Alexandre Symeonidis-Herzig, Jianhe Low, Ozge Mercanoglu Sincan
et al.
Sign language production requires more than hand motion generation. Non-manual features, including mouthings, eyebrow raises, gaze, and head movements, are grammatically obligatory and cannot be recovered from manual articulators alone. Existing 3D production systems face two barriers to integrating them: the standard body model provides a facial space too low-dimensional to encode these articulations, and when richer representations are adopted, standard discrete tokenization suffers from codebook collapse, leaving most of the expression space unreachable. We propose SMPL-FX, which couples FLAME's rich expression space with the SMPL-X body, and tokenize the resulting representation with modality-specific Finite Scalar Quantization VAEs for body, hands, and face. M3T is an autoregressive transformer trained on this multi-modal motion vocabulary, with an auxiliary translation objective that encourages semantically grounded embeddings. Across three standard benchmarks (How2Sign, CSL-Daily, Phoenix14T) M3T achieves state-of-the-art sign language production quality, and on NMFs-CSL, where signs are distinguishable only by non-manual features, reaches 58.3% accuracy against 49.0% for the strongest comparable pose baseline.
Vision-language models (VLMs) have advanced human-AI interaction but struggle with cultural understanding, often misinterpreting symbols, gestures, and artifacts due to biases in predominantly Western-centric training data. In this paper, we construct CultureVerse, a large-scale multimodal benchmark covering 19, 682 cultural concepts, 188 countries/regions, 15 cultural concepts, and 3 question types, with the aim of characterizing and improving VLMs' multicultural understanding capabilities. Then, we propose CultureVLM, a series of VLMs fine-tuned on our dataset to achieve significant performance improvement in cultural understanding. Our evaluation of 16 models reveals significant disparities, with a stronger performance in Western concepts and weaker results in African and Asian contexts. Fine-tuning on our CultureVerse enhances cultural perception, demonstrating cross-cultural, cross-continent, and cross-dataset generalization without sacrificing performance on models' general VLM benchmarks. We further present insights on cultural generalization and forgetting. We hope that this work could lay the foundation for more equitable and culturally aware multimodal AI systems.
Barnaby D. A. Levin, Kalani Moore, Nicolò M. Della Ventura
et al.
The use of highly sensitive pixelated direct detectors has dramatically improved the performance of high energy instrumentation such as transmission electron microscopy. Here, we describe a recently developed monolithic active pixel sensor designed for low energy scanning electron microscopy applications. This detector enables electron backscatter diffraction (EBSD) at lower energies and dose than are accessible with existing scintillator-coupled detectors, expanding grain orientation mapping capabilities to materials such as ceramics that are poor electron conductors. The high detector sensitivity allows collection of rich diffraction information - providing dislocation defect contrast that is otherwise not accessible via EBSD. Indeed, even the energy of single electron interaction events can be measured with this detector, which we demonstrate to energy filter diffraction patterns revealing details of how diffraction occurs at low energy.
Agricultural disease management in developing countries such as India, Kenya, and Nigeria faces significant challenges due to limited access to expert plant pathologists, unreliable internet connectivity, and cost constraints that hinder the deployment of large-scale AI systems. This work introduces a cost-effective self-consistency framework to improve vision-language model (VLM) reliability for agricultural image captioning. The proposed method employs semantic clustering, using a lightweight (80MB) pre-trained embedding model to group multiple candidate responses. It then selects the most coherent caption -- containing a diagnosis, symptoms, analysis, treatment, and prevention recommendations -- through a cosine similarity-based consensus. A practical human-in-the-loop (HITL) component is incorporated, wherein user confirmation of the crop type filters erroneous generations, ensuring higher-quality input for the consensus mechanism. Applied to the publicly available PlantVillage dataset using a fine-tuned 3B-parameter PaliGemma model, our framework demonstrates improvements over standard decoding methods. Evaluated on 800 crop disease images with up to 21 generations per image, our single-cluster consensus method achieves a peak accuracy of 83.1% with 10 candidate generations, compared to the 77.5% baseline accuracy of greedy decoding. The framework's effectiveness is further demonstrated when considering multiple clusters; accuracy rises to 94.0% when a correct response is found within any of the top four candidate clusters, outperforming the 88.5% achieved by a top-4 selection from the baseline.
