| Alexandra Amon |
Breaking Astrophysical Barriers for Next-Generation Cosmic Shear |
Realizing the full cosmological potential of surveys like the Legacy Survey of Space and Time (LSST) requires breaking the astrophysical barriers that currently cost more than half of the constraining power. I will present a new approach to calibrate the two dominant sources of uncertainty—baryonic feedback and intrinsic alignments— informed by multiplexed spectroscopic, CMB, and X-ray data. This approach has revealed new insights into galaxy formation, and it demonstrates a path to recover the small-scale information that holds the key to the next era of precision weak lensing cosmology. |
| Daniel Anglés-Alcázar |
Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) to model the impact of baryonic physics on cosmological structure formation |
The CAMELS project aims to overcome major obstacles limiting our understanding of the fundamental properties of the Universe by (1) providing thousands of state-of-the-art hydrodynamic simulations of cosmological structure formation covering a broad range of sub-grid model implementations of key physical processes driving galaxy evolution and (2) developing novel machine learning algorithms to maximize the extraction of information from cosmological surveys while marginalizing over uncertainties in galaxy formation physics. In this talk, I will review new simulations expanding the CAMELS public data repository and discuss recent progress towards understanding the impact of feedback from massive stars and supermassive black holes on the cosmic matter distribution and efficiently emulating their effects over large cosmological volumes. |
| M. Celeste Artale |
Halo Occupation Variations in LAEs, SMGs, and Star-Forming Galaxies |
We study the halo occupation of Lyman-alpha emitters (LAEs), submillimeter galaxies (SMGs), and the general star-forming galaxy population at high redshifts (z=2-4), using the cosmological hydrodynamical simulation IllustrisTNG. We investigate potential occupancy variations associated with secondary halo properties such as formation time, concentration, and environment. By comparing their halo occupation functions and satellite fractions, we aim to identify the physical drivers that link distinct tracers of star formation to the underlying dark matter distribution. This work provides a framework to interpret differences in their large-scale clustering of these populations. |
| Adrian Bayer |
Full-Sky Mocks for Multi-Probe Science |
Multi-probe analyses are now central to cosmology and astrophysics, enabling us to understand the Universe and the galaxy–halo connection like never before, but they demand mocks that are jointly consistent across surveys and observables. I’ll present the HalfDome and Backlight simulation projects, providing thousands of mocks of the multi-probe sky for LSS-CMB cross-correlations and beyond. I’ll discuss key considerations in making such mocks: the importance of large scales, the robustness of small scales, and ultimately, how we can leverage mocks of different fidelity to speed up cosmological simulation-based inference. |
| Peter Behroozi |
Enabling Bayesian Neural Networks for Everyone with the Ray Tracing Sampler |
Deep neural networks are now pervasive across astronomy, even as full uncertainty quantification of their outputs has remained an elusive goal. We introduce a sampling algorithm, ray tracing, which is able to perform accurate sampling with stochastic gradients, and show that it runs well on a range of neural network architectures (including MLPs, convolutional neural networks, and transformers), including networks of over a billion parameters---all on a single consumer-grade GPU. This method hence enables widespread use of Bayesian neural networks, with the associated promise of stronger uncertainty quantification. |
| Gillian Beltz-Mohrmann |
Results from the DESI Emulator Mock Challenge |
It has been shown that cosmological parameter constraints can be improved by a factor of 2 to 4 when smaller-scale information is included in the analysis. Additionally, including clustering measurements beyond standard two-point statistics can allow for obtaining tighter constraints on galaxy-halo connection parameters, as well as breaking degeneracies between galaxy model parameters and cosmology. Historically, using these non-standard clustering measurements has been limited by the challenges associated with modeling them analytically over a wide range of scales, as well as the complex impact of observational systematics on these clustering statistics, which is difficult to model. However, computational advancements in recent years have made these ambitious clustering analyses more feasible. In particular, improvements in simulations and machine learning techniques have allowed us to run large suites of high-resolution N-body simulations at cosmological volumes, and subsequently build emulators to model the relationship between cosmological parameters and summary statistics. This new ability to harness the statistical power of nonlinear and higher order clustering measurements comes with increased responsibility to demonstrate robustness of results. The DESI Emulator Mock Challenge (EMC) was designed to investigate the ability of several different simulation-based pipelines, which all rely on various summary statistics and methods of emulation, are able to recover unbiased cosmological parameter constraints when tested on a mock galaxy catalog generated from a novel galaxy-halo connection model. Such a result would demonstrate that the tracer models used in the different emulator pipelines are flexible enough to marginalize over unknown galaxy formation physics. A secondary aim of the DESI EMC is to examine the amount of constraining power that can be robustly extracted from different types of galaxy clustering and lensing measurements across a range of scales. In this talk, I will present the results from the DESI Emulator Mock Challenge, as well as takeaways and lessons learned for future mock challenges. |
