arXiv subjects

Francisco Villaescusa-Navarro

Publications and source records attributed to Francisco Villaescusa-Navarro.

At least 19 recordsLinked to original sources

arXiv2026

AI-assisted super-resolution cosmological simulations V: Cosmology-aware super-resolution

We present a cosmology-aware generative model for super-resolution (SR) of cosmological $N$-body simulations in particle phase space. The model jointly generates displacements and velocities, conditioning a generative adversarial network (GAN) based generator and discriminator on evolved low-resolution (LR) fields and five $Λ$CDM parameters $(Ω_m,Ω_b,h,n_s,σ_8)$. We train a single model on 100 matched low-high resolution pairs from the Quijote Latin-hypercube suite at $z = 0$, mapping $512^3$ to $1024^3$ particles in $1\,h^{-1}\,\mathrm{Gpc}$ boxes. On ten cosmologies excluded from SR training, the model reproduces the broad matter distribution and the cosmological variation of clustering and halo abundances. Over $0.1<k<3\,h\,\mathrm{Mpc}^{-1}$, the mean of the per-cosmology maximum absolute power-spectrum errors is $3.6\%$, while FoF halo abundances agree with HR to within $15\%$ in the analyzed mass bins below $10^{14.8}\,h^{-1}M_\odot$. Satellite abundances and occupations retain larger, mass-dependent discrepancies. Corrections measured at one near-fiducial cosmology reduce a shared component of these systematic errors when applied to the other nine test cosmologies. A power spectrum based inference test also shows that calibration reduces the mean shifts in all parameters. These results extend our SR model across cosmologies and provide 6D phase space realizations for subsequent studies of non-linear structure and field-level cosmological inference.

astro-ph.CO↗

arXiv2026

EurekaBench: Measuring Agentic Ability to Discover New Scientific Insights

When Isaac Newton discovered the law of gravitation, he did so through an iterative process of analyzing observed data such as planetary patterns, finding the underlying mechanisms by describing patterns in mathematical equations, and refining his theory against the Moon's orbit, revealing the startling insight that the same force governs both falling apples and orbiting planets. Would it be possible for AI agents to make similar discoveries? To measure this ability, we introduce EurekaBench, a cross-domain benchmark that tests AI agents' ability to conduct long-horizon experiments and discover mechanisms that explain observations. We evaluate these mechanisms by the scientific insights that can be derived from them. EurekaBench contains an expert-verified set of 26 long-horizon tasks across neuroscience, computer science, chemistry, astrophysics, geophysics, and plasma physics, with a total of 306 scientific insights that the discovered mechanisms are expected to support. Our evaluation framework tests three axes of scientific discovery: agents' ability to follow known scientific constraints, the predictive accuracy of the discovered mechanisms, and whether these mechanisms yield scientific insights or inform future research. Our results show that current AI agents often overly fixate on predictive accuracy optimization, surpassing human scientists, while falling substantially short in deriving scientific insights.

cs.CL↗

arXiv2026

Efficiently emulating distribution functions in gigaparsec volumes for varying cosmological parameters

We present a new method for emulating the halo mass function (HMF) and other distribution functions in large effective volumes, down to low halo masses, whilst simultaneously modifying large ranges of parameters, for a fraction of the cost of traditional periodic cosmological simulations. We demonstrate the method by selecting small regions, $V \sim (50 \,h^{-1}{\rm Mpc})^3$, with a range of overdensities from the Quijote suite, consisting of tens of thousands of $(1 \,h^{-1}{\rm Gpc})^3$ $N$-body simulation volumes run with varying $Λ$CDM parameters. We train a differentiable emulator, conditioned on the overdensity of the region and these global parameters, to reproduce the halo mass function in these regions. We then successfully recover the global distribution of halo masses of the entire box by integrating over the overdensity distribution. Our approach uses just $\sim\,$0.026% of the original simulation volume, and suggests that suites of targeted `zoom' simulations, extracted from low resolution parent volumes, can be used to emulate large volume simulations at a fraction of the computational cost, whilst simultaneously pushing the dynamic range to much lower masses than can be achieved in periodic simulations. We discuss emulation of other key dark matter and baryonic distribution functions, as well as higher order statistics, with implications for the interpretation of upcoming wide field surveys on observatories such as Euclid, Roman and Rubin.

