Researchers have proposed a new method for pre-training table models called TAPEX, which claims to be efficient and effective without requiring real-world data. The approach uses synthetic tables generated from existing datasets to train the model. This could potentially reduce the need for large amounts of labeled data in table-related AI tasks. --- Why it matters: This matters because it addresses a common challenge in AI research: obtaining high-quality, labeled data for training models. Efficient pre-training methods like TAPEX can help speed up development and deployment of table-based applications. Source: https://huggingface.co/blog/tapex