| Leah Bigwood |
Understanding the small-scale universe: bridging galaxy formation and cosmology |
Uncertain modelling of baryonic feedback, how energy from active galactic nuclei and supernovae redistributes gas within and beyond halos, poses a major obstacle to testing the cosmological model with large-scale structure (LSS) probes that extend into the non-linear regime. Indeed, it represents the dominant source of uncertainty limiting the cosmological precision of weak lensing (WL). I will share work confronting this challenge: demonstrating that astrophysical models calibrated by gas probes can recover this loss of power, and providing evidence that the impact of baryonic feedback on the matter distribution is stronger than assumed in cosmological hydrodynamical simulations. Whether such strong feedback is physically plausible, and how a consensus model can be achieved to fully exploit the cosmological precision of upcoming LSS surveys, remain open questions. I will discuss a path forward: to use new, publicly available multi-wavelength survey data to transform our understanding of feedback processes and, in tandem, to develop a data-driven baryonic feedback model to fully realize the potential of next-generation weak lensing cosmology. |
| Sownak Bose |
Modelling the galaxy-halo connection for low-mass dwarfs using DESI BGS |
We use the Photometric Objects Around Cosmic Webs (PAC) method, which combines deep photometric data with spectroscopic tracers to constrain the stellar-to-halo-mass relation (SHMR) of low-mass dwarfs. Using 349 cross-correlation measurements from DESI Y1 BGS and DECaLS, reaching stellar masses as low as M_*∼10^6.4 M⊙, we model the data with a SHMR-based subhalo abundance matching framework applied to high-resolution N-body simulations. We constrain the SHMR relation down to halo masses of M_halo∼10^8 M⊙ and find a clear upturn below ~10^10 M⊙, indicating rising star-formation efficiency in small haloes. This feature appears to be robust to extensions including mass-dependent scatter, reionisation suppression, assembly bias, and alternative cosmologies. We find that central red dwarfs dominate the low-mass population. Our results suggest that star formation was significantly more efficient in small haloes prior to reionisation, followed by quenching from the UV background. This has important implications for models of galaxy formation, as well as constraints on the nature of dark matter. |
| Sergio Contreras |
The effect of baryons on the positions and velocities of satellite galaxies in the MTNG simulation |
Mock galaxy catalogues are often constructed from dark-matter-only simulations based on the galaxy-halo connection. Although modern mocks can reproduce galaxy clustering to some extent, the absence of baryons affects the spatial and kinematic distributions of galaxies in ways that remain insufficiently quantified. We compare the positions and velocities of satellite galaxies in the MTNG hydrodynamic simulation with those in its dark-matter-only counterpart, assessing how baryonic effects influence galaxy clustering and contrasting them with the impact of galaxy selection, i.e. the dependence of clustering on sample definition. Using merger trees from both runs, we track satellite subhaloes until they become centrals, allowing us to match systems even when their z=0 positions differ. We then compute positional and velocity offsets as functions of halo mass and distance from the halo centre, and use these to construct a subhalo catalogue from the dark-matter-only simulation that reproduces the galaxy distribution in the hydrodynamic run. Satellites in the hydrodynamic simulation lie 3-4% closer to halo centres than in the dark-matter-only case, with an offset that is nearly constant with halo mass and increases toward smaller radii. Satellite velocities are also systematically higher in the dark-matter-only run. At scales of 0.1 Mpc/h, these spatial and kinematic differences produce 10-20% variations in clustering amplitude -- corresponding to 1-3$σ$ assuming DESI-like errors -- though the impact decreases at larger scales. These baryonic effects are relevant for cosmological and lensing analyses and should be accounted for when building high-fidelity mocks. However, they remain smaller than the differences introduced by galaxy selection, which thus represents the dominant source of uncertainty when constructing mocks based on observable quantities. |
| Shy Genel |
What we need versus what we can do: cosmological hydrodynamical simulations for cosmological inference |
Modern cosmological inference demands simulations that are simultaneously large, high-resolution, and physically flexible—three requirements that are fundamentally in tension. Large volumes are needed to capture rare structures and reduce sample variance; high resolution is required to resolve galaxies and baryonic processes that shape observables; and broad parameter spaces are essential to marginalize over uncertain physics and extract robust cosmological information. Yet increasing any one of these dimensions rapidly drives up computational cost, forcing tradeoffs between what we need for precision cosmology and what current technology allows. In this talk, I will discuss this challenge and how new approaches are emerging to bridge the gap. |
| Elizabeth Gonzalez |
Building a realistic Universe for the next generation of galaxy surveys |
