Hasil untuk "Animal culture"

Chaeyun Kim, YongTaek Lim, Kihyun Kim et al.

Existing red-teaming benchmarks, when adapted to new languages via direct translation, fail to capture socio-technical vulnerabilities rooted in local culture and law, creating a critical blind spot in LLM safety evaluation. To address this gap, we introduce CAGE (Culturally Adaptive Generation), a framework that systematically adapts the adversarial intent of proven red-teaming prompts to new cultural contexts. At the core of CAGE is the Semantic Mold, a novel approach that disentangles a prompt's adversarial structure from its cultural content. This approach enables the modeling of realistic, localized threats rather than testing for simple jailbreaks. As a representative example, we demonstrate our framework by creating KoRSET, a Korean benchmark, which proves more effective at revealing vulnerabilities than direct translation baselines. CAGE offers a scalable solution for developing meaningful, context-aware safety benchmarks across diverse cultures. Our dataset and evaluation rubrics are publicly available at https://github.com/selectstar-ai/CAGE-paper. (WARNING: This paper contains model outputs that can be offensive in nature.)

Darshan Singh, Arsha Nagrani, Kawshik Manikantan et al.

Recent advancements in video models have shown tremendous progress, particularly in long video understanding. However, current benchmarks predominantly feature western-centric data and English as the dominant language, introducing significant biases in evaluation. To address this, we introduce MINERVA-Cultural, a challenging benchmark for multicultural and multilingual video reasoning. MINERVA-Cultural comprises high-quality, entirely human-generated annotations from diverse, region-specific cultural videos across 18 global locales. Unlike prior work that relies on automatic translations, MINERVA-Cultural provides complex questions, answers, and multi-step reasoning steps, all crafted in native languages. Making progress on MINERVA-Cultural requires a deeply situated understanding of visual cultural context. Furthermore, we leverage MINERVA-Cultural's reasoning traces to construct evidence-based graphs and propose a novel iterative strategy using these graphs to identify fine-grained errors in reasoning. Our evaluations reveal that SoTA Video-LLMs struggle significantly, performing substantially below human-level accuracy, with errors primarily stemming from the visual perception of cultural elements. MINERVA-Cultural will be publicly available under https://github.com/google-deepmind/neptune?tab=readme-ov-file\#minerva-cultural

Vefa Tohumcu, Mehmet Cengiz, A. Hayirli et al.

ABSTRACT Background Isoproterenol (ISO) is a nonselective beta‐adrenergic receptor agonist known for its vasodilatory effects. This experiment aims to investigate whether intrauterine ISO administration could alter vascular indices and follicular development in postpartum Holstein cows. Objectives The objectives are to evaluate the effects of intrauterine ISO administration on vascular changes and its impact on follicular development compared to placebo groups. Study Design This randomized controlled study was conducted on 36 Holstein cows selected based on their health status, including only those free from reproductive, metabolic and infectious disorders. Methods The cows (n = 36) were divided into two groups as control received distilled water alone (CON, n = 18) and experiment received 4 mg ISO in 40 mL distilled water (ISO, n = 18) and four subgroups as CON‐I (n = 9), CON‐II (n = 9), ISO‐I (n = 9) and ISO‐II (n = 9) according to days of intrauterine administration (I or II represents to 1 or 2 days after ovulation, respectively). Uterine and ovarian artery blood flows were assessed before and after administration by Doppler ultrasonography. Blood samples were collected both before and after administration (on Day 1 or 2) and on Days 3, 6 and 9 post‐ovulation for hormonal analysis. Antral follicle count (AFC) was recorded on the blood sampling days. Data were analysed via mixed model ANOVA. Results Intrauterine ISO administration significantly increased the pulse rate (PR) in the ovaries (89.4 vs. 65.5 bpm, p < 0.0001) and uterus (90.6 vs. 64.2 bpm, p < 0.0001). Early AFC (1–2.9 mm) decreased, whereas small AFC (3–4.9 mm) increased in the ISO groups. The weighted average antral follicle size (WAAFS) significantly increased in the ISO group but remained unchanged in the controls. Hormonal analysis revealed elevated levels of FSH (626 vs. 468 mIU/mL), AMH (61.3 vs. 46.4 ng/L), E2 (138 vs. 122 ng/L), P4 (15.3 vs. 10.6 ng/mL), IGF‐1 (62.6 vs. 25.1 ng/mL) and IGFBP‐3 (28.4 vs. 16.5 ng/mL) in the ISO groups (p < 0.0001). Conclusions The findings indicate that intrauterine administration of ISO on Day 1 post‐ovulation could be a promising ‘adjunct technique’ for future research focussed on minimizing dependence on exogenous hormones or improving the sensitivity of follicles to endogenous hormonal signals, thereby potentially enhancing oocyte yield.

