Skip to main content
. 2024 Jan 23;3(1):e0000433. doi: 10.1371/journal.pdig.0000433

Fig 2. Variations in the scoring rates across languages, prompt adjusting levels, and GPT models.

Fig 2

Translating the Japanese questions into English text improved the scoring rates; however, it increased the output error rate. Upon further adjusting the prompts, the scoring rates improved, and the output error decreased. Moreover, switching from the GPT-3.5 model to the GPT-4 model enhanced the scoring rates and almost eliminated errors.