One of the main challenges of these observational projects is related to the identification and characterization of the halos by observing the galaxies that are expected to reside in them. Hence, galaxy simulations emerge as a powerful tool, both in the planning and analysing the galaxy surveys. "Science Pipeline at PIC" (SciPIC) is a galaxy simulation pipeline developed under a collaborative effort with multiple contributors. It relies on a Halo Occupation Distribution, HOD, and Abundance Matching, AM, methods to assign the number of galaxies and r-band luminosities. SciPIC deliver galaxy catalogues that aim to reproduce the observed galaxy number density and clustering above a given flux detection limit over a wide redshift range. In this work, we introduce a calibration pipeline, developed to automate the calibration procedure in which the HOD and AM parameters are optimised. Moreover, we present a calibration method for the intrinsic alignment modelling. The calibration method allows to explore the strategies applied aiming to enhance the realism of the assigned galaxy properties and ensure that remain accurate and up to date according to the latest constraints from upcoming wide-field surveys. |
| Henry Gray |
Linking Galaxy Properties to Dark Matter Halo Assembly Bias in the FLAMINGO Simulation |
Understanding how galaxies trace the Universe's large-scale structure is key to connecting observations of galaxy populations with the physics of cosmic structure formation. In the standard halo model, halo mass is assumed to be the dominant property governing the spatial clustering of halos and their galaxy content. However, numerical simulations have shown that halos of the same mass can exhibit different clustering strengths depending on their formation history and internal properties, a phenomenon known as halo assembly bias. Crucially, if not modeled correctly, assembly bias may skew inferences about cosmological parameters. Here, we investigate how different observable galaxy properties relate to the assembly bias strength of their host halos. In FLAMINGO, a very large hydrodynamical cosmological simulation, we first compute a noisy per-halo estimate of assembly bias before modeling the multi-dimensional dependence of assembly bias on galaxy properties with advanced machine learning methods. Ultimately, this study sheds light on which galaxy properties are most closely tied to assembly bias. |
| Andrew Hearin |
Multi-Survey Synthetic Cosmological Data with Diffsky |
I will give an overview of Diffsky, a forward model of the spectral energy distribution (SED) of galaxies that co-evolve with the dark matter halos they inhabit. Diffsky is a probabilistic and differentiable model, with SED predictions based on a JAX implementation of stellar population synthesis. Diffsky can populate simulated merger trees with galaxy SEDs across redshift, and a new, generative formulation of Diffsky can create Monte Carlo realizations of synthetic galaxy lightcones, including cosmology-dependence. In this talk, I will discuss ongoing efforts using Diffsky in the NASA OpenUniverse project to create synthetic data for analyses of the Roman survey, including multi-survey cross-correlations with other experiments such as DESI, and the Rubin Observatory LSST. |
| Michaela Hirschmann |
Modelling broad-line and narrow-line emission of AGN in simulated galaxies |
First JWST results revealed a growing population of faint, low-mass AGN at early cosmic epochs. However, identifying AGN, especially type-2, and estimating properties such as accretion rates and Eddington ratios for both type-1 and type-2 AGN from high-redshift galaxy spectra remains challenging. To address these issues, we developed novel, comprehensive grids of thousands of AGN photo-ionization models, covering both broad-line and narrow-line emission, considering the most recent incident AGN ionizing spectra, dependent on BH mass and Eddington ratio. These model grids are integrated into simulated galaxies from our new high-resolution, high-redshift cosmological zoom simulations (including updated models for BH accretion and winds) as well as the SantaCruz SAM and IllustrisTNG, making geometrical assumptions on the viewing angle to include broad emission lines and continuum emission. We revisit optical and UV emission-line diagnostics for both type-1 and type-2 AGN, presenting new criteria for identifying type-2 AGN, and novel tracers for BH accretion rates, Eddington-ratios, BH accretion-to-star formation rate ratios and gas metallicities in the vicinity of the BH. Comparisons with type-1 and type-2 AGN from JWST (applying the same selection criteria in simulations) indicate that simulations tend to underestimate type-2 AGN fractions and type-1 AGN number densities, likely due to insufficient gas accretion onto BHs at high redshift in current models. Finally, we outline future applications to JWST data and prospects for refining BH models in simulations. |
| Matt Ho |
Learning the Universe: Building a Scalable, Verifiable Emulation Pipeline for Astronomical Survey Science |
Learning the Universe is developing a large-scale, ML-accelerated pipeline for simulation-based inference in cosmology and astrophysics. By combining high-resolution physical models with fast emulators, we generate realistic training sets at the scale required for field-level inference from galaxy survey data. This enables us to constrain models of galaxy formation and cosmology from observations with unprecedented scale and precision. In designing this pipeline, we have also developed validation methodologies to assess emulator accuracy, identify sources of systematic error, and support blinded survey analysis. I will present results from its application to the SDSS BOSS CMASS spectroscopic galaxy sample and discuss how this approach is scaling to upcoming cosmological surveys. |
| Benjamin Horowitz |
Robust Differentiable Models for Galaxies and Stellar Mass |
In this talk, I will discuss various approaches to construct galaxy-related observables in a flexible differentiable framework suitable for large scale structure inference. These approaches include explicit formation modelling via differentiable hydrodynamics, approximate methods via particle-mesh co-evolution, and differentiable halo occupancy distribution models. These techniques will not only enable explicit field level inference, but can be used to construct and constrain flexible subgrid physics models. |
