Scholarlink

The paper evaluates the performance of Generative AI Tools - ChatGPT 3.5, ChatGPT 4.0, and Gemini - in the Portuguese Chartered Accountant Exam (CPA). This study shows a significant variation in accuracy across 600 questions covering financial accounting, management accounting, tax, and ethics. Results show that ChatGPT 4.0 performed best overall (average accuracy of 48%), followed by Gemini (38%) and ChatGPT 3.5 (36%). It should also be noted that both ChatGPT and Gemini exhibit greater difficulty in tackling questions related to Management and Financial Accounting as compared to questions on taxation and ethics. Ultimately, these findings highlight both the promise and limitations of Generative AI Tools in challenging domain-specific exams, suggesting that further work is necessary to attain passing grades.

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

AI in Accounting: Can AI Models Like ChatGPT and Gemini Successfully Pass the Portuguese Chartered Accountant Exam?

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Agostinho Sousa Pinto,
  • António Abreu,
  • Eusébio Costa,
  • Jerónimo Paiva

摘要

The paper evaluates the performance of Generative AI Tools - ChatGPT 3.5, ChatGPT 4.0, and Gemini - in the Portuguese Chartered Accountant Exam (CPA). This study shows a significant variation in accuracy across 600 questions covering financial accounting, management accounting, tax, and ethics. Results show that ChatGPT 4.0 performed best overall (average accuracy of 48%), followed by Gemini (38%) and ChatGPT 3.5 (36%). It should also be noted that both ChatGPT and Gemini exhibit greater difficulty in tackling questions related to Management and Financial Accounting as compared to questions on taxation and ethics. Ultimately, these findings highlight both the promise and limitations of Generative AI Tools in challenging domain-specific exams, suggesting that further work is necessary to attain passing grades.