Search

The Solar Neutrino and Astro-Particle PhYsics (SNAPPY) Cubesat is expected to launch in 2025 and it will carry into a polar orbit a prototype test detector for solar neutrino background studies while over the Earth's poles for the neutrino Solar Orbiting Laboratory future project ($\nu$SOL). During this flight it is possible to do other science measurements. One of these is an improved study of the solar wind particles through better particle identification and energy measurements. This study aimed to understand how well could the solar wind particles be identified using the planned detector but instead of using the veto array as an anti-coincidence it would be used as a $\Delta$E energy sampling of a phoswich particle ID system., Comment: 2024 NASA Jumpstart Student Report

We have developed a neutrino detector with threshold energies from ~0.115 to 105 MeV in a clean detection mode almost completely void of accidental backgrounds. It was initially developed for the NASA $\nu$SOL project to put a solar neutrino detector very close to the Sun with 1,000 to 10,000 times higher solar neutrino flux than on Earth. Similar interactions have been found for anti-neutrinos, which were initially intended for Beta decay neutrinos from reactors, geological sources, or for nuclear security applications. These techniques work at the 1 to 100 MeV region for neutrinos from the ORNL Spallation Neutron Source or low energy accelerator neutrino and anti-neutrino production targets less than $\sim$100 MeV. The identification process is clean, with a double pulse detection signature within a time window between the first interaction producing the conversion electron or positron and the secondary gamma emission 100 ns to ~1 $\mu$s, which removes most accidental backgrounds. These new modes for neutrino and anti-neutrino detection of low energy neutrinos and anti-neutrinos could allow improvements to neutrino interaction measurements from an accelerator beam on a target., Comment: Contribution to the 25th International Workshop on Neutrinos from Accelerators

NLP benchmarks rely on standardized datasets for training and evaluating models and are crucial for advancing the field. Traditionally, expert annotations ensure high-quality labels; however, the cost of expert annotation does not scale well with the growing demand for larger datasets required by modern models. While crowd-sourcing provides a more scalable solution, it often comes at the expense of annotation precision and consistency. Recent advancements in large language models (LLMs) offer new opportunities to enhance the annotation process, particularly for detecting label errors in existing datasets. In this work, we consider the recent approach of LLM-as-a-judge, leveraging an ensemble of LLMs to flag potentially mislabeled examples. Through a case study of four datasets from the TRUE benchmark, covering different tasks and domains, we empirically analyze the labeling quality of existing datasets, and compare expert, crowd-sourced, and our LLM-based annotations in terms of agreement, label quality, and efficiency, demonstrating the strengths and limitations of each annotation method. Our findings reveal a substantial number of label errors, which, when corrected, induce a significant upward shift in reported model performance. This suggests that many of the LLMs so-called mistakes are due to label errors rather than genuine model failures. Additionally, we discuss the implications of mislabeled data and propose methods to mitigate them in training to improve model performance.

Understanding protein motion within the cell is crucial for predicting reaction rates and macromolecular transport in the cytoplasm. A key question is how crowded environments affect protein dynamics through hydrodynamic and direct interactions at molecular length scales. Using megahertz X-ray Photon Correlation Spectroscopy (MHz-XPCS) at the European X-ray Free Electron Laser (EuXFEL), we investigate ferritin diffusion at microsecond time scales. Our results reveal anomalous diffusion, indicated by the non-exponential decay of the intensity autocorrelation function $g_2(q,t)$ at high concentrations. This behavior is consistent with the presence of cage-trapping in between the short- and long-time protein diffusion regimes. Modeling with the $\delta\gamma$-theory of hydrodynamically interacting colloidal spheres successfully reproduces the experimental data by including a scaling factor linked to the protein direct interactions. These findings offer new insights into the complex molecular motion in crowded protein solutions, with potential applications for optimizing ferritin-based drug delivery, where protein diffusion is the rate-limiting step.

Large Language Models (LLMs) show significant potential in economic and strategic interactions, where communication via natural language is often prevalent. This raises key questions: Do LLMs behave rationally? Can they mimic human behavior? Do they tend to reach an efficient and fair outcome? What is the role of natural language in the strategic interaction? How do characteristics of the economic environment influence these dynamics? These questions become crucial concerning the economic and societal implications of integrating LLM-based agents into real-world data-driven systems, such as online retail platforms and recommender systems. While the ML community has been exploring the potential of LLMs in such multi-agent setups, varying assumptions, design choices and evaluation criteria across studies make it difficult to draw robust and meaningful conclusions. To address this, we introduce a benchmark for standardizing research on two-player, sequential, language-based games. Inspired by the economic literature, we define three base families of games with consistent parameterization, degrees of freedom and economic measures to evaluate agents' performance (self-gain), as well as the game outcome (efficiency and fairness). We develop an open-source framework for interaction simulation and analysis, and utilize it to collect a dataset of LLM vs. LLM interactions across numerous game configurations and an additional dataset of human vs. LLM interactions. Through extensive experimentation, we demonstrate how our framework and dataset can be used to: (i) compare the behavior of LLM-based agents to human players in various economic contexts; (ii) evaluate agents in both individual and collective performance measures; and (iii) quantify the effect of the economic characteristics of the environments on the behavior of agents.

Large language models (LLMs) often produce errors, including factual inaccuracies, biases, and reasoning failures, collectively referred to as "hallucinations". Recent studies have demonstrated that LLMs' internal states encode information regarding the truthfulness of their outputs, and that this information can be utilized to detect errors. In this work, we show that the internal representations of LLMs encode much more information about truthfulness than previously recognized. We first discover that the truthfulness information is concentrated in specific tokens, and leveraging this property significantly enhances error detection performance. Yet, we show that such error detectors fail to generalize across datasets, implying that -- contrary to prior claims -- truthfulness encoding is not universal but rather multifaceted. Next, we show that internal representations can also be used for predicting the types of errors the model is likely to make, facilitating the development of tailored mitigation strategies. Lastly, we reveal a discrepancy between LLMs' internal encoding and external behavior: they may encode the correct answer, yet consistently generate an incorrect one. Taken together, these insights deepen our understanding of LLM errors from the model's internal perspective, which can guide future research on enhancing error analysis and mitigation.