| Miguel Icaza |
Calibrating Galaxy–Halo Connection Models for PNG Constraints with GP Emulators |
Primordial non-Gaussianities (PNG), parameterized by the amplitude 𝑓ₙₗ, provide a direct window into the physics of the early Universe and a means to discriminate among inflationary models. CMB constraints on 𝑓ₙₗ are limited by cosmic variance, making large-scale structure (LSS) surveys such as DESI a promising avenue for further progress. A central challenge is that the PNG signal through 𝑓ₙₗ is exactly degenerate with the galaxy bias parameter 𝑏φ, which quantifies how galaxy clustering responds to the presence of PNG. I present a Gaussian-Process (GP) emulator framework to efficiently calibrate semi-analytical models (SAMs) of the galaxy–halo connection across simulations with varying PNG, reducing the number of expensive model evaluations required. Trained on PNG-UNITsims—a suite of high-resolution N-body simulations with local PNG—populated with SAMs, the emulator uses iterative steps to select new training points and accelerate convergence. These emulators are designed to place tight priors on 𝑏φ thereby breaking the 𝑏φ - 𝑓ₙₗ degeneracy in LSS analyses. |
| Johannes U. Lange |
Cosmological Constraints from Highly Non-Linear Clustering and Galaxy–Galaxy Lensing with DESI |
The Dark Energy Spectroscopic Instrument (DESI) is a groundbreaking cosmology experiment measuring the spectra of tens of millions of extragalactic galaxies and quasars. Its unprecedented capabilities enable the creation of the most detailed three-dimensional maps of the Universe produced to date. In this talk, I will present constraints on cosmic structure growth from the analysis of galaxy clustering and galaxy–galaxy lensing with galaxies from DESI. Projected galaxy clustering measurements from DESI are supplemented with lensing measurements from the Dark Energy Survey (DES), the Kilo-Degree Survey (KiDS), and the Hyper Suprime-Cam (HSC) survey. I will discuss what our analysis, employing a sophisticated simulation-based method and a complex HOD model, implies for the galaxy-halo connection and the growth-of-structure tension. |
| Mingshau Liu |
Continuous representations of baryonic feedback |
Accurate modeling of baryonic physics remains a major challenge for precision cosmology due to our incomplete understanding of complex subgrid processes, like star formation and feedback from supernovae and active galactic nuclei below ~10 Mpc scales. This uncertainty leads to different hydrodynamical simulation suites to implement fundamentally different prescriptions for these unresolved physics. Current simulation-based inference approaches rely therefore on discrete sets of simulators, each encoding specific physical assumptions, making it difficult to robustly quantify theoretical uncertainties and learn about the underlying physics from observations. We introduce a machine learning framework that learns continuous representations of baryonic feedback across multiple simulation suites, to enable interpolation between different physical implementations while providing robust uncertainty quantification. Our approach addresses the key challenge of marginalizing over theoretical uncertainties represented by various simulators while simultaneously constraining the underlying baryonic physics from observations. We frame this as learning a shared continuous latent representation of the physics implemented across different simulators, allowing us to both marginalize over and constrain a continuous baryonic parameter space. Using the CAMELS simulation suite, we demonstrate our method on several baryonic fields including stellar mass, gas density, temperature, and pressure fields. This framework provides a path toward more robust cosmological inference by properly accounting for theoretical uncertainties in baryonic modeling while extracting maximum information about the underlying physical processes from current and future surveys. |
| Luisa Lucie-Smith |
Cosmological feedback from a halo assembly perspective |
I will discuss how baryonic feedback impacts halo assembly histories and thereby imprints on cosmological observables. I will show the sensitivity of the thermal Sunyaev-Zel'dovich effect (tSZ) power spectrum, X-ray number counts, weak lensing and kinetic Sunyaev-Zel'dovich (kSZ) stacked profiles to halo populations as a function of mass and redshift. I will then present results on the imprint of different feedback implementations in the FLAMINGO suite of cosmological simulations on the assembly histories of these halo populations, as a function of radial scale. The results provide a new perspective on the interpretation of recent kSZ/tSZ and weak lensing measurements. |
| Amanda Lue |
Emulators for Efficient Forward Modelling: Advancing Cosmological and Astrophysical Parameter Inference from Cosmological Hydrodynamical Simulations |
Cosmological inference aims to connect observational data to theoretical models, enabling precise estimation of cosmological and astrophysical parameters. However, the computational demands of hydrodynamical simulations pose significant challenges, especially when exploring large parameter spaces or emulating realistic observational effects. To address this gap, I present an accelerated forward model that leverages recent advances in score-based diffusion models. The model conditions on both the input dark-matter (DM) voxel field and the six CAMELS cosmology/feedback parameters, enabling parameter-aware synthesis of galaxy fields. It will be trained on 50 Mpc CAMELS simulations and then generalized to larger cosmological volumes, enabling cosmological inference with tighter constraints on astrophysical feedback processes and cosmological parameters. By combining the efficiency of N-body simulations with the fidelity of hydrodynamical simulations, this project aims to integrate hydrodynamic-level accuracy into cosmological inference pipelines, laying a foundation for exploring cosmological and astrophysical parameter spaces with unprecedented precision. |
