arXiv subjects

João Alves

Publications and source records attributed to João Alves.

At least 19 recordsLinked to original sources

arXiv2026

The star formation history of Orion is structured and episodic

The propagation of star formation throughout a molecular cloud complex can tell us much about its physical drivers and the importance of environmental versus intrinsic processes. We reconstruct the star formation history (SFH) of the Orion star-forming complex using the highest-resolution age map currently available. It comprises 47 co-spatial and co-moving groups identified by the SigMA clustering algorithm, with homogeneously derived ages and group memberships. We find four periods of increased star formation, separated by ~5 Myr each. The two periods of highest increase, at 6.4 and 11.3 Myr, can be statistically localized, together hold ca. 59% of all stars, and are robust against resampling and changes to the age fitting. The two older, smaller increases at 16.2 and 21.4 Myr can be identified in the age distribution, but cannot be shown to be robust at this point. The ~5 Myr spacing between the episodes of increased star formation is the same as found for the nearby Scorpius-Centaurus (Sco-Cen) star-forming region. We find no methodological bias that could produce the SFH structure. We identify spatio-temporally connected structures, including three cluster-chain candidates. We conclude that star formation in Orion seems to occur on the same timescales but with a different pattern than in Sco-Cen. The cause of the ~5 Myr separations remains a mystery, and its detection in two distinct star-forming regions needs further investigation.

astro-ph.GA↗

arXiv2026

The fading hierarchy of Galactic open clusters

Galactic open clusters provide a record of both hierarchical star formation and the subsequent dynamical evolution of the Milky Way disk. We use the two-point correlation function to characterize the spatial and kinematic clustering of open clusters in the solar neighborhood, considering their three-dimensional (3D) distributions and their projections onto the Galactic plane, as well as subsamples divided by age, mass, and survey volume. We quantify the clustering through fractal dimensions and characteristic correlation scales. We find that the youngest clusters have fractal dimensions comparable to those measured in the interstellar medium, suggesting that they retain part of the hierarchical structure of their natal molecular clouds. The clustering strength then decreases systematically with age and becomes weak on a characteristic timescale of order 100 Myr. This evolution is scale dependent: the excess correlation at tens of parsecs fades more rapidly than the weaker correlation at hundreds of parsecs. By contrast, we find only a weak dependence of the clustering properties on cluster mass and volume. Using the same cluster sample, we derive an empirical mapping between projected and 3D fractal dimensions, providing an observational calibration for comparing the Milky Way with studies of external galaxies, where only projected measurements are available. The open-cluster population exhibits characteristic spatial and kinematic clustering scales of approximately 370 pc and 11 km/s, respectively. These results support a picture in which young open clusters inherit a spatial hierarchy from star formation and progressively lose their spatial correlations through subsequent evolution.

astro-ph.GA↗

arXiv2026

The dispersal law of young star clusters: free expansion -- A Gaia/eROSITA census of Orion and Sco--Cen

We measure how young star clusters disperse over time in the two nearest star-forming complexes. Our sample comprises 44 \texttt{SigMA}-\textit{Gaia}-defined clusters in Orion and 31 in Sco--Cen, with isochronal ages of 2--25\,Myr. We extend each cluster's census with \textit{eROSITA}-selected young stars that share its parallax and motion and lie within the distance they could have traveled ballistically since birth. By 10--30\,Myr, about one in three former cluster stars lies outside the cluster's cataloged boundary. Across both complexes and the full age range, we find that (1) the half-member radius increases linearly with age, $\rfifty \simeq \sigbirth t$, (2) the velocity dispersion shows no measurable evolution, and (3) all studied clusters are unbound. These results support free expansion. Young clusters in Orion and Sco-Cen emerge unbound from their birth clouds and coast at their birth velocity dispersion without significant relaxation or a characteristic dispersal timescale. We further show that memberships defined by density contrast preferentially omit the fastest stars, making the same clusters appear to expand more slowly ($\propto t^{0.6}$), cool with age, and approach equilibrium.

