Researchers have developed a method to convert structured data into text across multiple domains without needing specific training data. They tested this approach against other methods and found that it outperformed them in various benchmarks. The technique uses knowledge distillation and augmentation strategies to improve performance. This work aims to address the challenge of generating text from diverse data sources, such as tables, graphs, and charts, without relying on task-specific training data or large language models. --- Why it matters: This research matters because it provides a solution for generating text from structured data across various domains, which is essential in applications like data summarization, knowledge graph-to-text generation, and conversational AI. The ability to perform well without requiring extensive domain-specific training data can significantly reduce the cost and effort of deploying such systems. Source: https://arxiv.org/abs/2608.23391