Erdős asked for positive integers $m<n$, such that $m$ and $n$ have the same set of prime factors, $m+1$ and $n+1$ have the same set of prime factors, and $m+2$ and $n+2$ have the same set of prime factors. No such integers are known. If one relaxes the problem and only considers the first two conditions, an infinite series of solutions is known: $m=2^k-2$, $n=(m+1)^2-1=2^k \cdot m$ for all integers $k\geq 2$. One additional solution is also known: $m=75=3\cdot 5^2$ and $n=1215=3^5 \cdot 5$ with $m+1=76=2^2\cdot 19$ and $n+1=1216=2^6 \cdot 19$. No other solutions with $n<2^{32}\approx 4.3\cdot 10^9$ were known. In this paper, we discuss an efficient algorithm to search for such integers, also known as Benelux pairs, using sieving and hashing techniques. Using highly parallel functioning algorithms on a modern consumer GPU, we could confirm the hitherto known results within a minute of computing time. Additionally, we have expanded the search space by a factor of more than $2^{16}$ and found no further solutions different from the infinite series given above up to $1.4\cdot 10^{12}>2^{40}$. For the analogous problem of integers $m<n$ with $m$ and $n+1$ having the same set of prime factors and $m+1$ and $n$having the same set of prime factors, the situation is very similar: An infinite series and one exceptional solution with $n\leq 2^{22}+2^{12}\approx 4.2\cdot 10^6$ were known. We prove that there are no other exceptional solutions with $n<1.4\cdot 10^{12}$.
Osvaldo Gallardo, Matías Milia, André Luiz Appel
et al.
This paper provides an empirical overview of the impact and practices of paying Article Processing Charges (APCs) by four nationally categorized groups of researchers in Argentina, Brazil, Mexico, and South Africa. The data was collected from 13,577 researchers through an online questionnaire. The analysis compares the practice of publishing in journals that charge APCs across different dimensions, including country, discipline, gender, and age of the researchers. The paper also focuses on the maximum amount APC paid and the methods and strategies researchers use to cover APC payments, such as waivers, research project funds, payment by coauthors, and the option to publish in closed access, where possible. Different tendencies were identified among the different disciplines and the national systems examined. Findings show that Argentine researchers apply for waivers most frequently and often use personal funds or international coauthors for APCs, with younger researchers less involved in APC payments. In contrast, Brazil, South Africa, and Mexico have more older researchers, yet younger researchers still publish more in APC journals. South African researchers lead in APC publications, likely due to better funding access and read and publish agreements. This study lays the groundwork for further analysis of gender asymmetries, funding access, and views on the commercial Open Access model of scientific dissemination.
AbstractUsing several cities in the late medieval Southern Low Countries as a case-study, this article deals with the relation between urban space and different forms of political protest. Urban commoners were aware of the powerful symbolism of certain places in the late medieval city and used that to their advantage during large-scale revolts. Yet the use of space was not limited to the dramatic occupations during these revolts. This article uncovers a wide range of strategies and tactics that common people used to act within given spaces to make their resistance possible. A spatial analysis of several instances of large- and smaller-scale resistance shows that space was intrinsically connected with how and when any form of resistance developed in late medieval cities. As such, the article aims to contribute to the literature on the importance of space in late medieval urban politics, in which attention to smaller-scale practices has been very limited.
Globally, the discourse of e-government has gathered momentum in public service delivery. No country has been left untouched in the implementation of e-government. Several government departments and agencies are now using information and communication technology (ICTs) to deliver government services and information to citizens, other government departments, and businesses. However, most of the government departments have not provided all of their services electronically or at least the most important ones. Thus, this creates a phenomenon of e-government service gaps. The objective of this study was to investigate the contextual factors enhancing e-government service gaps in a developing country. To achieve this aim, the TOE framework was employed together with a qualitative case study to guide data collection and analysis. The data was collected through semi-structured interviews from government employees who are involved in the implementation of e-government services in Zimbabwe as well as from citizens and businesses. Eleven (11) factors were identified and grouped under the TOE framework. This research contributes significantly to the implementation and utilisation of e-government services in Zimbabwe. The study also contributes to providing a strong theoretical understanding of the factors that enhance e-government service gaps explored in the research model.
