://arxiv.org/abs/2611.08583https://arxiv.org/abs/2611.08583Mon, 09 Nov 2026 06:00:00 +0000TL;DR An LLM pipeline extracts 6,145 dark-matter model cards from arXiv full texts, validates schemas, and maps high-priority infrastructure gaps.https://arxiv.org/abs/2611.08880https://arxiv.org/abs/2611.08880Mon, 09 Nov 2026 06:00:00 +0000TL;DR A search plus machine-learning classifier distinguishes gravitational-wave transients from instrumental glitches in simulated LISA data with about 98% accuracy and fast processing.https://arxiv.org/abs/2611.09198https://arxiv.org/abs/2611.09198Mon, 09 Nov 2026 06:00:00 +0000TL;DR New exact Kronecker-product Gaussian process inference in JAX enables scalable 2D modelling of correlated noise in JWST exoplanet spectroscopy and other astronomical time-series.https://arxiv.org/abs/2611.07162https://arxiv.org/abs/2611.07162Fri, 06 Nov 2026 06:00:00 +0000TL;DR Random forest models trained on XSTAR outputs predict line-driving force multipliers for Athena++ radiation hydrodynamics, outperforming trilinear interpolation across ionization, optical depth, and SED temperature.https://arxiv.org/abs/2611.07433https://arxiv.org/abs/2611.07433Fri, 06 Nov 2026 06:00:00 +0000TL;DR MP-DualNet, a dual-branch deep learning model, detects magnetopause boundaries in simulated SXI images, achieving >0.94 classification accuracy and >0.97 segmentation IoU.https://arxiv.org/abs/2611.07936https://arxiv.org/abs/2611.07936Fri, 06 Nov 2026 06:00:00 +0000TL;DR Machine-learning regression homogenises effective temperature, surface gravity, and metallicity across nine spectroscopic surveys, producing a 10.13-million-star catalogue with reduced survey offsets.https://arxiv.org/abs/2611.08634https://arxiv.org/abs/2611.08634Thu, 05 Nov 2026 06:00:00 +0000TL;DR Benchmarks regression models and physics foundation models on simulated chaotic three-body systems: early stability classification is accurate, but precise long-horizon trajectory prediction remains difficult.https://arxiv.org/abs/2611.05447https://arxiv.org/abs/2611.05447Thu, 05 Nov 2026 06:00:00 +0000TL;DR Adds a bounded output layer and flux-aware LoRA fine-tuning to Noise2Void denoising of PAUS images, reducing repeated-exposure flux scatter by 47% while keeping fluxes unbiased.https://arxiv.org/abs/2611.05790https://arxiv.org/abs/2611.05790Thu, 05 Nov 2026 06:00:00 +0000TL;DR These solvers enable scalable EoS inference that self-consistently accounts for rotation and repeated stellar-structure calculations during binary evolution, making spin predictions that account for rotational structure and EoS uncertainties computationally accessible to population synthesis.https://arxiv.org/abs/2611.06026https://arxiv.org/abs/2611.06026Thu, 05 Nov 2026 06:00:00 +0000TL;DR A cross-modal framework transferring NIRSpec spectroscopic information into NIRCam imaging representations finds 35 new Little Red Dot candidates and hundreds more in archival JWST imaging.https://arxiv.org/abs/2611.06190https://arxiv.org/abs/2611.06190Thu, 05 Nov 2026 06:00:00 +0000TL;DR The ML-based pipeline provides a scalable way to produce large catalogues for the ongoing EMU survey and future Square Kilometre Array observations and reveals that multi-wavelength analysis can uncover radio populations and trends often missed by colour cuts or spectroscopy alone.https://arxiv.org/abs/2611.06484https://arxiv.org/abs/2611.06484Thu, 05 Nov 2026 06:00:00 +0000TL;DR A 370-object three-band imaging benchmark with group-constrained splits shows CNN-Transformer, ResNet-50 and ViT-B/16 perform similarly, and that acquisition and catalogue metadata leak class information.https://arxiv.org/abs/2610.37287https://arxiv.org/abs/2610.37287Wed, 04 Nov 2026 06:00:00 +0000TL;DR Deep-learning pipeline using U-Net and DNNs on simulated Fermi-LAT data produces first all-sky gamma-ray source list of >3800 sources.https://arxiv.org/abs/2611.04139https://arxiv.org/abs/2611.04139Wed, 04 Nov 2026 06:00:00 +0000TL;DR COSMIO benchmark forward-renders aligned LSST, Euclid, Roman scenes to test cross-survey modality imputation, revealing poor transfer and rare-event lens challenges.https://arxiv.org/abs/2611.04389https://arxiv.org/abs/2611.04389Wed, 04 Nov 2026 06:00:00 +0000TL;DR Spin spherical Fourier neural operators extend SFNOs to equivariant vector and spin fields on the sphere, matching machine-precision equivariance and state-of-the-art benchmarks.https://arxiv.org/abs/2611.04664https://arxiv.org/abs/2611.04664Wed, 04 Nov 2026 06:00:00 +0000TL;DR Comparing feature-based ML and end-to-end deep learning on 40,000 ZTF light curves, classical features outperform deep networks for variable star classification.https://arxiv.org/abs/2611.05157https://arxiv.org/abs/2611.05157Wed, 04 Nov 2026 06:00:00 +0000TL;DR Macrocanonical generator networks learn data-efficient neural surrogates by matching multiscale statistics, recovering cosmology and fluid simulations from tiny training sets over 10^6 times faster.https://arxiv.org/abs/2611.05251https://arxiv.org/abs/2611.05251Wed, 04 Nov 2026 06:00:00 +0000TL;DR Semi-autonomous LLM multi-agent pipeline retrieves HEASARC data, writes and validates tools with provenance checks, reproducing IXPE Cyg X-1 results.https://arxiv.org/abs/2611.04484https://arxiv.org/abs/2611.04484Wed, 04 Nov 2026 06:00:00 +0000TL;DR CNN-derived position angles for 4087 SUMSS radio sources show no statistically significant local alignments after look-elsewhere calibration, providing a southern-sky morphology catalogue.https://arxiv.org/abs/2611.04713https://arxiv.org/abs/2611.04713Wed, 04 Nov 2026 06:00:00 +0000TL;DR Deep-learning generator produces black-hole, neutron-star, and mixed-binary waveforms including spin, matter, and postmerger effects, with mismatches 5.9e-4–5.3e-2 and up to 1800× speedup.https://arxiv.org/abs/2611.05367https://arxiv.org/abs/2611.05367Wed, 04 Nov 2026 06:00:00 +0000TL;DR Simulation-based inference of integrated three-point cosmic shear from DES Y3, combined with two-point functions, tightens cosmological constraints and yields Ωm=0.272, S8=0.784, w0=-1.04.https://arxiv.org/abs/2611.05309https://arxiv.org/abs/2611.05309Tue, 03 Nov 2026 06:00:00 +0000TL;DR Compact neural architecture with staged spectral token refinement predicts gravitational-wave signal-to-noise ratios for precessing binary black hole signals, reaching high accuracy and about 100-fold speedup.https://arxiv.org/abs/2611.00121https://arxiv.org/abs/2611.00121Tue, 03 Nov 2026 06:00:00 +0000TL;DR Conditional normalizing flows trained on DESI DR1 and Legacy Surveys photometry reconstruct tomographic redshift distributions, outperforming random-forest baselines at high redshift.https://arxiv.org/abs/2611.00150https://arxiv.org/abs/2611.00150Tue, 03 Nov 2026 06:00:00 +0000TL;DR Two-stage contrastive adaptation of ASTROMER light-curve representations improves frozen-encoder classification, reaching macro-F1 0.794 with 500 labeled examples per class.https://arxiv.org/abs/2611.00701https://arxiv.org/abs/2611.00701Tue, 03 Nov 2026 06:00:00 +0000TL;DR Physics-informed neural networks boosted by tree ensembles and support vector regression solve Lane-Emden equations for stellar-structure models, reducing error versus standard PINN baselines.https://arxiv.org/abs/2611.01356https://arxiv.org/abs/2611.01356Tue, 03 Nov 2026 06:00:00 +0000TL;DR Convolutional UNet, shifted-window transformer and Fourier neural operator surrogates emulate three-dimensional resistive MHD turbulence, matching simulation statistics and outperforming published benchmarks.https://arxiv.org/abs/2611.01988https://arxiv.org/abs/2611.01988Tue, 03 Nov 2026 06:00:00 +0000TL;DR Community hackathon evaluates autonomous systems, including language models, for allocating observing time in a synthetic multi-object survey with faults and competing priorities.https://arxiv.org/abs/2611.03234https://arxiv.org/abs/2611.03234Tue, 03 Nov 2026 06:00:00 +0000TL;DR Two-phase deep-learning pipeline forecasts monthly and next-day solar flare classes from GOES X-ray data, achieving skill but sometimes matching persistence.https://arxiv.org/abs/2611.03450https://arxiv.org/abs/2611.03450Tue, 03 Nov 2026 06:00:00 +0000TL;DR Physics-informed neural network incorporating external pressure reclassifies 2,091 Milky Way molecular clouds as pressure-confined and links stability to star formation.https://arxiv.org/abs/2611.00415https://arxiv.org/abs/2611.00415Tue, 03 Nov 2026 06:00:00 +0000TL;DR Normalizing-flow combination of gravitational-wave standard sirens with DES, BAO, supernovae, and CMB probes finds S8=0.793 and H0=69.4, leaving S8 and H0 tensions essentially unchanged.https://arxiv.org/abs/2611.01310https://arxiv.org/abs/2611.01310Tue, 03 Nov 2026 06:00:00 +0000TL;DR Differentiable physics-informed model maps SDO images and magnetograms to the 21.5 solar-radii wind boundary, improving >600 km/s forecasting over recurrence and linear baselines.https://arxiv.org/abs/2611.02064https://arxiv.org/abs/2611.02064Tue, 03 Nov 2026 06:00:00 +0000TL;DR XGBoost classification and SHAP attribution of Monte Carlo X-ray burst models identify competing (p,α) and (p,γ) reactions on 59Cu and 63Ga as the main source of secondary abundance peaks.https://arxiv.org/abs/2611.02627https://arxiv.org/abs/2611.02627Tue, 03 Nov 2026 06:00:00 +0000TL;DR Physics-informed latent corrections to a simulation-trained continuous normalizing flow recover variability lost to variance shrinkage in weak-lensing field-level inference, improving robustness to out-of-distribution shifts.