Will a Visual Language Model (VLM)-based bot warn us about slipping if it detects a wet floor? Recent VLMs have demonstrated impressive capabilities, yet their ability to infer outcomes and causes remains underexplored. To address this, we introduce NL-Eye, a benchmark designed to assess VLMs' visual abductive reasoning skills. NL-Eye adapts the abductive Natural Language Inference (NLI) task to the visual domain, requiring models to evaluate the plausibility of hypothesis images based on a premise image and explain their decisions. NL-Eye consists of 350 carefully curated triplet examples (1,050 images) spanning diverse reasoning categories: physical, functional, logical, emotional, cultural, and social. The data curation process involved two steps - writing textual descriptions and generating images using text-to-image models, both requiring substantial human involvement to ensure high-quality and challenging scenes. Our experiments show that VLMs struggle significantly on NL-Eye, often performing at random baseline levels, while humans excel in both plausibility prediction and explanation quality. This demonstrates a deficiency in the abductive reasoning capabilities of modern VLMs. NL-Eye represents a crucial step toward developing VLMs capable of robust multimodal reasoning for real-world applications, including accident-prevention bots and generated video verification.

The close link between cognitive decline and language has fostered long-standing collaboration between the NLP and medical communities in dementia research. To examine this, we reviewed over 200 papers applying NLP to dementia related efforts, drawing from medical, technological, and NLP-focused literature. We identify key research areas, including dementia detection, linguistic biomarker extraction, caregiver support, and patient assistance, showing that half of all papers focus solely on dementia detection using clinical data. However, many directions remain unexplored: artificially degraded language models, synthetic data, digital twins, and more. We highlight gaps and opportunities around trust, scientific rigor, applicability, and cross-community collaboration, and showcase the diverse datasets encountered throughout our review: recorded, written, structured, spontaneous, synthetic, clinical, social media based, and more. This review aims to inspire more creative approaches to dementia research within the medical and NLP communities.

Extreme continuum variability in AGNs can indicate extreme changes in accretion flows onto supermassive black holes. We explore the multiwavelength nature of a continuum flare in the Seyfert LCRS B040659.9$-$385922. The all-sky X-ray surveys conducted by the eROSITA showed that its X-ray flux increased by a factor of roughly five over six months, and concurrent optical photometric monitoring with the ATLAS showed a simultaneous increase. We triggered a multiwavelength follow-up monitoring program (XMM, NICER; optical spectroscopy) to study the evolution of the accretion disk, broad-line region, and X-ray corona. During the campaign, X-ray and optical continuum flux subsided over roughly six months. We detected a soft X-ray excess near the flare peak and after it subsided, both exhibiting a power-law (nonthermal) behavior. We modeled the broadband optical/UV/X-ray spectral energy distribution at both the flare peak and post-flare times with the AGNSED model, incorporating thermal disk emission into the optical/UV and warm thermal Comptonization in the soft X-rays. Additionally, we find that the broad Heii $\lambda$4686 emission line fades significantly as the optical/UV/X-ray continuum fades, which could indicate a substantial flare of disk emission above 54 eV. We also observed a redshifted broad component in the H${\beta}$ emission line that is present during the high flux state of the source and disappears in subsequent observations. We witnessed a likely sudden strong increase in local accretion rate, which manifested itself via an increase in accretion disk emission and thermal Comptonization emission in the soft X-rays, followed by a decrease in accretion and Comptonized luminosity. The physical processes leading to such substantial variations are still an open question, and future continuous monitoring along with multi-wavelength studies will shed some light on it., Comment: 19 pages, 7 figures

We present a detailed study of the MASTER OT J030227.28+191754.5 outburst in 2021-2022, reaching an amplitude of 10.2 mag and a duration of 60 d. The detections of (1) the double-peaked optical emission lines, and (2) the early and ordinary superhumps, established that MASTER OT J030227.28+191754.5 is an extremely energetic WZ Sge-type dwarf nova (DN). Based on the superhump observations, we obtained its orbital period and mass ratio as 0.05986(1) d and 0.063(1), respectively. These are within a typical range of low-mass-ratio DNe. According to the binary parameters derived based on the thermal-tidal instability model, our analyses showed that (1) the standard disk model requires an accretion rate $\simeq$ 10$^{20}$ g s$^{-1}$ to explain its peak optical luminosity and (2) large mass was stored in the disk at the outburst onset. These cannot be explained solely by the impact of its massive ($\gtrsim$ 1.15 M$_\odot$) primary white dwarf implied by Kimura et al. (2023). Instead, we propose that the probable origin of this enormously energetic DN outburst is the even lower quiescence viscosity than other WZ Sge-type DNe. This discussion is qualitatively valid for most possible binary parameter spaces unless the inclination is low ($\lesssim 40^\circ$) enough for the disk to be bright explaining the outburst amplitude. Such low inclinations, however, would not allow detectable amplitude of early superhumps in the current thermal-tidal instability model. The optical spectra at outburst maximum showed the strong emission lines of Balmer, He I, and He II series whose core is narrower than $\sim 800$ km s$^{-1}$. Considering its binary parameters, a Keplerian disk cannot explain this narrow component, but the presumable origin is disk winds., Comment: 23 pages, 13 figures, 2 tables. Accepted by PASJ. Part of the online supplemental information is included