Yameng Qiao, Wenzheng Liu, Fanzhen Wang et al.

Individual recognition of Hu sheep is a core requirement for precision livestock management, significantly improving breeding efficiency and fine management. However, traditional machine vision methods face challenges such as high annotation time costs, the inability to quickly annotate new sheep, and the need for manual intervention and retraining. To address these issues, this study proposes a solution that integrates automatic annotation and transfer learning, developing a sheep face recognition algorithm that adapts to complex farming environments and can quickly learn the characteristics of new Hu sheep individuals. First, through multi-view video collection and data augmentation, a dataset consisting of 82 Hu sheep and a total of 6055 images was created. Additionally, a sheep face detection and automatic annotation algorithm was designed, reducing the annotation time per image to 0.014 min compared to traditional manual annotation. Next, the YOLOv10n-CF-Lite model is proposed, which improved the recognition precision of Hu sheep faces to 92.3%, and the mAP@0.5 to 96.2%. To enhance the model’s adaptability and generalization ability for new sheep, transfer learning was applied to transfer the YOLOv10n-CF-Lite model trained on the source domain (82 Hu sheep) to the target domain (10 new Hu sheep). The recognition precision in the target domain increased from 91.2% to 94.9%, and the mAP@0.5 improved from 96.3% to 97%. Additionally, the model’s convergence speed was improved, reducing the number of training epochs required for fitting from 43 to 14. In summary, the Hu sheep face recognition algorithm proposed in this study improves annotation efficiency, recognition precision, and convergence speed through automatic annotation and transfer learning. It can quickly adapt to the characteristics of new sheep individuals, providing an efficient and reliable technical solution for the intelligent management of livestock.

Shilva Givanny Saiful, Mohandas Indradji, Diana Indrasanti et al.

Background: Infectious coryza, caused by Avibacterium paragallinarum, is an acute and highly contagious respiratory disease in chickens that results in high morbidity, growth retardation, and decreased egg production, leading to economic losses in poultry industries. Vaccination is considered the most effective preventive measure, and inactivated vaccines are widely used due to their safety and ability to stimulate protective immunity. Purpose: This study aimed to evaluate the safety and potency of a commercially produced inactivated coryza vaccine using Specific Pathogen-Free (SPF) chickens at the National Quality Testing and Certification Center for Veterinary Drugs (BPMSPH), Indonesia. Method: A descriptive observational approach was used to assess safety and potency in vaccinated SPF chickens. A total of 40 SPF chickens four weeks old were used and divided into a vaccinated group and a control group. The vaccinated group received the inactivated coryza vaccine according to standard test procedures, while the control group remained unvaccinated. Clinical observations and local reaction assessments were conducted to evaluate safety, and antibody titers against coryza serotype A were measured to determine potency. Results: Observations showed that 100% of both control and vaccinated chickens exhibited no abnormal clinical signs or coryza reactions. In addition, the vaccine potency test showed that 100% of vaccinated chickens had coryza serotype A antibody titers ≥10. Conclusion: This test confirms that the registered inactivated coryza vaccine formulation meets safety test criteria: 100% of control and vaccinated chickens remained clinically normal, and no abnormal local reactions were observed at the inoculation site. Potency testing demonstrated that all vaccinated chickens developed serotype A antibodies, which are expected to provide protection against coryza infection in the field and improve poultry survival.

Sharifa Sultana, Hafsah Mahzabin Chowdhury, Zinnat Sultana et al.

