Performance of ChatGPT-4, Gemini, and DeepSeek-V3 in answering the prompt-engineered questions from the Taiwan National Dental Technician Licensing Examination over a three-week period.

Three large language models (ChatGPT-4, Gemini, and DeepSeek-V3) were tested on Taiwan dental technician licensing exam questions over three weeks. All achieved moderate baseline accuracy (70–75%) with no significant improvement over time, though prompt-engineering benefited ChatGPT-4 and Gemini but not DeepSeek-V3.

A plain-English summary of the study's own findings — we don't add our own claims or interpretation. Read the full paper for the complete methods and results, and apply your own clinical judgement.