We investigate the spectroscopic characteristics of Intermediate Luminosity Red Transients (ILRTs), a class of elusive objects with peak luminosity between that of classical novae and standard supernovae. We present the extensive optical and near-infrared (NIR) spectroscopic monitoring of four ILRTs, namely NGC 300 2008OT-1, AT 2019abn, AT 2019ahd and AT 2019udc. First we focus on the evolution of the most prominent spectral features observed in the low resolution spectra, then we discuss more in detail the high resolution spectrum collected for NGC 300 2008OT-1 with the Very Large Telescope equipped with UVES. Finally we analyse late time spectra of NGC 300 2008OT-1 and AT 2019ahd through comparisons with both synthetic and observed spectra. Balmer and Ca lines dominate the optical spectra, revealing the presence of slowly moving circumstellar medium (CSM) around the objects. The line luminosity of H$\alpha$, H$\beta$ and Ca II NIR triplet presents a double peaked evolution with time, possibly indicative of interaction between fast ejecta and the slow CSM. The high resolution spectrum of NGC 300 2008OT-1 reveals a complex circumstellar environment, with the transient being surrounded by a slow ($\sim$30 km s$^{-1}$) progenitor wind. At late epochs, optical spectra of NGC 300 2008OT-1 and AT 2019ahd show broad ($\sim$2500 km s$^{-1}$) emission features at $\sim$6170 A and $\sim$7000 A which are unprecedented for ILRTs. We find that these lines originate most likely from the blending of several narrow lines, possibly of iron-peak elements., Comment: 28 pages, 19 figures. Submitted to A&A

We investigate the photometric characteristics of a sample of Intermediate Luminosity Red Transients (ILRTs), a class of elusive objects with peak luminosity between that of classical novae and standard supernovae. We present the multi-wavelength photometric follow-up of four ILRTs, namely NGC 300 2008OT-1, AT 2019abn, AT 2019ahd and AT 2019udc. Through the analysis and modelling of their spectral energy distribution and bolometric light curves we infer the physical parameters associated with these transients. All four objects display a single peaked light curve which ends in a linear decline in magnitudes at late phases. A flux excess with respect to a single black body emission is detected in the infrared domain for three objects in our sample, a few months after maximum. This feature, commonly found in ILRTs, is interpreted as a sign of dust formation. Mid infrared monitoring of NGC 300 2008OT-1 761 days after maximum allows us to infer the presence of $\sim$10$^{-3}$-10$^{-5}$ M$_{\odot}$ of dust, depending on the chemical composition and the grain size adopted. The late time decline of the bolometric light curves of the considered ILRTs is shallower than expected for $^{56}$Ni decay, hence requiring an additional powering mechanism. James Webb Space Telescope observations of AT 2019abn prove that the object has faded below its progenitor luminosity in the mid-infrared domain, five years after its peak. Together with the disappearance of NGC 300 2008OT-1 in Spitzer images seven years after its discovery, this supports the terminal explosion scenario for ILRTs. With a simple semi-analytical model we try to reproduce the observed bolometric light curves in the context of few M$_{\odot}$ of material ejected at few 10$^{3}$ km s$^{-1}$ and enshrouded in an optically thick circumstellar medium., Comment: 21 pages, 15 figures plus 20 additional pages of data in appendix. Submitted to A&A

Recent advancements in NLP systems, particularly with the introduction of LLMs, have led to widespread adoption of these systems by a broad spectrum of users across various domains, impacting decision-making, the job market, society, and scientific research. This surge in usage has led to an explosion in NLP model interpretability and analysis research, accompanied by numerous technical surveys. Yet, these surveys often overlook the needs and perspectives of explanation stakeholders. In this paper, we address three fundamental questions: Why do we need interpretability, what are we interpreting, and how? By exploring these questions, we examine existing interpretability paradigms, their properties, and their relevance to different stakeholders. We further explore the practical implications of these paradigms by analyzing trends from the past decade across multiple research fields. To this end, we retrieved thousands of papers and employed an LLM to characterize them. Our analysis reveals significant disparities between NLP developers and non-developer users, as well as between research fields, underscoring the diverse needs of stakeholders. For example, explanations of internal model components are rarely used outside the NLP field. We hope this paper informs the future design, development, and application of methods that align with the objectives and requirements of various stakeholders.

A small number of pulsars are known to emit giant pulses, single pulses much brighter than average. Among these is PSR J0534+2200, also known as the Crab pulsar, a young pulsar with high giant pulse rates. Long-term monitoring of the Crab pulsar presents an excellent opportunity to perform statistical studies of its giant pulses and the processes affecting them, potentially providing insight into the behavior of other neutron stars that emit bright single pulses. Here, we present an analysis of a set of 24,985 Crab giant pulses obtained from 88 hours of daily observations at a center frequency of 1.55 GHz by the 20-meter telescope at the Green Bank Observatory, spread over 461 days. We study the effects of refractive scintillation at higher frequencies than previous studies and compare methods of correcting for this effect. We also search for deterministic patterns seen in other single-pulse sources, possible periodicities seen in several rotating radio transients and fast radio bursts, and clustering of giant pulses like that seen in the repeating fast radio burst FRB121102., Comment: 21 pages, 14 figures, accepted to ApJ

This study empirically tests the $\textit{Narrative Economics}$ hypothesis, which posits that narratives (ideas that are spread virally and affect public beliefs) can influence economic fluctuations. We introduce two curated datasets containing posts from X (formerly Twitter) which capture economy-related narratives (Data will be shared upon paper acceptance). Employing Natural Language Processing (NLP) methods, we extract and summarize narratives from the tweets. We test their predictive power for $\textit{macroeconomic}$ forecasting by incorporating the tweets' or the extracted narratives' representations in downstream financial prediction tasks. Our work highlights the challenges in improving macroeconomic models with narrative data, paving the way for the research community to realistically address this important challenge. From a scientific perspective, our investigation offers valuable insights and NLP tools for narrative extraction and summarization using Large Language Models (LLMs), contributing to future research on the role of narratives in economics.