Reproductive well-being education in the Global South is often challenged as many communities perceive many of its contents as misinformation, misconceptions, and language-inappropriate. Our ten-month-long ethnographic study (n=41) investigated the impact of sociocultural landscape, cultural beliefs, and healthcare infrastructure on Bangladeshi people's access to quality reproductive healthcare and set four design goals: combating misinformation, including culturally appropriate language, professionals' accountable moderation, and promoting users' democratic participation. Building on the model of `\textit{Distributive Justice,}' we designed and evaluated \textit{`Socheton,'} a culturally appropriate AI-mediated tool for reproductive well-being that includes healthcare professionals, AI-language teachers, and community members to moderate and run the activity-based platform. Our user study (n=28) revealed that only combating misinformation and language inappropriateness may still leave the community with a conservative mob culture and patronize reproductive care-seeking. This guides well-being HCI design toward being culturally appropriate in the context of reproductive justice with sensitive marginalized communities.

Srishti Yadav, Zhi Zhang, Daniel Hershcovich et al.

Investigating value alignment in Large Language Models (LLMs) based on cultural context has become a critical area of research. However, similar biases have not been extensively explored in large vision-language models (VLMs). As the scale of multimodal models continues to grow, it becomes increasingly important to assess whether images can serve as reliable proxies for culture and how these values are embedded through the integration of both visual and textual data. In this paper, we conduct a thorough evaluation of multimodal model at different scales, focusing on their alignment with cultural values. Our findings reveal that, much like LLMs, VLMs exhibit sensitivity to cultural values, but their performance in aligning with these values is highly context-dependent. While VLMs show potential in improving value understanding through the use of images, this alignment varies significantly across contexts highlighting the complexities and underexplored challenges in the alignment of multimodal models.

Tram Thi Minh Tran, Xinyan Yu, Marius Hoggenmueller et al.

Autonomous mobility systems increasingly operate in environments shared with animals, from urban pets to wildlife. However, their design has largely focused on human interaction, with limited understanding of how non-human species perceive, respond to, or are affected by these systems. Motivated by research in Animal-Computer Interaction (ACI) and more-than-human design, this study investigates animal interactions with autonomous mobility through a multi-method approach combining a scoping review (45 articles), online ethnography (39 YouTube videos and 11 Reddit discussions), and expert interviews (8 participants). Our analysis surfaces five key areas of concern: Physical Impact (e.g., collisions, failures to detect), Behavioural Effects (e.g., avoidance, stress), Accessibility Concerns (particularly for service animals), Ethics and Regulations, and Urban Disturbance. We conclude with design and policy directions aimed at supporting multispecies coexistence in the age of autonomous systems. This work underscores the importance of incorporating non-human perspectives to ensure safer, more inclusive futures for all species.

Gang Cheng, Xin Gao, Li Hu et al.

We introduce Wan-Animate, a unified framework for character animation and replacement. Given a character image and a reference video, Wan-Animate can animate the character by precisely replicating the expressions and movements of the character in the video to generate high-fidelity character videos. Alternatively, it can integrate the animated character into the reference video to replace the original character, replicating the scene's lighting and color tone to achieve seamless environmental integration. Wan-Animate is built upon the Wan model. To adapt it for character animation tasks, we employ a modified input paradigm to differentiate between reference conditions and regions for generation. This design unifies multiple tasks into a common symbolic representation. We use spatially-aligned skeleton signals to replicate body motion and implicit facial features extracted from source images to reenact expressions, enabling the generation of character videos with high controllability and expressiveness. Furthermore, to enhance environmental integration during character replacement, we develop an auxiliary Relighting LoRA. This module preserves the character's appearance consistency while applying the appropriate environmental lighting and color tone. Experimental results demonstrate that Wan-Animate achieves state-of-the-art performance. We are committed to open-sourcing the model weights and its source code.

