From left: Professor Kim Hyunwoo of the KAIST School of Computing, researchers Lee Sang-hyeok and Song Tae-hoon, and researcher Park Ji-hwan of Korea University. Courtesy of KAIST.
■ KAIST introduced on the twenty seventh {that a} workforce led by Professor Kim Hyunwoo of the School of Computing, in a joint examine with a workforce from Korea University, has developed ‘TransMiter,’ a expertise that may successfully transplant learned knowledge between different synthetic intelligence (AI) models. The analysis findings had been orally offered on the twenty fifth (native time) at ‘AAAI 2026,’ a global convention within the subject of AI. The workforce’s method allows one AI to switch its learned know-how to one other by solely observing the predictive outputs, with out altering the advanced inner construction of the AI. This eliminates the necessity to repeat the advanced and time-consuming coaching course of, with virtually no discount in pace. Professor Kim acknowledged, “This can significantly reduce the post-training costs that had to be repeated every time a new, rapidly advancing large language model emerged,” including, “It will enable ‘model patching,’ which easily adds specialized knowledge in specific fields.”
■ The Gwangju Institute of Science and Technology (GIST) introduced on the twenty seventh {that a} analysis workforce led by Kang Hong-gyu, vice director of the Research Institute for Next-generation Energy Technology, has developed a expertise that improves the efficiency and stability of perovskite photo voltaic cells by exactly controlling the interface properties between inner layers, in collaboration with a workforce from Kangwon National University. The analysis findings had been printed within the worldwide educational journal ‘Small’ on the nineteenth of final month (native time). By growing and making use of a course of that improves the properties of the electron transport layer, which is answerable for electron motion inside the photo voltaic cell, the workforce achieved an influence conversion effectivity of twenty-two.56% in a large-area module 1 / 4 of the scale of an A4 paper. The module retained about 94% of its preliminary efficiency even after 400 hours of use. Vice Director Kang acknowledged, “This research is significant in that it solves a key challenge for large-area perovskite solar cells with a simple process.”
■ GIST introduced on the twenty seventh that it held the ‘twenty sixth Environment and Energy Olympiad Presentation and Awards Ceremony,’ hosted by the School of Earth Sciences and Environmental Engineering and sponsored by the Ministry of Climate, Energy, and Environment. This yr’s Olympiad happened for 3 weeks, from January 2nd to twenty third, with 31 college college students from throughout the nation collaborating. The collaborating college students stayed at GIST, performed analysis on assigned matters, and acquired steerage from the school. Yoo Ah-hyun, a pupil from GIST’s School of Earth Sciences and Environmental Engineering, received the Minister of Climate, Energy, and Environment Award for her analysis on ‘Evaluation of Whitlockite as an Adsorbent for Various Heavy Metals and Anions.’
– doi.org/10.48550/arXiv.2508.08604
– doi.org/10.1002/smll.202510775
Copyright ⓒ DongA Science. All rights reserved.