We report on findings from scintillation analyses using high-cadence observations of eight canonical pulsars with observing baselines ranging from one to three years. We obtain scintillation bandwidth and timescale measurements for all pulsars in our survey, scintillation arc curvature measurements for four, and detect multiple arcs for two. We find evidence of a previously undocumented scattering screen along the line of sight (LOS) to PSR J1645$-$0317, as well as evidence that a scattering screen along the LOS to PSR J2313$+$4253 may reside somewhere within the Milky Way's Orion-Cygnus arm. We report evidence of a significant change in the scintillation pattern in PSR J2022$+$5154 from the previous two decades of literature, wherein both the scintillation bandwidth and timescale decreased by an order of magnitude relative to earlier observations at the same frequencies, potentially as a result of a different screen dominating the observed scattering. By augmenting the results of previous studies, we find general agreement with estimations of scattering delays from pulsar observations and predictions by the NE2001 electron density model but not for the newest data we have collected, providing some evidence of changes in the ISM along various LOSs over the timespans considered. In a similar manner, we find additional evidence of a correlation between a pulsar's dispersion measure and the overall variability of its scattering delays over time. The plethora of interesting science obtained through these observations demonstrates the capabilities of the Green Bank Observatory's 20m telescope to contribute to pulsar-based studies of the interstellar medium., Comment: Accepted to ApJ

We report the study of a huge optical intraday flare on November 12, 2021, at 2 am UT, in the blazar OJ287. In the binary black hole model it is associated with an impact of the secondary black hole on the accretion disk of the primary. Our multifrequency observing campaign was set up to search for such a signature of the impact, based on a prediction made eight years earlier. The first I-band results of the flare have already been reported by \cite{2024ApJ...960...11K}. Here we combine these data with our monitoring in the R-band. There is a big change in the R-I spectral index by $1.0\pm0.1$ between the normal background and the flare, suggesting a new component of radiation. The polarization variation during the rise of the flare suggests the same. The limits on the source size place it most reasonably in the jet of the secondary black hole. We then ask why we have not seen this phenomenon before. We show that OJ287 was never before observed with sufficient sensitivity on the night when the flare should have happened according to the binary model. We also study the probability that this flare is just an oversized example of intraday variability, using the Krakow-dataset of intense monitoring between 2015 and 2023. We find that the occurrence of a flare of this size and rapidity is unlikely. In the Appendix, we give the full orbit-linked historical light curve of OJ287 as well as the dense monitoring sample of Krakow., Comment: to Appear in Astrophysical Journal Letters

When large language models are aligned via supervised fine-tuning, they may encounter new factual information that was not acquired through pre-training. It is often conjectured that this can teach the model the behavior of hallucinating factually incorrect responses, as the model is trained to generate facts that are not grounded in its pre-existing knowledge. In this work, we study the impact of such exposure to new knowledge on the capability of the fine-tuned model to utilize its pre-existing knowledge. To this end, we design a controlled setup, focused on closed-book QA, where we vary the proportion of the fine-tuning examples that introduce new knowledge. We demonstrate that large language models struggle to acquire new factual knowledge through fine-tuning, as fine-tuning examples that introduce new knowledge are learned significantly slower than those consistent with the model's knowledge. However, we also find that as the examples with new knowledge are eventually learned, they linearly increase the model's tendency to hallucinate. Taken together, our results highlight the risk in introducing new factual knowledge through fine-tuning, and support the view that large language models mostly acquire factual knowledge through pre-training, whereas fine-tuning teaches them to use it more efficiently., Comment: Accepted as a long paper at EMNLP 2024

We present photometric and spectroscopic observations of SN 2023fyq, a type Ibn supernova in the nearby galaxy NGC 4388 (D$\simeq$18~Mpc). In addition, we trace long-standing precursor emission at the position of SN 2023fyq using data from DLT40, ATLAS, ZTF, ASAS-SN, Swift, and amateur astronomer Koichi Itagaki. Precursor activity is observed up to nearly three years before the supernova explosion, with a relatively rapid rise in the final 100 days. The double-peaked post-explosion light curve reaches a luminosity of $\sim10^{43}~\rm erg\,s^{-1}$. The strong intermediate-width He lines observed in the nebular spectrum of SN 2023fyq imply the interaction is still active at late phases. We found that the precursor activity in SN 2023fyq is best explained by the mass transfer in a binary system involving a low-mass He star and a compact companion. An equatorial disk is likely formed in this process ($\sim$0.6$\rm M_{\odot}$), and the interaction of SN ejecta with this disk powers the main peak of the supernova. The early SN light curve reveals the presence of dense extended material ($\sim$0.3$\rm M_{\odot}$) at $\sim$3000$\rm R_{\odot}$ ejected weeks before the SN explosion, likely due to final-stage core silicon burning or runaway mass transfer resulting from binary orbital shrinking, leading to rapid rising precursor emission within $\sim$30 days prior to explosion. The final explosion could be triggered either by the core-collapse of the He star or by the merger of the He star with a compact object. SN 2023fyq, along with SN 2018gjx and SN 2015G, forms a unique class of Type Ibn SNe which originate in binary systems and are likely to exhibit detectable long-lasting pre-explosion outbursts with magnitudes ranging from $-$10 to $-$13., Comment: submitted to ApJ

Automatic conversion of free-text radiology reports into structured data using Natural Language Processing (NLP) techniques is crucial for analyzing diseases on a large scale. While effective for tasks in widely spoken languages like English, generative large language models (LLMs) typically underperform with less common languages and can pose potential risks to patient privacy. Fine-tuning local NLP models is hindered by the skewed nature of real-world medical datasets, where rare findings represent a significant data imbalance. We introduce SMP-BERT, a novel prompt learning method that leverages the structured nature of reports to overcome these challenges. In our studies involving a substantial collection of Crohn's disease radiology reports in Hebrew (over 8,000 patients and 10,000 reports), SMP-BERT greatly surpassed traditional fine-tuning methods in performance, notably in detecting infrequent conditions (AUC: 0.99 vs 0.94, F1: 0.84 vs 0.34). SMP-BERT empowers more accurate AI diagnostics available for low-resource languages.