Rachel Malkani, Sharmini Paramasivam, Sarah Wolfensohn

Behavioural disorders in dogs are common and have severe welfare consequences for dogs. This study aimed to assess the factors that are significant and predictive of behaviour problems in dogs using the animal welfare assessment grid (AWAG) to further understand what factors influence their welfare. 177 AWAG assessments were undertaken across 129 dogs that clinicians deemed to have a behavioural disorder. Wilcoxon rank-sum tests were used to assess the difference in scores between dogs with behaviour disorders and a cohort of healthy dogs (<i>n</i> = 117). This analysis showed that all physical factors besides body condition, all procedural factors besides procedure pain, and all psychological, and environmental factors were significantly different between healthy dogs and dogs with behaviour disorders. Spearman rank correlation coefficient (RS) revealed several significant strong positive correlations including the procedural impact on the dog’s daily routine with aggression towards unfamiliar people and procedure pain, as well as other correlations between the dog’s behaviour during assessment with the frequency at which they encounter fears and anxieties, clinical assessment and procedure pain, and reaction to stressors and social interactions. These findings highlight the interdependent nature of the various influences of welfare. Logistic regression analysis identified that aggression towards the caregiver, fears and anxieties frequency, and choice, control, and predictability were all significant predictors of behaviour disorders. The findings have important implications for veterinary, behaviour, and animal welfare professionals as any changes across these factors may indicate poor welfare linked to emotional disorders in dogs.

Aleksandr Algasov, Ekaterina Nepovinnykh, Tuomas Eerola et al.

Recent advancements in the automatic re-identification of animal individuals from images have opened up new possibilities for studying wildlife through camera traps and citizen science projects. Existing methods leverage distinct and permanent visual body markings, such as fur patterns or scars, and typically employ one of two strategies: local features or end-to-end learning. In this study, we delve into the impact of training set size by conducting comprehensive experiments across six different methods and five animal species. While it is well known that end-to-end learning-based methods surpass local feature-based methods given a sufficient amount of good-quality training data, the challenge of gathering such datasets for wildlife animals means that local feature-based methods remain a more practical approach for many species. We demonstrate the benefits of both local feature and end-to-end learning-based approaches and show that species-specific characteristics, particularly intra-individual variance, have a notable effect on training data requirements.

Ali Youssef, Kristina Vodorezova, Yannick Aarts et al.

IUMENTA (Latin for livestock) is an innovative software framework designed to construct and simulate digital twins of animals. By leveraging the powerful capability of the Open Digital Twin Platform (ODTP) alongside advanced software sensors, IUMENTA offers researchers a user-friendly tool to seamlessly develop adaptive digital replicas of animal-based processes. This framework establishes a dynamic ecosystem that integrates insights from diverse experiments, consequently enhancing our understanding of animal behavioural and physiological responses. Through real-time tracking of an animal's energy balance. IUMENTA provides valuable insights into metabolic rates, nutritional needs, emotional states, and overall well-being of animals. In this article, we explore the application of the IUMENTA framework in developing a digital twin focused on the animal's energy balance. IUMENTA includes the EnergyTag system, a state-of-the-art wearable software sensor, which facilitates real-time monitoring of energy expenditure, allowing for continuous updates and personalisation of the energy balance digital twin.

John Atanbori

Recognising animals based on distinctive body patterns, such as stripes, spots, or other markings, in night images is a complex task in computer vision. Existing methods for detecting animals in images often rely on colour information, which is not always available in night images, posing a challenge for pattern recognition in such conditions. Nevertheless, recognition at night-time is essential for most wildlife, biodiversity, and conservation applications. The SPOTS-10 dataset was created to address this challenge and to provide a resource for evaluating machine learning algorithms in situ. This dataset is an extensive collection of grayscale images showcasing diverse patterns found in ten animal species. Specifically, SPOTS-10 contains 50,000 32 x 32 grayscale images, divided into ten categories, with 5,000 images per category. The training set comprises 40,000 images, while the test set contains 10,000 images. The SPOTS-10 dataset is freely available on the project GitHub page: https://github.com/Amotica/SPOTS-10.git by cloning the repository.

Yuanjun Xu, Zhu An, Ning Huang et al.

This paper introduces the necessity and significance of the investigation of cultural heritage objects. The multi-technique method is useful for the study of cultural heritage objects, but a comprehensive analytical instrument is a better choice since it can guarantee that different types of information are always obtained from the same analytical point on the surface of cultural heritage objects, which may be crucial for some situations. Thus, the X-ray fluorescence (XRF)/X-ray diffraction (XRD) and X-ray fluorescence (XRF)/Raman spectroscopy (RS) comprehensive analytical instruments are more and more widely used to study cultural heritage objects. The two types of comprehensive analytical instruments are discussed in detail and the XRF/XRD instruments are further classified into different types on the basis of structure, type and number of detectors. A new comprehensive analytical instrument prototype that can perform XRF, XRD and RS measurements simultaneously has been successfully developed by our team and the preliminary application has shown the analysis performance and application potential. This overview contributes to better understand the research progress and development tendency of comprehensive analytical instruments for the study of cultural heritage objects. The new comprehensive instruments will make researchers obtain more valuable information on cultural heritage objects and further promote the study on cultural heritage objects.