| Francisco Maion |
Exploring the Strengths and Limitations of IllustrisTNG with Multi-Zoom Hydrodynamical Simulations |
We present a new analysis of the IllustrisTNG galaxy formation model based on a set of simultaneous multi-zoom hydrodynamical simulations of halos embedded within a common cosmological volume. This framework allows us to efficiently probe galaxy populations across a wide range of masses and environments while maintaining full cosmological context. We compare predictions for the stellar mass function and galaxy gas fractions to current observational measurements, using Gaussian-process emulation to explore model variations at low computational cost. Our results quantify where IllustrisTNG provides a robust description of galaxy populations and identify regimes in which tensions with the data reveal limitations of the model assumptions. |
| Kaustav Mitra |
Differentiable empirical modeling of the galaxy-halo-gas connection for mock-making and analyses |
Diffsky provides a physically motivated, differentiable empirical modeling framework to study halo growth, halo structure, star formation histories, and galaxy assembly. Together, these models form a powerful machinery for interpreting the wealth of data from current and upcoming galaxy surveys. I will present two ongoing projects that advance and apply this approach within the context of the Dark Energy Spectroscopic Instrument (DESI). The first focuses on developing a suite of mock galaxy catalogs that accurately reproduce the one-point and two-point statistics of the DESI Bright Galaxy Survey (BGS), Luminous Red Galaxy (LRG), and Emission Line Galaxy (ELG) samples. These mocks will be crucial for validating key science pipelines in the DESI collaboration, from group-finding algorithms to full-shape cosmological inference. The second project involves building a differentiable and highly flexible baryonification model to characterize the gas-halo connection and quantify the impact of galaxy formation physics on small-scale matter distributions. Once in place we will use this framework to jointly fit DESI BGS and LRG data with thermal-SZ cross-correlations to constrain both galaxy formation physics and cosmology. |
| Antonio Montero-Dorta |
The galaxy bias profile of cosmic voids |
Characterizing the mapping between galaxies, halos, and the matter density field is essential not only for extracting cosmological information from galaxy surveys, but also for uncovering the physical mechanisms that shape galaxy evolution. In this talk, I will concentrate on new results that connect the clustering of galaxies and halos with their internal and environmental properties. First, I will present the galaxy bias profile inside cosmic voids, which was measured for the first time from hydrodynamical simulations using innovative clustering techniques. Second, I will discuss several observational methods to probe the multiple dependencies of galaxy clustering, which provide observational constraints on secondary bias/assembly bias that can be directly compared with results from cosmological simulations. I will highlight the relevance of these connections for both galaxy evolution and cosmology. |
| Daisuke Nagai |
Baryon Pasting: Forward Modeling the Gas-Galaxy-Halo Connection for Cosmological Inference |
In the era of advanced multi-wavelength surveys, accurately inferring cosmological parameters and understanding galaxy astrophysics presents both unprecedented opportunities and significant challenges. This talk will focus on the pivotal role of the Baryon Pasting project in forward modeling the gas-galaxy-halo connection for cosmological inference. The Baryon Pasting project aims to profile the elusive missing baryons, manifesting as diffuse warm-hot gas in and around galaxies, groups, clusters, and large-scale structures. By systematically capturing these components, the project endeavors to enhance the precision and robustness of baryonic feedback models, thus addressing the S8 cosmological tension—one of the prominent enigmas at the intersection of galaxy formation and cosmology. Utilizing Baryon-Pasted (BP) Uchuu mock simulations, we model data from Sunyaev-Zel'dovich (SZ) and X-ray surveys and integrate them with optical galaxy surveys. Our innovative approach elucidates how detailed profiling of missing baryons can significantly advance our understanding of baryonic processes and their impact on galaxy formation models. The BP Uchuu simulations serve as powerful tools for comprehensive modeling of multi-wavelength data, presenting a coherent and intricate picture of the gas-galaxy-halo connection. I will explore potential pathways to an integrated understanding of baryon cycles that span the Circumgalactic Medium (CGM), Intragroup Medium (IGrM), Intra-cluster Medium (ICM), and Intergalactic Medium (IGM). Particular attention will be given to the critical roles of supernova and Active Galactic Nucleus (AGN) feedback in driving these baryon cycles and their broader implications for galaxy formation and evolution. Finally, I will present recent advances in forward modeling for X-ray, SZ, and Fast Radio Burst (FRB) surveys, and discuss the promises and challenges of cross-correlating these with CMB, X-ray, FRB, and galaxy surveys. I will highlight key challenges, propose potential solutions, and outline future directions for improving cosmological models through the insights gained from the Baryon Pasting project. This discussion aims to bridge the gap between observations and theoretical models, enhancing our understanding of the processes that shape our Universe. |
| Boon Kiat Oh |
The Role of Galaxy Formation Models and Simulation Codes in Shaping Simulated Universes |