astro-ph.CO↗

arXiv2026

FOTO: Figure-Level Semantic Search for Astronomy

Astronomy papers carry much of their content in figures, but literature search indexes text, so there is no way to find a plot by describing what it shows. We present FOTO, a figure-level search tool over 482,750 captions from 53,839 peer-reviewed astro-ph papers. Each figure is indexed once by its title and caption, embedded with a 110M-parameter open model that runs locally, and a large language model is applied only to the retrieved shortlist, where it verifies real figures rather than generating references to ones that do not exist. Held to the harder figure-level criterion, FOTO recovers the target for 29 to 79% of queries at recall 20 depending on register, against 6 to 20% for Pathfinder and 8 to 16% for Semantic Scholar on the easier paper-level criterion, and it beats the paid embedding API it replaced. Reranking the shortlist then raises recall at 1 from 0.667 to 0.875 on detailed queries and from 0.071 to 0.571 on vague ones. Openly available models match or beat paid LLM models for verification

astro-ph.IM↗

arXiv2026

The CAMELS-CROCODILE Simulation Suite: A New Cosmology--Astrophysics Playground for Machine Learning

We present CAMELS-CROCODILE, a suite of cosmological hydrodynamic simulations that extends the CAMELS framework with the \gadget{} smoothed particle hydrodynamics code and the Osaka feedback model. As a companion to the IllustrisTNG-based second-generation CAMELS suite, it provides an independent implementation of the baryonic physics that shapes galaxies, the intergalactic medium, and large-scale structure, broadening the range of subgrid models over which machine-learning (ML) inference can be marginalized and tested. The suite comprises $25$ and $50Mpc/h$ boxes organized into one-parameter, cosmic-variance, and Sobol-sequence sets that vary $26$ cosmological and astrophysical parameters around a fiducial Osaka model. We describe the simulation design and validate the suite against benchmarks and other CAMELS suites. Compared with IllustrisTNG, the fiducial model suppresses the matter power spectrum by only a few per cent, against up to ${\sim}27\%$ at $z=0$, and forms stars with similar efficiency below $10^{12.2} M_\odot/h$ but up to $2.5$ times more efficiently in group-scale halos, both consistent with weaker AGN feedback. We also present the stellar and black hole bias and the Ly$α$ absorption around galaxies. As a first ML application, we apply a graph neural network trained on IllustrisTNG galaxy catalogs, with frozen weights, to CAMELS-CROCODILE. It fails to recover $Ω_{\rm m}$ and $σ_8$ and returns overconfident posteriors, mainly for simulations containing more galaxies than any in its training set; within that range $Ω_{\rm m}$ is recovered with an error only $1.5$ times the in-distribution value. These results underline the need to train and test ML inference across physically distinct galaxy formation models.

