SURFing to the Fundamental Limit of Jet Tagging

The study introduces the Large Shape and Textures dataset (LAS&T), a comprehensive collection of diverse shapes and textures extracted from natural images. This dataset is utilized to evaluate the performance of leading Large Vision-Language Models (VLMs) in recognizing and representing shapes and textures in various contexts. Results indicate that VLMs still lag behind human capabilities in shape recognition, particularly when variations in orientation, texture, and color are present.

Students at the University of Staffordshire expressed feelings of being 'robbed' after discovering that a course intended to launch their digital careers was largely taught using AI-generated slides. In a recorded confrontation, student James challenged a lecturer about the reliance on AI for teaching a coding module, stating, 'I do not want to be taught by GPT.' The university has since posted a policy on its course website to justify the use of AI in academic settings.

Deep vision models are complex systems that perform computations difficult to interpret. Concept-based explanation methods (CBEMs) aim to enhance interpretability by using human-understandable concepts. However, ensuring the faithfulness of these explanations, which represent the model's internal workings, involves trade-offs. Recent advancements in unsupervised CBEMs (U-CBEMs) claim to improve both interpretability and faithfulness, but these improvements may stem from complex surrogates or deletion-based methods that compromise clarity.

The article introduces 'Spot The Ball', a benchmark designed to evaluate visual social inference in vision-language models (VLMs) using sports imagery. The task involves localizing a missing sports ball in images from soccer, basketball, and volleyball. The study compares human performance against four advanced VLMs, revealing that humans are significantly more accurate, achieving 20-34% accuracy compared to the models' maximum of 17%. This highlights the limitations of current AI in understanding complex visual cues.