Scholarlink

This study compares the linguistic creativity of humans and a Large Language Model (LLM) by examining their ability to modify common proverbs with single-word substitutions. A total of fifty modified proverbs are generated by substituting a single word, by both LLM and human, to enhance impact while preserving original meanings. Two experts evaluate each modified proverb on the basis of understandability and creativity. Results show that in approximately 72% of the instances, human made modifications are given preference by the experts over the modifications made by LLMs. Qualitative analysis indicates that the replacements made by LLMs exhibit lesser semantic understanding while innovating. These findings suggest that while the LLM under study possesses remarkable language capabilities, its outputs remain limited compared to human creativity.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

CLEAR Comparing LLM and Empirical Approaches to Reworking Proverbs

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Gyan Singh,
  • Vibhu Gupta,
  • Rohit Saluja

摘要

This study compares the linguistic creativity of humans and a Large Language Model (LLM) by examining their ability to modify common proverbs with single-word substitutions. A total of fifty modified proverbs are generated by substituting a single word, by both LLM and human, to enhance impact while preserving original meanings. Two experts evaluate each modified proverb on the basis of understandability and creativity. Results show that in approximately 72% of the instances, human made modifications are given preference by the experts over the modifications made by LLMs. Qualitative analysis indicates that the replacements made by LLMs exhibit lesser semantic understanding while innovating. These findings suggest that while the LLM under study possesses remarkable language capabilities, its outputs remain limited compared to human creativity.