astro-ph.GA↗

arXiv2026

Inductive Biases in Field-Level Cosmological Inference from Galaxy Catalogs

We perform field-level likelihood-free inference of the matter density parameter $Ω_m$ from simulated galaxy catalogs using machine learning models with differing inductive biases. Using hydrodynamic simulations from CAMELS, we examine how observable choice and architecture govern cosmological information extraction. We consider galaxy positions and line-of-sight peculiar velocities, separately and jointly, and compare permutation-invariant Deep Sets, implemented with either multilayer perceptrons (MLPs) or Kolmogorov-Arnold Networks (KANs), to graph neural networks (GNNs), which explicitly encode spatial relations. We test in-distribution and out-of-distribution (OOD) performance across simulations with different subgrid galaxy-formation prescriptions. Deep Sets infer $Ω_m$ from velocities alone with mean relative errors of approximately $18\%$ in-distribution and $\sim25\%$ OOD, with KANs and MLPs achieving comparable performance. In contrast, the same set-based approach does not yield useful $σ_8$ predictions in either in-distribution or cross-suite tests. Adding positions does not improve Deep Sets, while GNNs infer $Ω_m$ with mean relative errors of about $10\%$ in-distribution and $10$--$17\%$ OOD. These results indicate that peculiar velocities provide the dominant source of $Ω_m$ information for set-based models in this setting, while spatial information is most effectively used by architectures that explicitly encode galaxy-galaxy relations. Because the velocity inputs are exact simulated peculiar velocities, applications to survey data will require validation under realistic velocity-measurement noise, selection effects, and survey geometry.

astro-ph.CO↗

arXiv2026

Galaxy Phase-Space and Field-Level Cosmology: The Strength of Semi-Analytic Models

Semi-analytic models are a widely used approach to simulate galaxy properties within a cosmological framework, relying on simplified yet physically motivated prescriptions. They have also proven to be an efficient alternative for generating accurate galaxy catalogs, offering a faster and less computationally expensive option compared to full hydrodynamical simulations. In this paper, we demonstrate that using only galaxy $3$D positions and radial velocities, we can train a graph neural network coupled to a moment neural network to obtain a robust machine learning based model capable of estimating the matter density parameters, $Ω_{\rm m}$, with a precision of approximately 10%. The network is trained on ($25 h^{-1}$Mpc)$^3$ volumes of galaxy catalogs from L-Galaxies and can successfully extrapolate its predictions to other semi-analytic models (GAEA, SC-SAM, and Shark) and, more remarkably, to hydrodynamical simulations (Astrid, SIMBA, IllustrisTNG, and SWIFT-EAGLE). Our results show that the network is robust to variations in astrophysical and subgrid physics, cosmological and astrophysical parameters, and the different halo-profile treatments used across simulations. This suggests that the physical relationships encoded in the phase-space of semi-analytic models are largely independent of their specific physical prescriptions, reinforcing their potential as tools for the generation of realistic mock catalogs for cosmological parameter inference.

astro-ph.CO↗

arXiv2026

Deep Learning for Astrophysics: An Open Textbook from the NASA Cosmic Origins AI/ML Science and Technology Interest Group

Recent community assessments identify education as a principal barrier to adopting modern machine learning in astronomy. We present Deep Learning for Astrophysics, a freely available textbook at https://deeplearning4astro.com, curated from the NASA Cosmic Origins Artificial Intelligence and Machine Learning Science and Technology Interest Group (AI/ML STIG) lecture series. The book collects 23 chapters by 17 lecturers across six parts, moving from computational foundations and deep-learning architectures through generative modeling, simulation-based inference, reinforcement learning, and large-language-model agents to the practice of AI-laden science. Many include executable notebooks using astronomical data.

astro-ph.IM↗

arXiv2026

Learning Cosmology from Nearest Neighbour Statistics

Extracting cosmological parameters from galaxy/halo catalogues with sub-percent level accuracy is an important aspect of modern cosmology, especially in view of ongoing and upcoming surveys such as Euclid, DESI, and LSST. While traditional two-point statistics have been known to be suboptimal for this task, recently proposed k-Nearest Neighbour (kNN) based summary statistics have demonstrated tighter constraining power. Building on the kNN statistics, we introduce a new field-level representation of discrete halo catalogues - NN distance maps. We employ this technique on the halo catalogues obtained from Quijote N-body simulation suites. By combining these maps with kNN-based summary statistics, we train a hybrid neural network to infer cosmological parameters, showing that the resulting constraints achieve state-of-the-art, if not the best, accuracy. In addition, our hybrid framework is 5-10 times more computationally efficient than some of the existing point-cloud-based ML methods.