astro-ph.GA↗

arXiv2026

The superclumps of the local Milky Way. Supercloud fragmentation and the sites of star formation

Using a Gaia-based 3D dust map of the solar neighborhood, we analyze the internal structure of the seven local superclouds. We identify quasi-periodic density enhancements along their spines, which we term "superclumps" and show that 73% of the known star-forming regions in the dust map volume can be associated with them. Across the six superclouds with more than one recovered superclump, the spacings are characteristic per cloud: $\sim$150-250 pc for the Split, Malpolon Cloud and Vela Ridge Cloud, and $\sim$250-380 pc for the Radcliffe Wave, Natrix Cloud and Sagittarius Spur Extension. The observed separations are two to three times smaller than the $\sim$560 pc predicted for an isolated self-gravitating cylinder of the same effective diameter, indicating fragmentation under external pressure rather than in isolation. The recurring spacing suggests that giant molecular cloud assembly is not a local, stochastic process, but is instead influenced by the large-scale gravitational fragmentation of the parent superclouds. Regardless of the precise formation mechanism, the superclumps occupy a critical intermediate scale in the hierarchical organization of the interstellar medium, bridging the gap between the kiloparsec-scale gas lanes and the $\sim$10-100 pc scale of individual giant molecular clouds.

astro-ph.GA↗

arXiv2026

The Nearby Star Formation and Supernova Histories Reconstructed from Young Star Clusters

We reconstruct the recent star formation and core-collapse supernova (ccSN) histories of the Solar Neighborhood from the past trajectories of young star clusters. Using a \textit{Gaia}-based cluster sample with newly derived ages, masses, and bulk 3D velocities, we integrate orbits backward in an assumed axisymmetric Galactic potential and combine the trajectories with IMF sampling and stellar lifetimes to infer ccSN times and locations over the past 50 Myr. The result is an all-sky, 3D, time-resolved map of nearby ccSN activity for comparison with high-resolution 3D views of the local interstellar medium. The 0--15 Myr map shows strong enhancements toward Orion, Vela, Sco--Cen, and Cepheus, many within present-day cavities and shells. At earlier times, the dominant enhancements trace the Collinder 135, Messier 6, and Alpha Persei cluster families, showing how the remnants of massive star-forming complexes have shaped the recent local feedback history. We recover a bursty star formation history followed by a delayed, smoother ccSN history. Over the last 40 Myr, the mean star formation and ccSN rates are $823~M_\odot~\mathrm{Myr}^{-1}$ and $7.7~\mathrm{Myr}^{-1}$, respectively, corresponding to a Milky Way rate of $0.55\pm0.03~\mathrm{century}^{-1}$. Present-day OB-star catalogs yield rates ranging from agreement with the cluster reconstruction to several times higher. Because the catalogs overlap weakly and require different corrections, we do not rescale the ccSN map. Our reconstruction provides an empirical framework for connecting the recent history of massive-star feedback to the 3D structure and life cycle of gas in the nearby Milky Way.

astro-ph.GA↗

arXiv2026

A comprehensive cluster census of Orion. An application of the Significance Mode Analysis (SigMA) algorithm

