UniGen: A Unified Framework for Textual Dataset Generation Using Large Language Models

The results indicate that U NI G EN effectively supports dynamic and evolving benchmarking, and that data augmentation improves LLM capabilities in various domains, including agent-oriented abilities and reasoning skills.