astro-ph.CO↗

arXiv2026

AI Scientists as Engines of Discovery: A Case for Development within Reformed Institutions

Agentic artificial intelligence (AI) systems are beginning to assist, accelerate, and partially automate scientific discovery, performing tasks that span literature synthesis, code generation, data analysis, hypothesis proposal, and model criticism. We argue that this transition is qualitative rather than incremental, and that suitably designed multi-agent systems may evolve from passive computational tools into ``AI scientists'' that can expand the hypothesis-generating and verification capacity of science. Such systems must be developed and deployed within a scientific ecosystem fit for purpose: institutions must be redesigned for verification, accountability, interpretability, and dual-use safety. We sketch how multi-agent architectures, illustrated by the prototype framework \textit{Denario}, accelerate the discovery cycle and traverse model spaces beyond human reach; examine what this implies for authorship, peer review, and the enduring role of human scientists; and close with recommendations for governing AI as an epistemic actor rather than a mere instrument.

cs.AI↗

arXiv2026

Learning the Universe with the 2nd Generation of CAMELS: Varying 35 parameters of the IllustrisTNG model in (50Mpc/h)^3 boxes

We present a new set of 1,192 cosmological simulations as part of the CAMELS project, in which a space of 35 cosmological, astrophysical, and numerical parameters is explored around the fiducial IllustrisTNG model. The volume of each of these simulations is (50Mpc/h)^3, eight times larger than that of previous CAMELS simulations. This provides lower sample variance as well as access to more massive halos and more diverse environments. We focus this work on exploring the advantages these differences provide for parameter inference powered by neural networks. We generate training sets based on the matter power spectra, projected maps of the volumes, graphs representing galaxy spatial distributions, and thermodynamical properties of massive halos. We employ multilayer perceptrons, convolutional neural networks, graph neural networks, and Gaussian processes, respectively, to extract information on the simulation parameters from these inputs while comparing systematically to analogous results from our previous generation of (25Mpc/h)^3 simulations. We generally find that the new, larger volumes produce tighter marginal constraints on the parameters, to degrees that vary between the different inputs. The improvements, however, scale more weakly than with the square root of the increase in the amount of data (i.e., physical volume). We interpret this as originating either from information loss due to mode coupling or from complex degeneracies in parameter space. We also discuss the effects on statistics of the intergalactic medium temperature from four new parameters that are varied in these simulations, which control the amplitude and timing of the ionizing background radiation. We publicly release the simulation outputs and ancillary data at https://camels.readthedocs.io.

astro-ph.CO↗

arXiv2026

Reconciling the Fundamental Plane of Early-Type Galaxies with hydrodynamical simulations: The case of IllustrisTNG100-1

The Fundamental Plane (FP) of Early-Type Galaxies (ETGs) encapsulates a tight correlation among their structural and dynamical properties and provides an important benchmark for galaxy formation models. However, cosmological hydrodynamical simulations have historically struggled to reproduce the observed FP tilt, with discrepancies often attributed to to flawed feedback physics or insufficient resolution. Using the IllustrisTNG100-1 simulation, we show that adopting observationally motivated measurements, including Sérsic-derived photometric parameters and dynamically inferred velocity dispersions designed to minimise softening-length effects, substantially reduces the discrepancy between simulated and observed FPs. We further explore the impact of non-universal, mass-dependent Initial Mass Function (IMF) variations through forward modelling of their effects on galaxy structural and dynamical quantities. In particular, bottom-heavy IMF variations produce FP coefficients fully consistent with observational constraints for both direct and orthogonal fits. Our results suggest that a significant fraction of the long-standing FP tension arises from how galaxy observables are extracted and interpreted in simulations, although residual discrepancies may still reflect limitations in the underlying baryonic physics. These findings highlight the importance of observational realism and IMF variations for interpreting galaxy scaling relations and for improving the predictive power of hydrodynamical simulations of ETG formation.

astro-ph.GA↗

arXiv2026

Cosmology with one galaxy: An analytic formula relating $Ω_{\rm m}$ with galaxy properties