Precise astrometric surveys and modern clustering algorithms are working in step to transform our view of star-forming regions. By revealing a much richer substructure than previously accessible, they pave the way for reconstructing star formation histories by accurately resolving and age-dating individual sub-populations. The Orion star-forming complex is the best-studied stellar nursery in the solar neighborhood and the nearest one currently forming massive stars. Even so, a comprehensive characterization of its substructure, including a homogeneous age mapping and extinction analysis, is still incomplete. Here, we present the most complete census of stellar populations across the Orion complex from the newest version of the SigMA algorithm and outline our additions and improvements to the algorithm that have extended its usage to distant (>300 pc) regions. We separate the Orion complex into 47 co-spatial and co-moving stellar groups comprising 11,996 reliable members, with ages ranging from 1.5 to 25 Myr. To evaluate the statistical robustness of each group, we derive cluster persistence values and individual membership probabilities for each source from 10 independent clustering repetitions. Our group memberships agree well with the literature, but SigMA consistently finds a factor of ~2-3 more members. In particular, it resolves more very young populations, such as NGC 2024, RV Orionis, B30, NGC 1977, NGC 2068, and NGC 2071, than previous algorithms. In addition to recovering 28 known clusters and three groups previously classified as substructures, we present 16 new co-eval substructure candidates of the Orion star-forming complex. This work builds up a new high-resolution time-resolved picture of Orion. This spatio-temporal map allows us to relate its stellar content to the surrounding ISM and paves the way for a detailed analysis of its star formation history in the future.

astro-ph.GA↗

arXiv2026

A&A community survey on the future of scientific publishing: Credibility over speed, fairness over profit, human judgment over automation

(Abridged) Scientific publishing is undergoing major change, driven by a shift toward open access (OA), the rise of artificial intelligence (AI), and growing demands for transparency, reproducibility, and equity. At the same time, rapid growth in article output strains editors and reviewers and means that metrics and speed can eclipse quality and rigor. To better understand how the community is responding, Astronomy \& Astrophysics (A\&A) commissioned the {A\&A Survey on Trends and Challenges in Scientific Publishing}, which documents community opinion on journal choice, peer review, OA, research evaluation, and the role of AI, with the goal of informing future editorial policies and the wider conversation on sustainable, ethical, and equitable scientific communication. Distributed online in May 2025 to \SI{28787} A\&A authors and co-authors, the survey drew \SI{2944} responses from 69 countries by its closing date. The responses were clear. Journal quality and reputation are the most decisive factors in deciding where to publish, followed by cost. The principal worry about peer review is reviewer expertise and fairness rather than speed. Citation counts are still an important consideration, but many respondents want broader, more qualitative measures of impact. The majority prefers public or institutional funding for OA, and views on AI are polarized, with widespread acceptance of administrative and language assistance but firm opposition to autonomous decision-making or content generation. Integrity, credibility, and fairness are common themes in every section of the responses. Overall, the survey portrays an engaged community that values quality over speed, fairness over profit, and human oversight over automation, providing A\&A with clear insight into community preference and a solid framework for shaping future policies on OA, peer review, and the responsible integration of AI.

cs.DL↗

arXiv2026

A new Gaia census of OB associations within 1 kpc

OB associations are primordial tracers of star formation and Galactic structure. Originally defined about 80 years ago, their historical membership lists have been superseded thanks to the precise astrometry from ESA's \textit{Gaia}'s satellite. Recent studies have however been mostly focused on individual OB associations or limited by the coverage of spectroscopic surveys. In this paper, we exploit a complete census of $\sim$25,000 O- and B-type stars within 1 kpc of the Sun to produce a highly-reliable catalogue of 56 OB associations using the HDBSCAN clustering algorithm, increasing the number of known OB associations by a factor of two within this volume. We assess the validity of this catalogue by crossmatching our OB association members with other catalogues of OB associations, star clusters and young stellar groups, confirming the high-confidence of our census of OB associations. We characterize these OB associations physically (total initial stellar mass, number of OB stars, ...) and kinematically (velocity dispersion, linear expansion ages, ...). The majority of the OB associations (38 out of 56) exhibit a significant expansion pattern in at least one direction, including 12 in both plane-of-the-sky directions, though differences in expansion velocity suggest anisotropical expansion patterns. We compare the locations of these OB associations with superclouds and features in the local Milky Way such as the Radcliffe Wave and discuss the implications for star formation in the solar neighbourhood.

astro-ph.GA↗

arXiv2026

Horse Eye Blink Detection and Classification for Equine Affective State Assessment