Background: Recent advancements in Artificial Intelligence (AI) contributed significantly to suicide assessment, however, our theoretical understanding of this complex behavior is still limited. Objective: This study aimed to harness AI methodologies to uncover hidden risk factors that trigger or aggravate suicide behaviors. Method: The primary dataset included 228,052 Facebook postings by 1,006 users who completed the gold-standard Columbia Suicide Severity Rating Scale. This dataset was analyzed using a bottom-up research pipeline without a-priory hypotheses and its findings were validated using a top-down analysis of a new dataset. This secondary dataset included responses by 1,062 participants to the same suicide scale as well as to well-validated scales measuring depression and boredom. Results: An almost fully automated, AI-guided research pipeline resulted in four Facebook topics that predicted the risk of suicide, of which the strongest predictor was boredom. A comprehensive literature review using APA PsycInfo revealed that boredom is rarely perceived as a unique risk factor of suicide. A complementing top-down path analysis of the secondary dataset uncovered an indirect relationship between boredom and suicide, which was mediated by depression. An equivalent mediated relationship was observed in the primary Facebook dataset as well. However, here, a direct relationship between boredom and suicide risk was also observed. Conclusions: Integrating AI methods allowed the discovery of an under-researched risk factor of suicide. The study signals boredom as a maladaptive 'ingredient' that might trigger suicide behaviors, regardless of depression. Further studies are recommended to direct clinicians' attention to this burdening, and sometimes existential experience.

EX Lupi, a low-mass young stellar object, went into an accretion-driven outburst in March of 2022. The outburst caused a sudden phase change of ~ 112$^{\circ}$ $\pm$ 5$^{\circ}$ in periodically oscillating multiband lightcurves. Our high resolution spectra obtained with HRS on SALT also revealed a consistent phase change in the periodically varying radial velocities, along with an increase in the radial velocity amplitude of various emission lines. The phase change and increase of radial velocity amplitude morphologically translates to a change in the azimuthal and latitudinal location of the accretion hotspot over the stellar surface, which indicates a reconfiguration of the accretion funnel geometry. Our 3D MHD simulations reproduce the phase change for EX Lupi. To explain the observations we explored the possibility of forward shifting of the dipolar accretion funnel as well as the possibility of an emergence of a new accretion funnel. During the outburst, we also found evidence of the hotspot's morphology extending azimuthally, asymmetrically with a leading hot edge and cold tail along the stellar rotation. Our high cadence photometry showed that the accretion flow has clumps. We also detected possible clumpy accretion events in the HRS spectra, that showed episodically highly blue-shifted wings in the Ca II IRT and Balmer H lines., Comment: Accepted for publication in The Astrophysical Journal

We conducted a search for luminous outbursts prior to the explosion of Type IIn Supernovae (SNe IIn). We built a sample of 27 objects spectroscopically classified as SNe IIn, all located at $z<0.015$. Using deep archival SN fields images taken up to nearly 20 years prior from transient surveys (PTF, ZTF, DES, CHASE) and major astronomical observatories (ESO and NOAO), we found at least one outburst years to months before the explosion of seven SNe IIn, the earliest precursor being 10 years prior to the explosion of SN 2019bxq. The maximum absolute magnitudes of the outbursts range between -11.5 mag and -15 mag, and the eruptive phases last for a few weeks to a few years. The $g-r$ colour measured for three objects during their outburst is relatively red, with $g-r$ ranging between 0.5 and 1.0 mag. This is similar to the colour expected during the eruptions of Luminous Blue Variables. We noticed that the SNe with pre-SN outbursts have light curves with faster decline rates than those that do not show pre-SN outbursts. SN 2011fh is remarkable, as it is still visible 12 years after the luminous SN-like event, indicating that the progenitor possibly survived, or that the interaction is still on-going. We detect precursor activity in 29% of bona-fide SNe~IIn in our sample. However, a quantitative assessment of the observational biases affecting the sample suggests that this fraction underestimates the intrinsic precursor occurrence rate., Comment: 14 pages, 13 figures, 2 tables, accepted for publication on A&A

Queer youth face increased mental health risks, such as depression, anxiety, and suicidal ideation. Hindered by negative stigma, they often avoid seeking help and rely on online resources, which may provide incompatible information. Although access to a supportive environment and reliable information is invaluable, many queer youth worldwide have no access to such support. However, this could soon change due to the rapid adoption of Large Language Models (LLMs) such as ChatGPT. This paper aims to comprehensively explore the potential of LLMs to revolutionize emotional support for queers. To this end, we conduct a qualitative and quantitative analysis of LLM's interactions with queer-related content. To evaluate response quality, we develop a novel ten-question scale that is inspired by psychological standards and expert input. We apply this scale to score several LLMs and human comments to posts where queer youth seek advice and share experiences. We find that LLM responses are supportive and inclusive, outscoring humans. However, they tend to be generic, not empathetic enough, and lack personalization, resulting in nonreliable and potentially harmful advice. We discuss these challenges, demonstrate that a dedicated prompt can improve the performance, and propose a blueprint of an LLM-supporter that actively (but sensitively) seeks user context to provide personalized, empathetic, and reliable responses. Our annotated dataset is available for further research.