Li Zhou, Antonia Karamolegkou, Wenyu Chen et al.

The increasing ubiquity of language technology necessitates a shift towards considering cultural diversity in the machine learning realm, particularly for subjective tasks that rely heavily on cultural nuances, such as Offensive Language Detection (OLD). Current understanding underscores that these tasks are substantially influenced by cultural values, however, a notable gap exists in determining if cultural features can accurately predict the success of cross-cultural transfer learning for such subjective tasks. Addressing this, our study delves into the intersection of cultural features and transfer learning effectiveness. The findings reveal that cultural value surveys indeed possess a predictive power for cross-cultural transfer learning success in OLD tasks and that it can be further improved using offensive word distance. Based on these results, we advocate for the integration of cultural information into datasets. Additionally, we recommend leveraging data sources rich in cultural information, such as surveys, to enhance cultural adaptability. Our research signifies a step forward in the quest for more inclusive, culturally sensitive language technologies.

Andres Ferraro, Gustavo Ferreira, Fernando Diaz et al.

Recommender systems have become the dominant means of curating cultural content, significantly influencing individual cultural experience. Since recommender systems tend to optimize for personalized user experience, they can overlook impacts on cultural experience in the aggregate. After demonstrating that existing metrics do not center culture, we introduce a new metric, commonality, that measures the degree to which recommendations familiarize a given user population with specified categories of cultural content. We developed commonality through an interdisciplinary dialogue between researchers in computer science and the social sciences and humanities. With reference to principles underpinning public service media systems in democratic societies, we identify universality of address and content diversity in the service of strengthening cultural citizenship as particularly relevant goals for recommender systems delivering cultural content. We develop commonality as a measure of recommender system alignment with the promotion of content toward a shared cultural experience across a population of users. We empirically compare the performance of recommendation algorithms using commonality with existing metrics, demonstrating that commonality captures a novel property of system behavior complementary to existing metrics. Alongside existing fairness and diversity metrics, commonality contributes to a growing body of scholarship developing `public good' rationales for machine learning systems.

Minh-Tan Pham, Hugo Gangloff, Sébastien Lefèvre

This paper studies a reconstruction-based approach for weakly-supervised animal detection from aerial images in marine environments. Such an approach leverages an anomaly detection framework that computes metrics directly on the input space, enhancing interpretability and anomaly localization compared to feature embedding methods. Building upon the success of Vector-Quantized Variational Autoencoders in anomaly detection on computer vision datasets, we adapt them to the marine animal detection domain and address the challenge of handling noisy data. To evaluate our approach, we compare it with existing methods in the context of marine animal detection from aerial image data. Experiments conducted on two dedicated datasets demonstrate the superior performance of the proposed method over recent studies in the literature. Our framework offers improved interpretability and localization of anomalies, providing valuable insights for monitoring marine ecosystems and mitigating the impact of human activities on marine animals.

Welligton Conceição da Silva, Éder Bruno Rebelo da Silva, Maria Roseane Pereira dos Santos et al.

This study aimed to evaluate the behavior and thermal comfort of 20 Girolando cows (5/8-H/G), with light and dark coats, in the wettest period of the year, in Santarém, Pará, Brazil, in pasture with access to shade, and plenty of drinking water and mineral salt. Animal behavior categories were computed for 12 h a day, on 3 days in a row, by trained observers. Three day shifts were considered: Morning (6:00 a.m. to 9:55 a.m.), Intermediate (10:00 a.m. to 01:55 p.m.) and Afternoon (2:00 p.m. to 05:55 p.m.). The Temperature Index (TI), the Black Globe Humidity Index (BGHI) and the Comfort Index (CI) were calculated to measure thermal comfort. At all times studied, BGHI pointed that the environment was outside the thermal comfort zone. Dark-coated animals spent more 34.26% of the time in activities in the shade. The light-coated animals remained more 11.88% of the time in the sun, performing their natural behaviors. Both light and dark coat animals remained more 77 and 74.44% of the time in the sun, respectively. The behavior “in the sun while grazing” was the most evident, in both coats, in the studied shifts. The behaviors “in the shade while walking” and “in the shade while standing idle” were more evident (p &lt; 0.01) in dark-coated cattle. The grazing behavior was higher in animals with dark coat (p &lt; 0.05). In all evaluated shifts, there was a positive correlation between the behavior “in the sun while grazing” with the CI (r = 0.44211; p &lt; 0.0305). Behaviors performed in the shade, such as “idleness while lying down,” “ruminating while lying down and standing up,” and behaviors “in the sun,” “idleness while lying down” and “ruminating while lying down,” were negatively correlated with CI. It is concluded that, even in the wettest period of the year, in the Eastern Amazon, Girolando dairy cows are exposed to hot environments, which causes thermal discomfort and changes in their natural behavior, as they spend more time standing in shaded areas, usually in rumination. Also, light-coated cows spend more time in the sun, while dark-coated cows spend more time in the shade. Thus, light-coated cows tend to have health and zootechnical performance negatively affected.