Simulating galaxies and the large-scale structure of the Universe relies on a diverse set of computational codes and galaxy formation models, each incorporating distinct physical assumptions and numerical methods. Despite their differences, many of these models successfully reproduce a common set of observational constraints, raising important questions about the origin and implications of their variations. Collaborative efforts such as CAMELS and AGORA provide valuable frameworks for systematically comparing these simulations and understanding how modelling choices influence galaxy properties. In this talk, I will explore the similarities and differences among widely used simulation codes and galaxy formation models, and discuss their impact on the resulting simulated universes and the mock galaxy catalogues derived from them. |
| Nelson Padilla |
Galactic conformity: a simple two-halo model and its link to assembly bias |
Galactic conformity has been the subject of intense study because it probes correlations between the baryon physics of galaxies over very large separations. However, simulations show that removing galaxies near massive halos strongly suppresses the signal. I will show that the same operation also reduces the amplitude of assembly bias, defined as the difference in large-scale bias at fixed halo mass between quenched and star-forming galaxies, and that changes in conformity closely track changes in this assembly-bias amplitude. This points to a tight connection between the two phenomena. I will then present a simple model for the two-halo component of galactic conformity, written entirely in terms of standard galaxy auto- and cross-correlation functions, which accurately reproduces the measured signal down to small separations. In this framework, two-halo conformity is not an independent statistic, but a specific combination of familiar correlation functions weighted by assembly bias. |
| Enrique Paillas |
Simulation-based emulators for DESI galaxy clustering |
I will present a pipeline to emulate galaxy clustering statistics at the two-point level and beyond, down to the non-linear regime, including many alternative summary statistics for which no complete analytic models exist in the literature, including the wavelet scattering transform, density-split clustering, Minkowski functionals, void statistics, and more. Our theory models are based on neural networks trained on high-fidelity N-body simulations that meet the requirements to reproduce the properties of the luminous red galaxy sample from the Dark Energy Spectroscopic Instrument. We test the performance of our pipeline at recovering cosmological parameters within the context of the LCDM model and its extensions, and validate its applicability beyond the halo occupation distribution framework that is used to model the connection between dark matter halos and galaxies. We combine the bits of each summary statistic that maximize the Fisher information into a single data vector through a greedy algorithm, which achieves the tightest cosmological constraining power in all cases that are studied. |
| Daniela Palma |
Tracing Large-scale conformity: Does the cosmic environment shape the signal? |
We investigate the origin of the large-scale conformity signal by examining its connection to the cosmic web. In particular, we focus on low-mass galaxies—the main contributors to the signal—and explore how those residing in different environments, such as cluster and group outskirts, filaments, and voids, either strengthen or weaken the signal. Our results demonstrate that the conformity amplitude depends on the density of the environment surrounding primary galaxies. This highlights the role of environmental processes in correlating the star formation activity of surrounding galaxies. |
| Shivam Pandey |
Painting galaxies onto dark matter only simulations using a transformer-based model |
Connecting the formation and evolution of galaxies to the large-scale structure is crucial for interpreting cosmological observations. While hydrodynamical simulations accurately model the correlated properties of galaxies, they are computationally prohibitive to run over volumes that match modern surveys. We address this by developing a framework to rapidly generate mock galaxy catalogs conditioned on inexpensive dark-matter-only simulations. We present a multi-modal, transformer-based model that takes 3D dark matter density and velocity fields as input, and outputs a corresponding point cloud of galaxies with their physical properties. We demonstrate that our trained model faithfully reproduces a variety of galaxy summary statistics and correctly captures their variation with changes in the underlying cosmological and astrophysical parameters, making it the first accelerated forward model to capture all the relevant galaxy properties, their full spatial distribution, and their conditional dependencies in hydrosimulations. |
| Dante Paz |
Voids as Cosmological and Astrophysical Laboratories |
In this talk, I will show how cosmic voids serve as privileged environments for investigating key phenomena in cosmology and astrophysics, including topics as diverse as primordial magnetic fields, primordial non-Gaussianities, modified gravity models and galaxy formation. I will present the main results of the research lines I am involved highlighting the tools, methodologies, and data we have developed. These resources may also be of interest for collaborative use. I will also describe how the unique dynamics of voids contribute to advancing our understanding of the structure and evolution of the Universe. |
| Lucia Perez |
The Impact of Galaxy Formation on Galaxy Biasing, and Implications for Primordial non-Gaussianity Constraints |
