The field of tabular data analysis is moving towards more innovative and effective methods for prediction, embedding, and recommendation. Researchers are exploring the use of judicial data and graph structures to improve the accuracy and reliability of predictions. Additionally, there is a growing interest in developing universal models that can handle diverse domains and tasks, such as outlier detection and bias mitigation. The development of novel frameworks and algorithms is also a key trend, with a focus on improving the performance and fairness of models in various applications, including art classification and data journalism. Noteworthy papers include:
- Universal Embeddings of Tabular Data, which presents a novel framework for generating universal embeddings of tabular data.
- UniOD, a universal outlier detection framework that can detect outliers across diverse domains.
- BOOST, a novel OOD-informed model bias adaptive sampling method for bias mitigation in stylistic convolutional neural networks.