Publications

Alicia Parrish et al. (including Jun Seong Kim)

arXiv preprint arXiv:2607.06196

A culture-first AI safety benchmark spanning six Asia-Pacific countries, eight languages, and multimodal risks that emerge from the interaction of text and images.

Eunsu Kim, Junyeong Park, Na Min An, Jun Seong Kim, Hitesh Laxmichand Patel, Jiho Jin, Julia Kruk, Amit Agarwal, Srikant Panda, Fenal Ashokbhai Ilasariya, Hyunjung Shim, and Alice Oh

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2477–2489

Introduces CultureMix, a 23,000-image benchmark for studying how vision-language models preserve cultural identities when elements from different cultures appear together.

Jun Seong Kim, Kyaw Ye Thu, Javad Ismayilzada, Junyeong Park, Eunsu Kim, Huzama Ahmad, Na Min An, James Thorne, and Alice Oh

Proceedings of the 3rd Workshop on Cross-Cultural Considerations in NLP (C3NLP), pp. 143–154

Introduces MIXCUBE to evaluate cultural bias and robustness in multimodal large language models across mixed-cultural visual contexts.

Jun Seong Kim, Husnu Baris Baydargil, Jose Eduardo Silva, Yeon-Joo Lee, and Meeyoung Cha

한국정보과학회 2023 한국소프트웨어종합학술대회 논문집, pp. 1815–1817

Uses self-supervised anomaly-detection models on ultraviolet imager (UVI) and longwave infrared (LIR) data to detect stationary waves in Venus' atmosphere.