The emergence of Large Language Models (LLMs), has opened exciting possibilities for constructing computational simulations designed to replicate human behavior accurately. Current research suggests that LLM-based agents become increasingly human-like in their performance, sparking interest in using these AI agents as substitutes for human participants in behavioral studies. However, LLMs are complex statistical learners without straightforward deductive rules, making them prone to unexpected behaviors. Hence, it is crucial to study and pinpoint the key behavioral distinctions between humans and LLM-based agents. In this study, we highlight the limitations of LLMs in simulating human interactions, particularly focusing on LLMs' ability to simulate political debates on topics that are important aspects of people's day-to-day lives and decision-making processes. Our findings indicate a tendency for LLM agents to conform to the model's inherent social biases despite being directed to debate from certain political perspectives. This tendency results in behavioral patterns that seem to deviate from well-established social dynamics among humans. We reinforce these observations using an automatic self-fine-tuning method, which enables us to manipulate the biases within the LLM and demonstrate that agents subsequently align with the altered biases. These results underscore the need for further research to develop methods that help agents overcome these biases, a critical step toward creating more realistic simulations.

Human choice prediction in economic contexts is crucial for applications in marketing, finance, public policy, and more. This task, however, is often constrained by the difficulties in acquiring human choice data. With most experimental economics studies focusing on simple choice settings, the AI community has explored whether LLMs can substitute for humans in these predictions and examined more complex experimental economics settings. However, a key question remains: can LLMs generate training data for human choice prediction? We explore this in language-based persuasion games, a complex economic setting involving natural language in strategic interactions. Our experiments show that models trained on LLM-generated data can effectively predict human behavior in these games and even outperform models trained on actual human data.

Recent years have seen broad observational support for the presence of a clumpy component within the circumnuclear gas around SMBHs. In the X-ray band, individual clouds can manifest themselves when they transit the line of sight to the X-ray corona, temporarily obscuring the X-ray continuum and thereby indicating the characteristics and location of these clouds. X-ray flux monitoring with SRG/eROSITA has revealed that in the Seyfert 1 AGN EC 04570-5206, the soft X-ray flux dipped abruptly for about 10-18 months over 2020-2021, only to recover and then drop a second time by early 2022. Here, we investigate whether these flux dips and recoveries could be associated with cloud occultation events. We complemented the eROSITA scans with multiwavelength follow-up observations, including X-ray/UV observations with Swift, XMM-Newton, and NICER, along with ground-based optical photometric and spectroscopic observations to investigate the spectral and flux variability. XMM-Newton spectra confirm that the soft X-ray flux dips were caused by partial-covering obscuration by two separate clouds. The 2020-2021 event was caused by a cloud with column density near 1e22 /cm2 and a covering fraction near 0.6. The cloud in the 2022 event had a column density near 3e23 /cm2 and a covering fraction near 0.8. The optical/UV continuum flux varied minimally and the optical emission line spectra showed no variability in Balmer profiles or intensity. The transiting gas clouds are neutral or lowly-ionized, while the lower limits on their radial distances are commensurate with the dust sublimation zone (cloud 1) or the optical broad line region (cloud 2). One possible explanation is a dust-free, outflowing wind with embedded X-ray clumps. These events are the first cloud obscuration events detected in a Seyfert galaxy using eROSITA's X-ray monitoring capabilities., Comment: Accepted for publication in Astronomy and Astrophysics

We present comprehensive optical observations of SN~2021gmj, a Type II supernova (SN~II) discovered within a day of explosion by the Distance Less Than 40~Mpc (DLT40) survey. Follow-up observations show that SN~2021gmj is a low-luminosity SN~II (LL~SN~II), with a peak magnitude $M_V = -15.45$ and Fe~II velocity of $\sim 1800 \ \mathrm{km} \ \mathrm{s}^{-1}$ at 50 days past explosion. Using the expanding photosphere method, we derive a distance of $17.8^{+0.6}_{-0.4}$~Mpc. From the tail of the light curve we obtain a radioactive nickel mass of $0.014 \pm 0.001$ M$_{\odot}$. The presence of circumstellar material (CSM) is suggested by the early-time light curve, early spectra, and high-velocity H$\alpha$ in absorption. Analytical shock-cooling models of the light curve cannot reproduce the fast rise, supporting the idea that the early-time emission is partially powered by the interaction of the SN ejecta and CSM. The inferred low CSM mass of 0.025 M$_{\odot}$ in our hydrodynamic-modeling light curve analysis is also consistent with our spectroscopy. We observe a broad feature near 4600 \AA, which may be high-ionization lines of C, N, or/and He~II. This feature is reproduced by radiation-hydrodynamic simulations of red supergiants with extended atmospheres. Several LL~SNe~II show similar spectral features, implying that high-density material around the progenitor may be common among them., Comment: Accepted version at ApJ

This paper investigates the problem-solving capabilities of Large Language Models (LLMs) by evaluating their performance on stumpers, unique single-step intuition problems that pose challenges for human solvers but are easily verifiable. We compare the performance of four state-of-the-art LLMs (Davinci-2, Davinci-3, GPT-3.5-Turbo, GPT-4) to human participants. Our findings reveal that the new-generation LLMs excel in solving stumpers and surpass human performance. However, humans exhibit superior skills in verifying solutions to the same problems. This research enhances our understanding of LLMs' cognitive abilities and provides insights for enhancing their problem-solving potential across various domains.