M.J. Doran, F.J. Mulligan, M.B. Lynch et al.

ABSTRACT: The objectives of this study are to evaluate the effects of (1) a potential interaction between supplement crude protein (CP) concentration and differing cow genotypes on milk production, (2) differing cow genotypes on milk production, and (3) decreasing the supplement CP concentration on milk production and N excretion during the main grazing season within a spring-calving herd. A 2 × 2 factorial arrangement experiment, with 2 feeding strategies [14%; n = 30 (lower CP; LCP) and 18%; n = 28 (higher CP; HCP) CP concentrate supplements] offered at varying levels according to pasture availability and days in milk (DIM) was conducted over the main grazing season from April 3 to September 3, 2019, at University College Dublin Lyons Farm. Cows were also grouped into 2 genotype groups: lower milk genotype; n = 30 [LM; milk kg predicted transmitting ability (PTA): 45 ± 68.6 (mean ± SD); fat kg PTA: 10 ± 4.9; and protein kg PTA: 7 ± 2.3] and higher milk genotype; n = 28 [HM; milk kg PTA: 203 ± 55.0; fat kg PTA: 13 ± 3.8; and protein kg PTA: 10 ± 2.4]. A total of 46 multiparous and 12 primiparous (total; 58) Holstein Friesian dairy cows were blocked on parity and balanced on DIM, body condition score, and Economic Breeding Index. Cows were offered a basal diet of grazed perennial ryegrass pasture. The N partitioning study took place from August 25 to 30, 2019 (187 ± 15.2 DIM). No interactions were observed for any milk production or milk composition parameter. No effect of supplement CP concentration was observed for any total accumulated milk production, daily milk production, or milk composition parameter measured. The HM cows had increased daily milk yield (+1.9 kg), fat and protein (+0.15 kg), and energy-corrected milk (+1.7 kg), compared with the LM cows. Furthermore, HM cows had decreased milk protein concentration (−0.1%) compared with LM cows. For the N partitioning study, cows offered LCP had increased pasture dry matter intake (PDMI; +0.9 kg/d), dietary N intake (+0.022 kg/d), feces N excretion (+0.016 kg/d), and decreased N partitioning to milk (−2%), and N utilization efficiency (−2.3%). In conclusion, offering cows LCP had no negative influence on milk production or milk composition over the main grazing season where high pasture quality was maintained. However, any potential negative effects of offering LCP on milk production may have been offset by the increased PDMI. Furthermore, offering cows LCP decreased N utilization efficiency due to the higher PDMI and feed N intake associated with cows on this treatment in our study.

Conor Moloney, Antonella Puggioni, Myles McKenna

Abstract A 9‐week‐old male intact Cavalier King Charles Spaniel was presented for evaluation of acute onset dyspnea caused by left‐sided pneumothorax. Thoracic computed tomography (CT) identified multiple pulmonary bullae and blebs in multiple lung lobes. Rupture of ≥1 pulmonary blebs or bullae, precipitated by low impact trauma, was the suspected cause of pneumothorax. A volume of 7.5 mL/kg of fresh whole blood was collected from a type‐matched donor dog and administered into the left pleural space using a thoracostomy tube. The pneumothorax was successfully resolved and no adverse effects of blood patch pleurodesis were noted. The dog was clinically normal 12 months later.