Automated detection of equine facial action units (AUs) is a promising yet under-explored avenue for pain and affective state assessment in horses. Half and full-blink movements are recognised indicators of pain and stress, but as micro-expressions, their subtle, fine-grained nature makes them easily missed by the naked eye and only discernible through frame-by-frame video inspection, making reliable automated detection from video a particularly demanding task. We develop and evaluate three methods for automated blink classification from horse videos: a frame-based YOLOv12 detector, an optical flow magnitude thresholding approach, and a fine-tuned VideoMAE model, tested on a publicly available dataset. We achieve a macro-F1 score of 0.898 when doing blink classification and 0.926 on binary blink detection. Our results highlight both the potential and the inherent challenges of fine-grained AU detection for equine welfare monitoring.

cs.CV↗

arXiv2026

EuroBERT: Scaling Multilingual Encoders for European Languages

General-purpose multilingual vector representations, used in retrieval, regression and classification, are traditionally obtained from bidirectional encoder models. Despite their wide applicability, encoders have been recently overshadowed by advances in generative decoder-only models. However, many innovations driving this progress are not inherently tied to decoders. In this paper, we revisit the development of multilingual encoders through the lens of these advances, and introduce EuroBERT, a family of multilingual encoders covering European and widely spoken global languages. Our models outperform existing alternatives across a diverse range of tasks, spanning multilingual capabilities, mathematics, and coding, and natively supporting sequences of up to 8,192 tokens. We also examine the design decisions behind EuroBERT, offering insights into our dataset composition and training pipeline. We publicly release the EuroBERT models, including intermediate training checkpoints, together with our training framework.

cs.CL↗

arXiv2026

Requiem for a belt: A spatial and kinematical reinterpretation of Gould's Belt in light of Gaia

We reassess the long-standing idea of Gould's Belt using Gaia DR3 for a sample of young massive stars and nearby young clusters. The structure surrounding the Sun, often interpreted as an inclined, expanding, and rotating ring, emerges in our analysis as a transient alignment of a few cluster families rather than an individual, coherent dynamical feature. By combining the ALS III catalog of OB stars with a homogeneous sample of clusters younger than 70 Myr, and by tracing their motions in a realistic Galactic potential, we show that neither the spatial distribution nor the kinematics form a unified system. The inferred expansion, rotation, and bulk motion of the Belt can be reproduced by the superposition of the $α$Per, Cr135, M6, and $γ$Vel cluster families and are further amplified by solar reflex motion and historical assumptions about the local standard of rest (LSR). The classic inclined geometry is largely explained by the oscillatory pattern of the Radcliffe Wave, which contributes a major arc of the supposed ring. Taken together, these results indicate that Gould's Belt is not a physical structure but a 3D asterism shaped by a complex local star formation history, observational biases, and projection effects.

astro-ph.GA↗

arXiv2026

The evolution of velocity dispersion in the Sco-Cen OB association

We study how the stellar velocity dispersion within the Scorpius-Centaurus OB association (Sco-Cen) has evolved over approximately 20 million years, from its formation to the present day, by investigating 32 stellar clusters in Sco-Cen. Using data from the Gaia mission along with supplementary stellar radial velocities, we identified a surprising sequence of abrupt jumps and intervening plateaus in the evolution of velocity dispersion correlating with times of star formation bursts. We find that the association is almost isotropically expanding and that star formation propagated from inside-out with a speed of about 5-6 km/s. We measure a present-day expansion rate of about 10-12 pc/Myr and observe that younger star clusters within the association exhibit higher velocities compared to older ones. This result, along with the stepwise increase in velocity dispersion over time, suggests a structured and sequential star formation process rather than a random one. This phased evolution suggests that stellar feedback is the primary driver of Sco-Cen's star formation history, expansion, and eventual dispersal. Our findings emphasise the value of precisely characterising stellar populations within OB associations, particularly through the creation of detailed, high-resolution age maps.

