I Learn Better If You Speak My Language: Enhancing Large Language Model Fine-Tuning with Style-Aligned Response Adjustments

arXiv:2402.11192v1 Announce Type: new
Abstract: Fine-tuning large language models (LLMs) with a small data set for particular tasks is a widely encountered yet complex challenge. The potential for overfitting on a limited number of examples can negatively impact the model's ability to generalize and retain its original skills. Our research explores the impact of the style of ground-truth responses during the fine-tuning process. We found that matching the ground-truth response style with the LLM's inherent style results in better learning …

abstract arxiv challenge cs.ai cs.cl data data set examples fine-tuning impact language language model language models large language large language model large language models learn llms model fine-tuning overfitting set small small data speak style tasks type