Abstract: A preregistered audit of 600 images by generative AI across 150 different prompts explores the link between humor and discrimination in consumer-facing AI solutions. When ChatGPT updates images to make them "funnier", the prevalence of stereotyped groups changes. While stereotyped groups for politically sensitive traits (i.e., race and gender) are less likely to be represented after making an image funnier, stereotyped groups for less politically sensitive traits (i.e., older, visually impaired, and people with high body weight groups) are more likely to be represented.

Abstract: Improving energy efficiency, reducing consumption, and enhancing indoor thermal comfort are key concerns in sustainable architecture. While much research has addressed minimizing heating demands during winter, fewer studies have explored strategies to improve thermal comfort and reduce cooling loads during summer. This study aims to bridge that gap by analyzing the combined effects of window opening ratios, solar shading devices, and HVAC systems on summer energy performance and indoor comfort in faculty offices at Bingöl University. The hypothesis suggests that optimizing window openings, implementing suitable shading strategies, and selecting effective HVAC systems can significantly enhance thermal comfort and lower cooling energy use. The study explores four main questions: (1) How effective is natural ventilation through varying window openings? (2) How much can solar shading reduce overheating and cooling loads? (3) How do mechanical systems interact with passive design strategies? (4) What is the combined effect of all three parameters on performance? A parametric simulation approach was applied using DesignBuilder software. Scenarios included window openings from 5% to 50%, ten solar shading configurations, and five HVAC types. A total of 498 simulations generated a robust dataset for performance analysis. Results show that integrated optimization can reduce cooling energy use by up to 62% and improve thermal comfort by up to 54% compared to the base case. These findings confirm the initial hypothesis and underscore the value of holistic design strategies. In conclusion, this research offers a structured framework for improving summer thermal performance in educational office spaces. It provides actionable insights for architects, engineers, and policymakers seeking to enhance indoor environmental quality and energy efficiency in warm climate zones.

Abstract: Action recognition is a crucial task in artificial intelligence, with significant implications across various domains. We initially perform a comprehensive analysis of seven prominent action recognition methods across five widely-used datasets. This analysis reveals a critical, yet previously overlooked, observation: as the velocity of actions increases, the performance of these methods variably declines, undermining their robustness. This decline in performance poses significant challenges for their application in real-world scenarios. Building on these findings, we introduce the Velocity-Aware Action Recognition (VA-AR) framework to obtain robust action representations across different velocities. Our principal insight is that rapid actions (e.g., the giant circle backward in uneven bars or a smash in badminton) occur within short time intervals, necessitating smaller temporal attention windows to accurately capture intricate changes. Conversely, slower actions (e.g., drinking water or wiping face) require larger windows to effectively encompass the broader context. VA-AR employs a Mixture of Window Attention (MoWA) strategy, dynamically adjusting its attention window size based on the action's velocity. This adjustment enables VA-AR to obtain a velocity-aware representation, thereby enhancing the accuracy of action recognition. Extensive experiments confirm that VA-AR achieves state-of-the-art performance on the same five datasets, demonstrating VA-AR's effectiveness across a broad spectrum of action recognition scenarios.

Abstract: Understanding how neuronal diversity is generated is a major goal of neuroscience. Here we characterize the first step in generating neuronal diversity in the Drosophila embryo: spatial transcription factors (STFs) expressed in orthogonal rows and columns of neural progenitors. These factors give spatial identity to neural progenitors (neuroblasts, NBs), and are highly conserved in mammals. Here we investigate the roles of Engrailed (En+; posterior row) and Vnd+ (medial column) in specifying the well-characterized progenitor: neuroblast 7-1 (NB7-1). We show that NB7-1 is located at the intersection of Vnd and En, and we identify NB7-1 using a newly characterized gene, fd4 , that we show is specifically expressed in NB7-1 and its progeny, giving us a specific assay for NB7-1 identity. We show that En and Vnd are both required for Fd4 expression, and that Vnd and En co-expression is sufficient to induce ectopic Fd4 expression in other NBs and their lineages. Finally, we show that NBs gradually lose competence to respond to En or Vnd. We conclude that En and Vnd are STFs that act combinatorially to specify the identity of an individual progenitor, NB7-1.

Abstract: As people spend extended periods of time indoors, stress and negative emotions caused by work have become increasingly difficult to ignore. Observing window views is widely considered an effective method to alleviate stress and promote mental health. However, the specific visual elements within these views that contribute to stress reduction and the differential restorative benefits across varying compositions remain insufficiently understood. This study focuses on four major visual elements commonly seen through windows: sky, buildings, greenery, and roads. Using a horizontal layering approach, nine window views were created based on different proportions of these elements. Participants were exposed to these views, and their responses were evaluated through the positive and negative affect scale (PANAS), as well as electroencephalographic (EEG) data acquisition. The findings indicate that greenery exhibits the most pronounced positive effect on stress mitigation and the enhancement of positive affect, while the presence of roads is more likely to elicit negative emotional responses. Additionally, the visual richness and structural completeness of the window scenes are found to significantly impact restorative outcomes. These findings provide empirical insights for landscape and architectural design aimed at improving psychological well-being.

Abstract: Processing long input remains a significant challenge for large language models (LLMs) due to the scarcity of large-scale long-context training data and the high computational cost of training models for extended context windows. In this paper, we propose Ada ptive Gro uped P ositional E ncoding (AdaGroPE), a training-free, plug-and-play method to enhance long-context understanding in existing LLMs. AdaGroPE progressively increases the reuse count of relative positions as the distance grows and dynamically adapts the positional encoding mapping to sequence length, thereby fully exploiting the range of pre-trained position embeddings. Its design is consistent with the principles of rotary position embedding (RoPE) and aligns with human perception of relative distance, enabling robust performance in real-world settings with variable-length inputs. Extensive experiments across various benchmarks demonstrate that our AdaGroPE consistently achieves state-of-the-art performance, surpassing baseline methods and even outperforming LLMs inherently designed for long-context processing on certain tasks.