astro-ph.GA↗

arXiv2026

EuroLLM-22B: Technical Report

This report presents EuroLLM-22B, a large language model trained from scratch to support the needs of European citizens by covering all 24 official European Union languages and 11 additional languages. EuroLLM addresses the issue of European languages being underrepresented and underserved in existing open large language models. We provide a comprehensive overview of EuroLLM-22B's development, including tokenizer design, architectural specifications, data filtering, and training procedures. Across a broad set of multilingual benchmarks, EuroLLM-22B demonstrates strong performance in reasoning, instruction following, and translation, achieving results competitive with models of comparable size. To support future research, we release our base and instruction-tuned models, our multilingual web pretraining data and updated EuroBlocks instruction datasets, as well as our pre-training and evaluation codebases.

cs.CL↗

arXiv2025

The superclouds of the local Milky Way

Recent 3D dust maps of the local Milky Way are revolutionizing our understanding of the Sun's Galactic neighborhood, providing much needed insight into the large-scale organization of the interstellar medium. Focusing on the largest scales in $\textit{Gaia}$-based 3D dust maps, we find a pattern of seven highly elongated, mostly parallel structures in the local $\sim 5\,\mathrm{kpc}^2$, five of which were previously unknown. These structures show pitch angles of $33.5 \pm 4.0 ^\circ$ and masses ranging from $10^5$ to $10^6$ $\mathrm{M}_\odot$. We refer to these structures as superclouds. Nearly all known star-forming regions in the solar neighborhood lie within the superclouds, primarily along their central axes, supporting the idea that they act as gas reservoirs for the formation of giant molecular clouds. All but one of the seven superclouds show an underlying undulation, indicating that this is not a property unique to the Radcliffe Wave. We find that while the superclouds have linear masses that vary by about a factor of 4, their volume densities only vary by about 10$\%$. This suggests that superclouds self-regulate their physical sizes and internal structure to maintain pressure equilibrium with their environment. These findings establish a new framework for understanding how large-scale Galactic structures shape the conditions for star formation in the solar vicinity, and likely in galaxies like the Milky Way.

astro-ph.GA↗

arXiv2025

Precise determination of circumstellar disk lifetimes: Disk evolution in a single star-forming region

Determining how long circumstellar disks last is key to understanding the timescale of planet formation. Typically, this is done by measuring the fraction of young stars with infrared-excess, a sign of circumstellar material, in stellar clusters of different ages. However, comparing data from different star-forming regions at different distances introduces uncertainties and biases because of the different sample completeness and environment. This study addresses these challenges by analyzing 33 clusters, aged 3 to 21 million years (PARSEC isochrones), within the Scorpius-Centaurus OB association, sampling the stellar IMF from the hydrogen burning limit to about 8 M$_\odot$. By using $\mathit{Gaia}$, 2MASS, and WISE data, we identified stars with infrared-excess through color-color diagrams and spectral energy distributions, ensuring a consistent selection of disk-bearing sources. Our results indicate a disk lifetime of $5.8 \pm 0.3$ Myr, about a factor of two longer than most previous estimates, suggesting that planet formation may have more time than previously thought. We also find that an exponential decay model best describes disk dispersal. These findings emphasize the importance of studying disk evolution in a single star-forming region to reduce uncertainties and refine our understanding of planet formation timelines.

