<p>The quality of classification using large language models in automatic mode decreases significantly as the number of alternatives increases. The results of an experimental study of this problem are presented using the example of automatically assigning advertisements to predefined categories. An approach that combines prompt engineering methods and pairwise comparisons is proposed.</p>

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

A Method for Improving the Quality of Multiclass Classification Using LLMs

  • D. Zharkov,
  • V. Tulchinsky

摘要

The quality of classification using large language models in automatic mode decreases significantly as the number of alternatives increases. The results of an experimental study of this problem are presented using the example of automatically assigning advertisements to predefined categories. An approach that combines prompt engineering methods and pairwise comparisons is proposed.