The parameter fNL measures the local non-Gaussianity in the primordial energy fluctuations of the Universe, with any deviation from fNL = 0 providing key constraints on inflationary models. Galaxy clustering is sensitive to fNL at large scale modes and the next generation of galaxy surveys will approach a statistical error of σfNL ∼ 1. However, the systematic errors on these constraints are dominated by the degeneracy of fNL with the galaxy bias parameters b1 (galaxy overdensities caused by mass perturbations) and bϕ (galaxy overdensities caused by primordial potential perturbations). It has been shown that the assumed scaling of bϕ(z) = 2δc(b1(z) − 1) is not accurate for realistically simulated galaxies, and depends both on the galaxy selection and the way that galaxies are modeled. To address this, we leverage the CAMELS-SAM pipeline to explore how varying parameters of galaxy formation affects bϕ and b1 for various galaxy selections. We run separate-universe N-body simulations of L = 205h−1 cMpc and N = 12803 to measure bϕ, and run 55 unique instances of the Santa Cruz semi-analytic model with varying parameters of stellar and AGN feedback. We find the behavior and evolution of a SC-SAM model’s stellar-, SFR- and sSFR- to halo mass relationships track well with how b1 and bϕ(b1) change across redshift and selection for the SC-SAM. We find our variations of the SC-SAM encapsulate the bϕ behavior previously measured in IllustrisTNG, the Munich SAM, and Galacticus. Finally, we identify sSFR selections as particularly robust to varied galaxy modeling. |
| Chad Popik |
Constraining Galaxy Models using SZ Cross-Correlation Measurements |
Statistical studies of the circumgalactic medium using Sunyaev-Zeldovich maps and galaxy redshift surveys offer a promising method of studying the gas properties of galaxies. Forward modeling thermodynamic profiles allows them to be fit off data, but current results show significant differences between observed signals and those predicted from simulations. We present results from our inference pipeline on stacking/cross-correlation measurements of DESI LRGs and ACT DR6 Compton y maps, and review various implementation effects and systematics that potentially bias forwards models. We also discuss using emulators built off the CAMELS simulations to directly constrain astrophysical parameters, such as those that quantify the strength of feedback. |
| Nesar Ramachandra |
From Exascale Simulations to Multi-Modal Foundation models: Synthetic Galaxies for Cosmological Inference |
Modern cosmological surveys demand simulated universes that capture the diversity of galaxies and their observables across multiple wavelengths. This talk presents complementary frameworks for creating realistic synthetic galaxies from gravity-only and hydrodynamical simulations, leveraging the distinct computational technologies behind each. Starting from dark-matter halo or galaxy catalogs, we generate spectral energy distributions using empirical and machine-learned models, along with photometry calibrated to survey data. These synthetic datasets serve as controlled testbeds for evaluating biases in cosmological analyses and for training generalizable machine-learning models. I will then discuss how these data products integrate with transformer-based multi-modal foundation models that bridge simulations and observations through shared latent representations spanning redshift, halo and stellar mass, star-formation histories, spectra, and broadband magnitudes. This enables cross-modality prediction and reconstruction—such as estimating redshifts from incomplete photometry or recovering spectral segments—demonstrating strong generalization across survey-like domains. Together, these developments illustrate an emerging paradigm for cosmology: simulation-informed, data-driven inference. By combining physically grounded synthetic catalogs with large-scale multi-modal learning, we can build interpretable, flexible, and survey-ready representations of the Universe, paving the way for integrated analyses across next-generation telescopes and simulation frameworks. |
| Andrew Robertson |
Accelerated calibration of semi-analytic galaxy formation models |
I will present an accelerated calibration framework for semi-analytic galaxy formation models, demonstrated with Galacticus. Rather than fitting directly to properties such as the low-redshift stellar mass function (SMF) - which requires evolving thousands of halos per likelihood evaluation - I constructed a fast likelihood from the stellar-to-halo mass relation (SHMR; mean and scatter) evaluated at a small set of target halo masses, reducing each evaluation to simulating only tens of galaxies. I will discuss the pros and cons of such an approach, as well as attempts to extend this approach to handle calibrating to properties other than stellar masses (stellar sizes, black hole masses, etc.). |
| Facundo Rodriguez |
Alignment of central galaxies across different scales |
Observations and simulations show that central galaxies align with both their satellite systems and the large-scale cosmic web. In this talk, we will dissect the physical drivers of these alignments using the TNG300 hydrodynamical simulation and SDSS data. Our analysis reveals a strong dependence on halo mass, but demonstrates that the apparent influence of the large-scale environment is primarily a secondary effect of local density. These results fit naturally into the multi-scale alignment framework, where the central galaxy aligns with its subhalo, the subhalo with the main dark matter halo, and the halo with the surrounding cosmic web. |
| Andres N. Salcedo |
Dark Energy Survey cluster cosmology constraints with simulation-based forward modeling |