Deep Language Models (DLMs) provide a novel computational paradigm for understanding the mechanisms of natural language processing in the human brain. Unlike traditional psycholinguistic models, DLMs use layered sequences of continuous numerical vectors to represent words and context, allowing a plethora of emerging applications such as human-like text generation. In this paper we show evidence that the layered hierarchy of DLMs may be used to model the temporal dynamics of language comprehension in the brain by demonstrating a strong correlation between DLM layer depth and the time at which layers are most predictive of the human brain. Our ability to temporally resolve individual layers benefits from our use of electrocorticography (ECoG) data, which has a much higher temporal resolution than noninvasive methods like fMRI. Using ECoG, we record neural activity from participants listening to a 30-minute narrative while also feeding the same narrative to a high-performing DLM (GPT2-XL). We then extract contextual embeddings from the different layers of the DLM and use linear encoding models to predict neural activity. We first focus on the Inferior Frontal Gyrus (IFG, or Broca's area) and then extend our model to track the increasing temporal receptive window along the linguistic processing hierarchy from auditory to syntactic and semantic areas. Our results reveal a connection between human language processing and DLMs, with the DLM's layer-by-layer accumulation of contextual information mirroring the timing of neural activity in high-order language areas.

Text-To-Image (TTI) models, such as DALL-E and StableDiffusion, have demonstrated remarkable prompt-based image generation capabilities. Multilingual encoders may have a substantial impact on the cultural agency of these models, as language is a conduit of culture. In this study, we explore the cultural perception embedded in TTI models by characterizing culture across three hierarchical tiers: cultural dimensions, cultural domains, and cultural concepts. Based on this ontology, we derive prompt templates to unlock the cultural knowledge in TTI models, and propose a comprehensive suite of evaluation techniques, including intrinsic evaluations using the CLIP space, extrinsic evaluations with a Visual-Question-Answer (VQA) model and human assessments, to evaluate the cultural content of TTI-generated images. To bolster our research, we introduce the CulText2I dataset, derived from six diverse TTI models and spanning ten languages. Our experiments provide insights regarding Do, What, Which and How research questions about the nature of cultural encoding in TTI models, paving the way for cross-cultural applications of these models., Comment: Project page: https://venturamor.github.io/CulText2IWeb/

Causal explanations of the predictions of NLP systems are essential to ensure safety and establish trust. Yet, existing methods often fall short of explaining model predictions effectively or efficiently and are often model-specific. In this paper, we address model-agnostic explanations, proposing two approaches for counterfactual (CF) approximation. The first approach is CF generation, where a large language model (LLM) is prompted to change a specific text concept while keeping confounding concepts unchanged. While this approach is demonstrated to be very effective, applying LLM at inference-time is costly. We hence present a second approach based on matching, and propose a method that is guided by an LLM at training-time and learns a dedicated embedding space. This space is faithful to a given causal graph and effectively serves to identify matches that approximate CFs. After showing theoretically that approximating CFs is required in order to construct faithful explanations, we benchmark our approaches and explain several models, including LLMs with billions of parameters. Our empirical results demonstrate the excellent performance of CF generation models as model-agnostic explainers. Moreover, our matching approach, which requires far less test-time resources, also provides effective explanations, surpassing many baselines. We also find that Top-K techniques universally improve every tested method. Finally, we showcase the potential of LLMs in constructing new benchmarks for model explanation and subsequently validate our conclusions. Our work illuminates new pathways for efficient and accurate approaches to interpreting NLP systems.

We present optical, infrared, ultraviolet, and radio observations of SN 2022xkq, an underluminous fast-declining type Ia supernova (SN Ia) in NGC 1784 ($\mathrm{D}\approx31$ Mpc), from $<1$ to 180 days after explosion. The high-cadence observations of SN 2022xkq, a photometrically transitional and spectroscopically 91bg-like SN Ia, cover the first days and weeks following explosion which are critical to distinguishing between explosion scenarios. The early light curve of SN 2022xkq has a red early color and exhibits a flux excess which is more prominent in redder bands; this is the first time such a feature has been seen in a transitional/91bg-like SN Ia. We also present 92 optical and 19 near-infrared (NIR) spectra, beginning 0.4 days after explosion in the optical and 2.6 days after explosion in the NIR. SN 2022xkq exhibits a long-lived C I 1.0693 $\mu$m feature which persists until 5 days post-maximum. We also detect C II $\lambda$6580 in the pre-maximum optical spectra. These lines are evidence for unburnt carbon that is difficult to reconcile with the double detonation of a sub-Chandrasekhar mass white dwarf. No existing explosion model can fully explain the photometric and spectroscopic dataset of SN 2022xkq, but the considerable breadth of the observations is ideal for furthering our understanding of the processes which produce faint SNe Ia., Comment: 38 pages, 16 figures, accepted for publication in ApJ, the figure 15 input models and synthetic spectra are now available at https://zenodo.org/record/8379254

We present optical and near-infrared observations of SN~2022crv, a stripped envelope supernova in NGC~3054, discovered within 12 hrs of explosion by the Distance Less Than 40 Mpc Survey. We suggest SN~2022crv is a transitional object on the continuum between SNe Ib and SNe IIb. A high-velocity hydrogen feature ($\sim$$-$20,000 -- $-$16,000 $\rm km\,s^{-1}$) was conspicuous in SN~2022crv at early phases, and then quickly disappeared around maximum light. By comparing with hydrodynamic modeling, we find that a hydrogen envelope of $\sim 10^{-3}$ \msun{} can reproduce the behaviour of the hydrogen feature observed in SN~2022crv. The early light curve of SN~2022crv did not show envelope cooling emission, implying that SN~2022crv had a compact progenitor with extremely low amount of hydrogen. The analysis of the nebular spectra shows that SN~2022crv is consistent with the explosion of a He star with a final mass of $\sim$4.5 -- 5.6 \msun{} that has evolved from a $\sim$16 -- 22 \msun{} zero-age main sequence star in a binary system with about 1.0 -- 1.7 \msun{} of oxygen finally synthesized in the core. The high metallicity at the supernova site indicates that the progenitor experienced a strong stellar wind mass loss. In order to retain a small amount of residual hydrogen at such a high metallicity, the initial orbital separation of the binary system is likely larger than $\sim$1000~$\rm R_{\odot}$. The near-infrared spectra of SN~2022crv show a unique absorption feature on the blue side of He I line at $\sim$1.005~$\mu$m. This is the first time that such a feature has been observed in a Type Ib/IIb, and could be due to \ion{Sr}{2}. Further detailed modelling on SN~2022crv can shed light on the progenitor and the origin of the mysterious absorption feature in the near infrared., Comment: accepted for publication in ApJ