astro-ph.SR↗

arXiv2025

Kinematics of young stellar objects in NGC 2024 based on infrared proper motions

The most recently formed young stellar objects (YSOs) in active star forming regions are excellent tracers of their parent cloud motion. Their positions and dynamics provide insight into cluster formation and constrain kinematic decoupling timescales between stars and gas. However, because of their strong extinction and young age, embedded YSOs are mainly visible at infrared wavelengths and thus absent from astrometric surveys such as Gaia. We measured the proper motions of 6,769 sources toward the NGC 2024 cluster in the Flame Nebula using multi-epoch near-infrared observations from three ESO public surveys: VISIONS, VHS, and the VISTA/VIRCAM science verification program. Cross-validation of our results with Gaia using optically visible stars shows excellent agreement, with uncertainties on the same order of magnitude. For 362 YSO candidates identified from the literature, we derived proper motions on the order of <5 mas/yr with mean measurement uncertainties of ~0.22 mas/yr. This is the first homogeneous proper motion measurement of this quality for more than half of these stars. For Class I and flat-spectrum sources, our results provide a >13-fold increase in available proper motion measurements. We analyzed the positional and kinematic differences between YSO classes and confirmed a previously reported inside-out age segregation from younger to older stars, likely driven by an outward movement of older stars. No evidence of prolonged hierarchical assembly was found. Instead, the results support a rapid (<1 Myr) cluster collapse. This scenario also accounts for the observed slightly higher 1D velocity dispersion of Class I sources relative to Class flat objects. YSO radial velocities generally align with the gas velocities measured from 12CO(3-2), HNC(1-0), HCN(1-0), and show a weaker correlation with N_2H+(1-0). Some Class II and III objects appear to be already decoupling.

astro-ph.GA↗

arXiv2025

Toward the fabric of the Milky Way I. The density of disk streams from a local $250^3$ pc$^3$ volume

We studied 12 disk streams found in a 250$^3$ pc$^3$ volume in the solar neighborhood, which we define as coeval and comoving stellar structures with aspect ratios greater than 3:1. Using Gaia Data Release 3 data and the advanced clustering algorithms SigMA and Uncover, we identified and characterized these streams beyond the search volume, doubling, on average, their known populations. We estimate the number density of disk streams to be $\approx 820$ objects / kpc$^3$ (for $|Z| < 100$ pc), or surface densities of $\approx 160$ objects / kpc$^2$. These estimates surpass N-body estimates by one to two orders of magnitude and challenge the prevailing understanding of their destruction mechanisms. Our analysis reveals that these 12 disk streams are dynamically cold with 3D velocity dispersions between 2 and 5 km s$^{-1}$, exhibit narrow sequences in the Hertzsprung-Russell diagram, and are highly elongated with average aspect ratios of 7:1, extending up to several hundred parsecs. We find evidence suggesting that one of the disk streams, currently embedded in the Scorpius-Centaurus association, is experiencing disruption, likely due to the primordial gas mass of the association.

astro-ph.GA↗

arXiv2025

Tower+: Bridging Generality and Translation Specialization in Multilingual LLMs

Fine-tuning pretrained LLMs has been shown to be an effective strategy for reaching state-of-the-art performance on specific tasks like machine translation. However, this process of adaptation often implies sacrificing general-purpose capabilities, such as conversational reasoning and instruction-following, hampering the utility of the system in real-world applications that require a mixture of skills. In this paper, we introduce Tower+, a suite of models designed to deliver strong performance across both translation and multilingual general-purpose text capabilities. We achieve a Pareto frontier between translation specialization and multilingual general-purpose capabilities by introducing a novel training recipe that builds on Tower (Alves et al., 2024), comprising continued pretraining, supervised fine-tuning, preference optimization, and reinforcement learning with verifiable rewards. At each stage of training, we carefully generate and curate data to strengthen performance on translation as well as general-purpose tasks involving code generation, mathematics problem solving, and general instruction-following. We develop models at multiple scales: 2B, 9B, and 72B. Our smaller models often outperform larger general-purpose open-weight and proprietary LLMs (e.g., Llama 3.3 70B, GPT-4o). Our largest model delivers best-in-class translation performance for high-resource languages and top results in multilingual Arena Hard evaluations and in IF-MT, a benchmark we introduce for evaluating both translation and instruction-following. Our findings highlight that it is possible to rival frontier models in general capabilities, while optimizing for specific business domains, such as translation and localization.

cs.CL↗