Standard cosmological analyses typically treat galaxy formation and cosmological parameter inference as decoupled problems, relying on population-level statistics such as clustering, lensing, or halo abundances. However, classical studies of baryon fractions in massive galaxy clusters have long suggested that gravitationally bound systems may retain cosmological information through their baryonic content. Building on this insight, we present the first analytic and physically interpretable cosmological tracer that links the matter density parameter, $Ω_m$, directly to intrinsic galaxy-scale observables, demonstrating that cosmological information can be extracted from individual galaxies. Using symbolic regression applied to state-of-the-art hydrodynamical simulations from the CAMELS project, we identify a compact functional form that robustly recovers $Ω_m$ across multiple simulation suites (IllustrisTNG, ASTRID, SIMBA, and Swift-EAGLE), requiring only modest recalibration of a small number of coefficients. The resulting expression admits a transparent physical interpretation in terms of baryonic retention and enrichment efficiency regulated by gravitational potential depth, providing a clear explanation for why $Ω_m$ is locally encoded in galaxy properties. Our work establishes a direct, interpretable bridge between small-scale galaxy physics and large-scale cosmology, opening a complementary pathway to cosmological inference that bypasses traditional clustering-based statistics and enables new synergies between galaxy formation theory and precision cosmology.

astro-ph.CO↗

arXiv2026

The DREAMS Project: Disentangling the Impact of Halo-to-Halo Variance and Baryonic Feedback on Milky Way Dark Matter Density Profiles

In this work, we utilize a new suite of Milky Way-mass halos from the DREAMS Project, simulated with Cold Dark Matter (CDM), to quantify the influence of baryon feedback and intrinsic halo-to-halo variance on dark matter density profiles. Our suite of 1024 halos varies over supernova and black hole feedback parameters from the IllustrisTNG model, as well as variations in two cosmological parameters. We find that, for the DREAMS parameter variations, Milky Way-mass dark matter density profiles in the IllustrisTNG model are largely insensitive to astrophysics and cosmology variations, with the dominant source of scatter instead arising from halo-to-halo variance. However, most of the (comparatively minor) feedback-driven variations come from the changes to supernova prescriptions. By comparing to dark matter-only simulations, we find that the strongest supernova wind energies are so effective at preventing galaxy formation that the halos are nearly entirely collisionless dark matter. Finally, regardless of physics variation, all the DREAMS halos are roughly consistent with a halo contracting adiabatically from the presence of baryons, unlike models that have bursty stellar feedback. This work represents a step toward assessing the uncertainty in Milky Way dark matter profiles, with direct implications for dark matter searches where systematic uncertainty in the density profile remains a major challenge.

astro-ph.GA↗

arXiv2026

Field-Level Inference of Primordial Non-Gaussianity with the Quijote Simulation Suite

Local primordial non-Gaussianity, parameterised as $f_{\rm NL}^{\rm local}$, will be stringently constrained using state-of-the-art methods applied to next-generation galaxy redshift survey data. In this paper, in preparation for the upcoming data sets, we demonstrate for the first time the joint field-level inference of $f_{\rm NL}^{\rm local}$, nuisance parameters, and the initial conditions in realistic halo catalogues, ones which are generated through full dark-matter-only $N$-body simulations. The field-level inference algorithm optimally constrains $f_{\rm NL}^{\rm local}$ through a Bayesian forward-modelling approach at the field level, which outperforms traditional methods by leveraging the full statistical power of the data at the scales considered. In addition, we assess its performance under various design choices in the forward model, including tests of the structure formation model and resolution. We demonstrate the robustness of our approach by applying it to a subset of the \textit{Quijote} simulation suite, performing the inference at scales down to $k_{\rm max} \approx 0.1 h \rm{Mpc}^{-1}$. Compared with a power spectrum and bispectrum estimator, we find a $\sim1.3$ improvement in $σ(f_{\rm NL}^{\rm local})$ when applying \borg{}, while marginalising over the initial conditions and bias parameters. From the small-scale information sensitivity tests, we show that the constraints on $f_{\rm NL}^{\rm local}$ improve as we increase the resolution of the inference. These findings underscore the transformative potential of field-level inference to leverage the information available in ongoing surveys such as \textit{Euclid}, providing accurate insights into the physics of cosmic inflation and the number of fields driving it.

