Analysis of Instruction Tuning for LLM in Generating Effective Advertising Copy and Related Advertising Attributes

Abstract

In recent years, Large Language Models (LLMs) have seen remarkable development, with numerous open-source LLMs being made available, facilitating their democratization. Particularly, a tuning method known as instruction tuning has emerged as a crucial technique for enhancing the performance of LLMs, and research on its effectiveness is being actively conducted. However, studies on the application of instruction tuning in specific domains are still limited, leaving its potential largely unexplored. This study analyzes to what extent LLMs can generate effective advertising text using instruction tuning. It also presents analysis results on the relationship between the advertising attributes specified in the instruction and the effectiveness of the generated advertising text, discussing key advertising attributes for LLMs to produce highly effective advertising content.