Controllable Learning (CL) is emerging as a crucial component of trustworthy machine learning. It emphasizes ensuring that learning models meet predefined targets and adapt to changing requirements without retraining. Let’s delve into the methods and applications of CL, particularly focusing on its implementation within Information Retrieval (IR) systems presented by researchers from Renmin University of…
Adversarial attacks are attempts to trick a machine learning model into making a wrong prediction. They work by creating slightly modified versions of real-world data (like images) that a human wouldn’t notice as different but that cause the model to misclassify them. Neural networks are known to be vulnerable to adversarial attacks, raising concerns about…
Microsoft has unveiled an extensive AI learning journey designed to deal with the diverse needs of various personas within a business, ranging from business leaders to citizen developers. This initiative is structured into four stages: Understanding AI, Preparing for AI, Using AI, and Building AI Solutions. Each stage equips organizations with the necessary skills to…
Having high-quality photographs of products is crucial in the ever-changing realm of online marketing and e-commerce. The use of artificial intelligence (AI) has the potential to alter the product photography industry completely. Users can streamline, save money, and unleash their creativity with this platform’s professional-grade product photo editing tools—all without the need for real samples.…
The research on vision-language models (VLMs) has gained significant momentum, driven by their potential to revolutionize various applications, including visual assistance for visually impaired individuals. However, current evaluations of these models often need to pay more attention to the complexities introduced by multi-object scenarios and diverse cultural contexts. Two notable studies shed light on these…
Graph comprehension and complex reasoning in artificial intelligence involve developing and evaluating the abilities of Large Language Models (LLMs) to understand and reason about graph-structured data. This field is critical for various applications, including social network analysis, drug discovery, recommendation systems, and spatiotemporal predictions. The goal is to advance the capabilities of AI to handle…
Accurately modeling magnetic hysteresis is a significant challenge in the field of AI, especially for optimizing the performance of magnetic devices such as electric machines and actuators. Traditional methods often struggle to generalize to novel magnetic fields, limiting their effectiveness in real-world applications. Addressing this challenge is crucial for developing efficient and generalizable models that…
Large Language Models (LLMs) with parametric memory of rules and knowledge have shown limitations in implicit reasoning. Research has shown that these models, even more complex ones like GPT-4, have trouble applying and integrating internalized facts reliably. For instance, even when they are aware of the entities in question, they frequently make inaccurate comparisons of…
Complex Human Activity Recognition (CHAR) in ubiquitous computing, particularly in smart environments, presents significant challenges due to the labor-intensive and error-prone process of labeling datasets with precise temporal information of atomic activities. This task becomes impractical in real-world scenarios where accurate and detailed labeling is scarce. The need for effective CHAR methods that do not…
In a recent study by Innodata, various large language models (LLMs) such as Llama2, Mistral, Gemma, and GPT were benchmarked for their performance in factuality, toxicity, bias, and propensity for hallucinations. The research introduced fourteen novel datasets designed to evaluate the safety of these models, focusing on their ability to produce factual, unbiased, and appropriate…