Background Hepatitis B virus and hepatitis C virus are the greatest threats to blood safety for the recipient. This study aimed to determine the seroprevalence and trends of HBV and HCV infections among blood donors over a period of 5 years at Nekemte blood bank, Ethiopia. Methods A retrospective study was conducted from January 2015 to December 2019 at Nekemte blood bank. The recorded blood donors’ history and laboratory tests were reviewed by data collectors analyzed with Statistical Package for the Social Sciences version 20 software. A p-value of less than 0.005 was considered statistically significant. Results A total of 17,810 consecutive blood donors were screened between January 2015 and December 2019. The seroprevalence of HBV and HCV was 3.06% and 0.64%, respectively. The prevalence of HBV was significantly associated with male (AOR: 2.51; 95% CI: 1.17, 2.91), unmarried (AOR: 2.81; 95% CI: 1.79, 2.51) and rural (AOR: 2.11; 95% CI: 1.17, 3.05) blood donors. The prevalence of HCV was significantly associated with blood donor those were male (AOR: 3.01; 95% CI: 1.17, 3.91), within 45–65 years of age (AOR: 3.56: 95% CI: 1.14, 3.99) and unmarried (AOR: 3.14; 95% CI: 1.65, 3.96). Conclusion The current study shows the prevalence of hepatitis B virus was higher among study participants. However, the prevalence of HCV was low compared to the study conducted in other countries in Africa, a substantial percentage of the blood donors harbor HCV infections. Therefore, it is recommended to increase awareness of people (particularly on unmarried, male and rural resident) on modes of transmission and prevention of infection could help in reducing the burden of both HBV and HCV.
In 1553-1554, one Pieter Bruegel retired to Sint-Janshuis, Bergen op Zoom: a home where former servants of the Marquises of Bergen could spend their old age. The main argument of the article is that this retiree, who was the former barber-surgeon of Marquis Jan IV (1541-1567), should be considered as the father of the painter Pieter Bruegel the Elder. The latter’s origins are almost completely unknown, yet heavily debated: was he the son of a peasant who painted scenes of life in the countryside, or was he born and raised in an urban environment and did he satirise peasants in his artistic work? An historical reconstruction of the background of the retired barber-surgeon, and the retirement home he spent his final years in, shows he is a strong candidate for having fathered the famous painter. Evidence from the discipline of art history provides further support for the claim that the painter Pieter Bruegel the Elder was the son of a barber-surgeon and came from an urban social-middling-group background with close ties to one of the most important courts and artistic milieus in the Low Countries, the Renaissance palace Markiezenhof in Bergen op Zoom.
M. Moghadami, Nazanin Dadashpour, A. Mokhtari
et al.
Introduction Hepatitis B virus (HBV) is responsible for one of the most common human viral infections. An estimated 257 million people are living with chronic HBV infection worldwide, and mortality has reached 900,000 deaths in recent years. In 2001, the World Health Organization reported a prevalence of chronic hepatitis B infection in Iran between 2–7%. Objective To assess the effect of the national HBV mass vaccination program after 25 years. Methods A retrospective cohort study was conducted in vaccinated and unvaccinated people according to the year of birth. Blood samples were obtained from each enrolled person and data about demographic variables, and medical and vaccination history were collected using a standardized questionnaire. Persons were considered uninfected if they were negative for both HBsAg and anti-HBc. Also, Vaccine effectiveness was measured by calculating the risk of disease among vaccinated and unvaccinated persons and defining the percentage risk reduction of infection in the vaccinated group. Results A total of 2720 persons were interviewed. The rate of HBV breakthrough infection among the vaccinated group was significantly lower than in unvaccinated group. One hundred ninety-four cases with positive HBV markers of infection were identified. The risk ratio of HBV infection was 0.71, 95% CI: 0.54–0.94 (vaccinated/unvaccinated). The estimated vaccination effectiveness against Hepatitis B infection was 29% (95% CI: 6%–46%). Conclusions Iran has successfully combined hepatitis B vaccination into regular immunization programs. The WHO goal of reducing HBsAg prevalence to an equivalent of 1% by 2020 has been reached. With respect to vaccination effectiveness and low prevalence of the disease in the country, catch-up hepatitis B vaccination programs for adolescents can guarantee the immunity of the population.
