Search

Long-duration gamma-ray bursts are thought to be associated with the core-collapse of massive, rapidly spinning stars and the formation of black holes. However, efficient angular momentum transport in stellar interiors, currently supported by asteroseismic and gravitational-wave constraints, leads to predominantly slowly-spinning stellar cores. Here, we report on binary stellar evolution and population synthesis calculations, showing that tidal interactions in close binaries not only can explain the observed sub-population of spinning, merging binary black holes but also lead to long gamma-ray bursts at the time of black-hole formation. Given our model calibration against the distribution of isotropic-equivalent energies of luminous long gamma-ray bursts, we find that ~10% of the GWTC-2 reported binary black holes had a luminous long gamma-ray burst associated with their formation, with GW190517 and GW190719 having a probability of ~85% and ~60%, respectively, being among them. Moreover, given an assumption about their average beaming fraction, our model predicts the rate density of long gamma-ray bursts, as a function of redshift, originating from this channel. For a constant beaming fraction $f_\mathrm{B}\sim 0.05$ our model predicts a rate density comparable to the observed one, throughout the redshift range, while, at redshift $z \in [0,2.5]$, a tentative comparison with the metallicity distribution of observed LGRB host galaxies implies that between 20% to 85% of the observed long gamma-ray bursts may originate from progenitors of merging binary black holes. The proposed link between a potentially significant fraction of observed, luminous long gamma-ray bursts and the progenitors of spinning binary black-hole mergers allows us to probe the latter well outside the horizon of current-generation gravitational wave observatories, and out to cosmological distances., Comment: Accepted for publication in A&A Letters, 12 pages, 6 figures, 1 table

In recent years, the extraction of opinions and information from user-generated text has attracted a lot of interest, largely due to the unprecedented volume of content in Social Media. However, social researchers face some issues in adopting cutting-edge tools for these tasks, as they are usually behind commercial APIs, unavailable for other languages than English, or very complex to use for non-experts. To address these issues, we present pysentimiento, a comprehensive multilingual Python toolkit designed for opinion mining and other Social NLP tasks. This open-source library brings state-of-the-art models for Spanish, English, Italian, and Portuguese in an easy-to-use Python library, allowing researchers to leverage these techniques. We present a comprehensive assessment of performance for several pre-trained language models across a variety of tasks, languages, and datasets, including an evaluation of fairness in the results.

Deep neural networks are vulnerable to small input perturbations known as adversarial attacks. Inspired by the fact that these adversaries are constructed by iteratively minimizing the confidence of a network for the true class label, we propose the anti-adversary layer, aimed at countering this effect. In particular, our layer generates an input perturbation in the opposite direction of the adversarial one and feeds the classifier a perturbed version of the input. Our approach is training-free and theoretically supported. We verify the effectiveness of our approach by combining our layer with both nominally and robustly trained models and conduct large-scale experiments from black-box to adaptive attacks on CIFAR10, CIFAR100, and ImageNet. Our layer significantly enhances model robustness while coming at no cost on clean accuracy., Comment: Accepted to AAAI Conference on Artificial Intelligence (AAAI'22)

The main two families of real hypersurfaces in complex space forms are Hopf and ruled. However, very little is known about real hypersurfaces in the indefinite complex projective space $\cpn$. In a previous work, Kimura and the second author introduced Hopf real hypersurfaces in $\cpn$. In this paper, ruled real hypersurfaces in the indefinite complex projective space are introduced, as those whose maximal holomorphic distribution is integrable, and such that the leaves are totally geodesic holomorphic hyperplanes. A detailed description of the shape operator is computed, obtaining two main different families. A method of construction is exhibited, by gluing in a suitable way totally geodesic holomorphic hyperplanes along a non-null curve. Next, the classification of all minimal ruled real hypersurfaces is obtained, in terms of three main families of curves, namely geodesics, totally real circles and a third case which is not a Frenet curve, but can be explicitly computed. Four examples of minimal ruled real hypersurfaces are described.

We study the impact of mass-transfer physics on the observable properties of binary black hole populations formed through isolated binary evolution. We investigate the impact of mass-accretion efficiency onto compact objects and common-envelope efficiency on the observed distributions of $\chi_{eff}$, $M_{chirp}$ and $q$. We find that low common envelope efficiency translates to tighter orbits post common envelope and therefore more tidally spun up second-born black holes. However, these systems have short merger timescales and are only marginally detectable by current gravitational-waves detectors as they form and merge at high redshifts ($z\sim 2$), outside current detector horizons. Assuming Eddington-limited accretion efficiency and that the first-born black hole is formed with a negligible spin, we find that all non-zero $\chi_{eff}$ systems in the detectable population can come only from the common envelope channel as the stable mass-transfer channel cannot shrink the orbits enough for efficient tidal spin-up to take place. We find the local rate density ($z\simeq 0.01$) for the common envelope channel is in the range $\sim 17-113~Gpc^{-3}yr^{-1}$ considering a range of $\alpha_{CE} \in [0.2,5.0]$ while for the stable mass transfer channel the rate density is $\sim 25~Gpc^{-3}yr^{-1}$. The latter drops by two orders of magnitude if the mass accretion onto the black hole is not Eddington limited because conservative mass transfer does not shrink the orbit as efficiently as non-conservative mass transfer does. Finally, using GWTC-2 events, we constrain the lower bound of branching fraction from other formation channels in the detected population to be $\sim 0.2$. Assuming all remaining events to be formed through either stable mass transfer or common envelope channels, we find moderate to strong evidence in favour of models with inefficient common envelopes., Comment: 26 pages, 13 figures, accepted for publication in A&A

Clusters of galaxies are useful tools to constrain cosmological parameters, only if their masses can be correctly inferred from observations. In particular, X-ray and Sunyaev-Zeldovich (SZ) effect observations can be used to derive masses within the framework of the hydrostatic equilibrium. Therefore, it is crucial to have a good control of the possible mass biases that can be introduced when this hypothesis is not valid. In this work, we analyzed a set of 260 synthetic clusters from the MUSIC simulation project, at redshifts $0 \leq z \leq 0.82$. We estimate the hydrostatic mass of the MUSIC clusters from X-ray only (temperature and density) and from X-ray and SZ (density and pressure). Then, we compare them with the true 3D dynamical mass. The biases are of the order of 20%. We find that using the temperature instead of the pressure leads to a smaller bias, although the two values are compatible within 1$\sigma$. Non-thermal contributions to the total pressure support, arising from bulk motion and turbulence of the gas, are also computed and show that they are sufficient to account for this bias. We also present a study of the correlation between the mass bias and the dynamical state of the clusters. A clear correlation is shown between the relaxation state of the clusters and the bias factor. We applied the same analysis on a subsample of 32 objects, already selected for supporting the NIKA2 SZ Large Program., Comment: 19 pages, 16 figures. Accepted for publication in MNRAS

This paper describes our participation in SemEval-2020 Task 12: Multilingual Offensive Language Detection. We jointly-trained a single model by fine-tuning Multilingual BERT to tackle the task across all the proposed languages: English, Danish, Turkish, Greek and Arabic. Our single model had competitive results, with a performance close to top-performing systems in spite of sharing the same parameters across all languages. Zero-shot and few-shot experiments were also conducted to analyze the transference performance among these languages. We make our code public for further research, Comment: Github repo: https://github.com/finiteautomata/offenseval2020