In Fontanot et al. 2026, we present predictions from a realization of our GAEA model, implementing a prescription for a variable stellar initial mass function (IMF). This is inspired by high-resolution numerical simulations that account for the role of cosmic rays (CR) as regulators of the star formation rate (SFR) in giant molecular clouds. In our approach, SFR density is assumed to be a proxy for the CR density, providing a link between the IMF shape and the predicted physical conditions of the star forming interstellar medium. Our results show that, in our model framework, assuming such a variable IMF reproduces several properties of the z>6 galaxy population, with no further modification of the feedback model, including their UV luminosity functions up to z~13. In order to compare model predictions with available estimates for the galaxy stellar mass function (GSMF), we reconstruct stellar masses from the model's synthetic photometry assuming a universal IMF, reflecting standard observational practice. Under this approach, we show that the model can reproduce the evolution of the GSMF up to the highest redshifts accessible. Our findings highlight the need to consider a variable IMF shape in the error budget associated with stellar mass estimates. We show that the evolution of both the slope and normalization of the gas-phase mass metallicity relation can be used as powerful discriminant between models of early galaxy formation assuming different IMF evolution. 

In Osman et al. 2026, we present a novel self-consistent implementation of dust formation by stars, destruction by supernova shocks and hot gas, and growth within the dense interstellar medium (ISM) in our GAEA model. Our new model DUSTY-GAEA is able to reproduce the dust build-up as a function of stellar mass out to z ∼ 6, the scaling relations between the dust-to-gas and dust-to-metal ratios and the stellar mass and metallicity in the local Universe, and the dust mass function both in the local Universe and out to z ∼ 1. In the framework of our model, dust growth dominates the cosmic dust budget out to z ∼ 8, and we find that observational constraints beyond the local Universe can be reproduced only when assuming such efficient dust growth in the dense ISM. Nevertheless, reproducing the estimated number densities of dust-rich galaxies at higher redshifts remains challenging, as also found in independent theoretical work. We discuss our model predictions in comparison with both observational data and independent theoretical efforts, and we highlight how further observational constraints at high redshifts would help constrain dust models.

In Kamran et al. 2026, we characterize the HI--halo mass relation, its redshift evolution, and its intrinsic scatter, identifying its secondary dependences. We use the latest version of our GAEA model, applied to the Millennium-I and Millennium-II simulations, to predict the HI mass function (HIMF) and the HI--halo mass relation from the present day to redshift z~5. At z=0, the model reproduces the observed HIMF and its decomposition by host-halo mass. The median HI--halo mass relation rises with halo mass, peaks near 10^{11.7} Msun, declines as central galaxies are quenched by feedback from active galactic nuclei, and rises again where satellites dominate, approaching a single power law at high redshift. We show that the substantial scatter, of about 0.5 dex, is not random but is governed by halo assembly: at fixed mass, higher-spin, later-forming, and less-concentrated halos are systematically HI-richer, with spin together with either concentration or formation time accounting for part of this scatter and leaving an intrinsic dispersion of about 0.3 dex. We encode the median relation, these secondary trends, and the intrinsic scatter in a compact, physically motivated prescription expressed entirely in terms of quantities available in dark-matter halo catalogs. This prescription reproduces the full scatter and enables the construction of large-volume 21-cm mock catalogs for interpreting ongoing intensity-mapping measurements with SKA precursor facilities, such as MeerKAT, and for preparing for forthcoming surveys with the SKA.

In Zakharova et al. 2026, we investigate how galaxy properties, such as stellar mass and star formation rate, correlate with their position within the cosmic web. We use our GAEA model and the IllustrisTNG simulation to reconstruct the environmental histories of galaxies that today reside in filaments between z = 0 and z = 4. We find that filament galaxies at z = 0 are a heterogeneous mix of populations with distinct environmental histories, and a clear dependency on the infall times into filaments. The vast majority of filament galaxies at z = 0 have experienced group processing at some stage of their evolution, with only ∼20% of galaxies remaining centrals throughout their life. For low-mass filament galaxies (9 < log10(Mstar/Msun)< 10), both GAEA and TNG100 confirm that environmental effects are primarily driven by group processing: satellite galaxies in this mass range stop growing stellar mass and exhibit elevated quenched fractions, whereas their central counterparts in filaments have properties that are similar to those of field galaxies. In contrast, massive galaxies (log10(Mstar/Msun)> 10) are affected by the filament environment, regardless of being centrals or satellites. Massive galaxies that have never been satellites and that entered filaments more than 9 Gyr ago show accelerated stellar mass assembly and higher quenched fractions relative to the field, due to a higher frequency of merger events inside filaments. Moreover, the most massive log((Mstar/Msun)> 11) galaxies typically accreted onto filaments over 9 Gyr ago and have never become satellites within a larger halo, highlighting the role of filaments in building up the high-mass end of the galaxy population.

In De Lucia et al. 2025, we investigate the environments of massive quiescent galaxies at 3 < z < 5 using our GAEA model. We show that model high-z quiescent galaxies are alpha-enhanced and exhibit a wide range of stellar metallicities, in broad agreement with current observational estimates. Massive high-z quiescent galaxies in our model occupy a range of environments, from void-like regions to dense knots at the intersections of filaments. Quiescent galaxies in underdense regions typically reside in halos that collapsed early and grew rapidly at high redshift, though this trend becomes difficult to identify observationally due to a large intrinsic scatter in star formation histories. The descendants of high-z massive quiescent galaxies display a broad distribution in mass and environments at z=0, reflecting the stochastic nature of mergers. About one third of these systems remain permanently quenched in our model, while most rejuvenation events are merger-driven and more common in overdense regions.