Efficient Embedding-based Synthetic Data Generation for Complex Reasoning Tasks

ArXi:2603.22294v1 Announce Type: cross Synthetic Data Generation (SDG), leveraging Large Language Models (LLMs), has recently been recognized and broadly adopted as an effective approach to improve the performance of smaller but resource and compute efficient LLMs through fine-tuning. A key challenge in SDG is ensuring the quality and diversity of the generated data.