Galaxy clusters are embedded in the most massive bound structures in the Universe making their properties sensitive to cosmology. In the standard approach the abundances of clusters as a function of their mass are used to constrain cosmology. Because cluster masses cannot be measured directly this approach relies on observable proxies of cluster mass that are calibrated using weak gravitational lensing. Robustly marginalizing over these mass-observable relations is therefore critical to accurately measuring cosmology with clusters. Recently, data from clusters identified in the Dark Energy Survey (DES) were found to be in significant tension with a variety of other probes, including a 5.6-sigma tension with cosmic microwave background data. We have developed a novel framework to forward model cluster selection in cosmological simulations to address this discrepancy. When applied to the same DES cluster data we have found that this framework can consistently describe the DES cluster data assuming a Planck cosmology, thereby resolving the tension observed in earlier analyses of the DES cluster data. We present cosmological structure constraints from a reanalysis of DES-Y1 cluster data using our new framework for clusters down to richness 10 as well as constraints on evolving dark energy from combinations with other cosmogical probes. |
| Jared Siegel |
Toward a consensus picture of baryon feedback: joint constraints from X-ray gas fractions, kSZ effect profiles, and galaxy-galaxy lensing |
There is no consensus on how baryon feedback shapes the non-linear matter power spectrum from either simulations or observations. With improvements in survey size and methodology, this uncertainty is now a limiting systematic for cosmic shear inference at small scales. In this talk, we address the uncertain landscape with a multi-observation view of feedback: kSZ effect profiles, X-ray gas mass fractions, and galaxy-galaxy lensing. Across a wide range of halo masses and redshifts, we find evidence of more efficient gas expulsion than predicted by most state-of-the-art simulations. We incorporate these constraints into data-driven priors for cosmic shear, improving cosmological constraining power on small scales. |
| Simona Sotiri |
Testing Cosmological Parameter Inference on Small, Non-Linear Scales with the FLAMINGO Simulations |
Next-generation surveys, such as those conducted with the Rubin Observatory and Roman Space Telescope, will measure galaxy clustering and gravitational lensing with extremely high precision. Studying the distribution of galaxies and matter on non-linear scales promises the most stringent constraints on cosmology but recovering unbiased constraints requires accurate modeling of the galaxy–halo connection. Here, we present a mock analysis of simulated galaxy clustering and lensing data from the FLAMINGO simulation, one of the largest and most realistic hydrodynamical simulations of the Universe carried out to date. Crucially, this simulation incorporates important features of galaxy physics such as baryonic feedback and galaxy assembly bias. We apply a complex Bayesian inference framework based on simulation-based modeling to check the reliability of cosmological constraints from non-linear scales. |
| Kate Storey-Fisher |
100,000 Universes: Simulation-Based Inference for Galaxy Clustering with muchísimocks |
The 3D distribution of galaxies encodes information about the composition and evolution of the universe. Simulation-based inference (SBI) offers an approach to extracting this high-dimensional information using a forward model of the galaxy field, allowing for access to higher order statistics and smaller scales. I will present our SBI approach using a novel library of tracer density fields called muchísimocks, which span 10,000 cosmologies using a full-field N-body emulator. We apply a flexible galaxy bias prescription, the hybrid Lagrangian bias expansion, allows us to robustly marginalize over our uncertainties about galaxy formation; our model is accurate down to k~0.4. I will discuss our results on inferring key cosmological parameters in mock data and implications for the analysis of cosmological galaxy surveys. |
| Chun-Hao To |
Simulations and lensing survey science |
Simulations play an important role in lensing survey analysis such as DES, Rubin, and Roman. I will review requirements for simulations that will be essential for survey science and give several examples on how they are used in DES, Rubin, and Roman. I will talk about how to use advanced galaxy/gas--halo connection model to create galaxy catalogs to enable survey sciences. |
| Francisco Villaescusa-Navarro |
New methods to tackle the Halo-galaxy connection |
In this talk, I will discuss new ways to use AI to tackle difficult problems like the halo-galaxy connection. I will illustrate how AI agents can propose, implement, and improve hypotheses and present the benefits this technology can have to address complex problems. |
| Ian Williams |
The Galaxy-Halo Connection from DESI BGS Galaxy Groups |
Galaxy group catalogs can function as models of the galaxy-halo connection. We present a new galaxy group catalog built from Data Release 1 of the Dark Energy Spectroscopic Instrument (DESI) Bright Galaxy Survey (BGS), using the Self Calibrated Group Finder. After introducing a new technique to mitigate systematics from fiber incompleteness, we calibrate group finder parameters with new measurements of luminosity and color dependent clustering in BGS. We present Halo Occupancy Distributions from the catalog of 2M galaxies, along with satellite fractions and luminosity/stellar halo mass relations. |
| Zheng Zheng |
Modeling Galaxy Clustering with the Conditional Color-Magnitude Distribution (CCMD) |
From simultaneously modeling the projected two-point correlation functions of SDSS galaxies in fine bins of color and luminosity, we obtain the joint distribution of galaxy color and luminosity as a function of halo mass, dubbed as the Conditional Color-Magnitude Distribution (CCMD). I will show the modeling results and compare the CCMD inferred from data with that predicted by galaxy formation models. I then present further tests of the CCMD results with various galaxy clustering measurements, including galaxy groups, galaxy lensing, and counts-in-cell statistics. Based on the results, I will discuss the implications for galaxy assembly bias and the role of the CCMD model in cosmological inference. |