We present the results of our analysis of Gaia19dke, an extraordinary microlensing event in the Cygnus constellation that was first spotted by the {\gaia} satellite. This event featured a strong microlensing parallax effect, which resulted in multiple peaks in the light curve. We conducted extensive photometric, spectroscopic, and high-resolution imaging follow-up observations to determine the mass and the nature of the invisible lensing object. Using the Milky Way priors on density and velocity of lenses, we found that the dark lens is likely to be located at a distance of $D_L =(3.05^{+4.10}_{-2.42})$kpc, and has a mass of $M_L =(0.51^{+3.07}_{-0.40}) M_\odot$. Based on its low luminosity and mass, we propose that the lens in Gaia19dke event is an isolated white dwarf., Comment: 16 pages, submitted to A&A

Intense outbursts in blazars are among the most extreme phenomena seen in extragalactic objects. Studying these events can offer important information about the energetic physical processes taking place within the innermost regions of blazars, which are beyond the resolution of current instruments. This work presents some of the largest and most rapid flares detected in the optical band from the sources 3C 279, OJ 49, S4 0954+658, Ton 599, and PG 1553+113, which are mostly TeV blazars. The source flux increased by nearly ten times within a few weeks, indicating the violent nature of these events. Such energetic events might originate from magnetohydrodynamical instabilities near the base of the jets, triggered by processes modulated by the magnetic field of the accretion disc. We explain the emergence of flares owing to the injection of high-energy particles by the shock wave passing along the relativistic jets. Alternatively, the flares may have also arisen due to geometrical effects related to the jets. We discuss both source-intrinsic and source-extrinsic scenarios as possible explanations for the observed large amplitude flux changes., Comment: 8 pages, 2 figures, 38th International Cosmic Ray Conference (ICRC2023) proceedings

The highly variable blazar OJ~287 is commonly discussed as an example of a binary black hole system. The 130 year long optical light curve is well explained by a model where the central body is a massive black hole of 18.35$\times$10$^9$ solar mass that supports a thin accretion disc. The secondary black hole of 0.15$\times$10$^9$ solar mass impacts the disc twice during its 12 year orbit, and causes observable flares. Recently, it has been argued that an accretion disc with a typical AGN accretion rate and above mentioned central body mass should be at least six magnitudes brighter than OJ~287's host galaxy and would therefore be observationally excluded. Based on the observations of OJ~287's radio jet, detailed in Marscher and Jorstad (2011), and up-to-date accretion disc models of Azadi et al. (2022), we show that the V-band magnitude of the accretion disc is unlikely to exceed the host galaxy brightness by more than one magnitude, and could well be fainter than the host. This is because accretion power is necessary to launch the jet as well as to create electromagnetic radiation, distributed across many wavelengths, and not concentrated especially on the optical V-band. Further, we note that the claimed V-band concentration of accretion power leads to serious problems while interpreting observations of other Active Galactic Nuclei. Therefore, we infer that the mass of the primary black hole and its accretion rate do not need to be smaller than what is determined in the standard model for OJ~287.

We present a summary of the results of the OJ 287 observational campaign, which was carried out during the 2021/2022 observational season. This season is special in the binary model because the major axis of the precessing binary happens to lie almost exactly in the plane of the accretion disc of the primary. This leads to pairs of almost identical impacts between the secondary black hole and the accretion disk in 2005 and 2022. In 2005, a special flare called "blue flash" was observed 35 days after the disk impact, which should have also been verifiable in 2022. We did observe a similar flash and were able to obtain more details of its properties. We describe this in the framework of expanding cloud models. In addition, we were able to identify the flare arising exactly at the time of the disc crossing from its photo-polarimetric and gamma-ray properties. This is an important identification, as it directly confirms the orbit model. Moreover, we saw a huge flare that lasted only one day. We may understand this as the lighting up of the jet of the secondary black hole when its Roche lobe is suddenly flooded by the gas from the primary disk. Therefore, this may be the first time we directly observed the secondary black hole in the OJ 287 binary system., Comment: 24 pages, 13 figures, published in Galaxies as part of the Special Issue Distant Glowing Objects: Quest for Quasars

We perform a comprehensive search for optical precursor emission at the position of SN~2023ixf using data from the DLT40, ZTF and ATLAS surveys. By comparing the current data set with precursor outburst hydrodynamical model light curves, we find that the probability of a significant outburst within five years of explosion is low, and the circumstellar material (CSM) ejected during any possible precursor outburst is likely smaller than $\sim$0.015\msun. By comparing to a set of toy models, we find that, if there was a precursor outburst, the duration must have been shorter than $\sim$100 days for a typical brightness of $M_{r}\simeq-9$ mag or shorter than 200 days for $M_{r}\simeq-8$ mag; brighter, longer outbursts would have been discovered. Precursor activity like that observed in the normal type II SN~2020tlf ($M_{r}\simeq-11.5$) can be excluded in SN~2023ixf. If the dense CSM inferred by early flash spectroscopy and other studies is related to one or more precursor outbursts, then our observations indicate that any such outburst would have to be faint and only last for days to months, or it occurred more than five years prior to the explosion. Alternatively, any dense, confined CSM may not be due to eruptive mass loss from a single red supergiant (RSG) progenitor. Taken together, the results of SN~2023ixf and SN~2020tlf indicate that there may be more than one physical mechanism behind the dense CSM inferred around some normal type II SNe., Comment: 10 pages, 4 figures