astro-ph.CO↗

arXiv2026

The Future of Artificial Intelligence and the Mathematical and Physical Sciences (AI+MPS)

This community paper developed out of the NSF Workshop on the Future of Artificial Intelligence (AI) and the Mathematical and Physics Sciences (MPS), which was held in March 2025 with the goal of understanding how the MPS domains (Astronomy, Chemistry, Materials Research, Mathematical Sciences, and Physics) can best capitalize on, and contribute to, the future of AI. We present here a summary and snapshot of the MPS community's perspective, as of Spring/Summer 2025, in a rapidly developing field. The link between AI and MPS is becoming increasingly inextricable; now is a crucial moment to strengthen the link between AI and Science by pursuing a strategy that proactively and thoughtfully leverages the potential of AI for scientific discovery and optimizes opportunities to impact the development of AI by applying concepts from fundamental science. To achieve this, we propose activities and strategic priorities that: (1) enable AI+MPS research in both directions; (2) build up an interdisciplinary community of AI+MPS researchers; and (3) foster education and workforce development in AI for MPS researchers and students. We conclude with a summary of suggested priorities for funding agencies, educational institutions, and individual researchers to help position the MPS community to be a leader in, and take full advantage of, the transformative potential of AI+MPS.

cs.AI↗

arXiv2026

Interpreting Cosmological Information from Neural Networks in the Hydrodynamic Universe

What happens when a black box (neural network) meets a black box (simulation of the Universe)? Recent work has shown that convolutional neural networks (CNNs) can infer cosmological parameters from the matter density field in the presence of complex baryonic processes. A key question that arises is, which parts of the cosmic web is the neural network obtaining information from? We shed light on the matter by identifying the Fourier scales, density scales, and morphological features of the cosmic web that CNNs pay most attention to. We find that CNNs extract cosmological information from both high and low density regions: overdense regions provide the most information per pixel, while underdense regions -- particularly deep voids and their surroundings -- contribute significantly due to their large spatial extent and coherent spatial features. Remarkably, we demonstrate that there is negligible degradation in cosmological constraining power after aggressive cutting in both maximum Fourier scale and density. Furthermore, we find similar results when considering both hydrodynamic and gravity-only simulations, implying that neural networks can marginalize over baryonic effects with minimal loss in cosmological constraining power. Our findings point to practical strategies for optimal and robust field-level cosmological inference in the presence of uncertainly modeled astrophysics.

astro-ph.CO↗

arXiv2026

One latent to fit them all: a unified representation of baryonic feedback on matter distribution

Accurate and parsimonious quantification of baryonic feedback on matter distribution is of crucial importance for understanding both cosmology and galaxy formation from observational data. This is, however, challenging given the large discrepancy among different models of galaxy formation simulations, and their distinct subgrid physics parameterizations. Using 5,072 simulations from 4 different models covering broad ranges in their parameter spaces, we find a unified 2D latent representation. Compared to the simulations and other phenomenological models, our representation is independent of both time and cosmology, much lower-dimensional, and disentangled in its impacts on the matter power spectra. The common latent space facilitates the comparison of parameter spaces of different models and is readily interpretable by correlation with each. The two latent dimensions provide a complementary representation of baryonic effects, linking black hole and supernova feedback to distinct and interpretable impacts on both the matter power spectrum, and field, level. Our approach enables developing robust and economical analytic models for optimal gain of physical information from data, and is generalizable to other fields with significant modeling uncertainty.

astro-ph.CO↗