Johis Ortega, Mary Hooshmand, Cynthia L. Foronda
et al.
ABSTRACT Development of leadership capacities in the nursing workforce is essential to achieving universal health in the Region of the Americas. This evaluation considered the effectiveness of an online leadership course offered in English and Spanish to nurses throughout Latin America and the Caribbean. The online course was an asynchronous eight-module leadership nursing course created and offered by the Pan American Health Organization (PAHO) Virtual Campus. A retrospective, descriptive design was used to evaluate learner performance data from the modules. Learner performance on the course was analyzed. Electronic surveys were distributed to individuals who withdrew prior to course completion to obtain information regarding the reason for withdrawal. In all, 289 individuals from 41 countries participated in the online course. Learner performance demonstrated improvement from pre- to post-test. The most frequent reason for not completing the course was being too busy with other obligations. The Spanish-language course version received more enrollment applications than any other course in the virtual campus’ 12-year history. The evaluation concluded that continuing education that develops nursing leadership is desired across Latin America and the Caribbean. Online education through the PAHO Virtual Campus may be a low-cost yet powerful means of disseminating knowledge to the nursing workforce throughout the Americas.
Andrew Emerick, Greg L. Bryan, Mordecai-Mark Mac Low
In order to better understand the relationship between feedback and galactic chemical evolution, we have developed a new model for stellar feedback at grid resolutions of only a few parsecs in global disk simulations, using the adaptive mesh refinement hydrodynamics code Enzo. For the first time in galaxy scale simulations, we simulate detailed stellar feedback from individual stars including asymptotic giant branch winds, photoelectric heating, Lyman-Werner radiation, ionizing radiation tracked through an adaptive ray-tracing radiative transfer method, and core collapse and Type Ia supernovae. We furthermore follow the star-by-star chemical yields using tracer fields for 15 metal species: C, N, O, Na, Mg, Si, S, Ca, Mn, Fe, Ni, As, Sr, Y, and Ba. We include the yields ejected in massive stellar winds, but greatly reduce the winds' velocities due to computational constraints. We describe these methods in detail in this work and present the first results from 500~Myr of evolution of an isolated dwarf galaxy with properties similar to a Local Group, low-mass dwarf galaxy. We demonstrate that our physics and feedback model is capable of producing a dwarf galaxy whose evolution is consistent with observations in both the Kennicutt-Schmidt relationship and extended Schmidt relationship. Effective feedback drives outflows with a greater metallicity than the ISM, leading to low metal retention fractions consistent with observations. Finally, we demonstrate that these simulations yield valuable information on the variation in mixing behavior of individual metal species within the multi-phase interstellar medium.
Mordecai-Mark Mac Low, Andreas Burkert, Juan C. Ibáñez-Mejía
A large fraction of the gas in the Galaxy is cold, dense, and molecular. If all this gas collapsed under the influence of gravity and formed stars in a local free-fall time, the star formation rate in the Galaxy would exceed that observed by more than an order of magnitude. Other star-forming galaxies behave similarly. Yet observations and simulations both suggest that the molecular gas is indeed gravitationally collapsing, albeit hierarchically. Prompt stellar feedback offers a potential solution to the low observed star formation rate if it quickly disrupts star-forming clouds during gravitational collapse. However, this requires that molecular clouds must be short-lived objects, raising the question of how so much gas can be observed in the molecular phase. This can occur only if molecular clouds form as quickly as they are destroyed, maintaining a global equilibrium fraction of dense gas. We therefore examine cloud formation timescales. We first demonstrate that supernova and superbubble sweeping cannot produce dense gas at the rate required to match the cloud destruction rate. On the other hand, Toomre gravitational instability can reach the required production rate. We thus argue that, although dense, star-forming gas may last only around a single global free-fall time, the dense gas in star-forming galaxies can globally exist in a state of dynamic equilibrium between formation by gravitational instability, and disruption by stellar feedback. At redshift z >~ 2, the Toomre instability timescale decreases, resulting in a prediction of higher molecular gas fractions at early times, in agreement with observations.