Evaluation of Google Bard on Vietna mese High School Biology Examination

Abstract

This paper presents an evaluation of the performance of Google Bard on the biology test of the Vietnamese National High School Graduation Examination (VNHSGE). The study aims to assess the accuracy and effectiveness of Google Bard in solving biology problems commonly encountered in the Vietnamese curriculum. To facilitate the evaluation, the VNHSGE dataset was translated from Vietnamese to English, as Google Bard currently lacks support for the Vietnamese language. The findings reveal that Google Bard achieved a performance rate of 49.5%, trailing behind competing models such as ChatGPT and BingChat, which achieved relatively higher scores of 58% and 69%, respectively. Notably, the translation process from Vietnamese to English might have impacted Bard's performance. Therefore, it is not recommended for Vietnamese students to solely depend on Google Bard for solving biology problems, especially if the Vietnamese language is not supported. Students should consider using language models explicitly designed for biology, such as ChatGPT or BingChat.