RECOST: External Knowledge Guided Data-efficient Instruction Tuning

arXiv:2402.17355v1 Announce Type: new
Abstract: In the current landscape of large language models (LLMs), the process of instruction tuning serves as an essential step. Considering the high computing power overhead, data-efficient instruction tuning was proposed to reduce the training data size in this process, aiming at selecting high-quality instructional data. Nevertheless, we argue that most current data-efficient instruction-tuning methods are highly dependent on the quality of the original instruction-tuning dataset. When it comes to datasets synthesized by LLMs, a common …

abstract arxiv computing computing power cs.cl current data knowledge landscape language language models large language large language models llms power process quality reduce training training data type