KCC Holds Explainable AI (XAI) Workshop for 2026

A significant workshop on Explainable Artificial Intelligence (XAI) convened in Jeju, gathering leading researchers from KAIST, Seoul National University, Sogang University, and the Electronics and Telecommunications Research Institute (ETRI). The event, held on June 25 at the Jeju International Convention Center, served as a major academic exchange platform for the 'Human-Centered AI Core Source Technology Development Project.'

Following the implementation of the 'Act on the Basic Law on the Development of Artificial Intelligence and Trust Building' in January, this workshop provided a crucial forum for the four XAI sub-projects. The focus was not merely on sharing achievements but on fostering genuine collaboration. Researchers engaged in heated discussions regarding methods to ensure the transparency and accountability of generative AI, particularly large language models (LLMs).

“

Interpreting and assigning reliability to black-box artificial intelligence is a monumental task that cannot be achieved through the independent research of a single lab or a specific sub-project.

Koo Myung-wan, the lead researcher for the third sub-project at Sogang University, emphasized the scale of the challenge in his opening remarks. "Interpreting and assigning reliability to black-box artificial intelligence is a monumental task that cannot be achieved through the independent research of a single lab or a specific sub-project," he stated. He stressed the necessity of "unrestricted convergence and cooperation among the top experts in each field from sub-projects one through four" to implement advanced, multimodal XAI beyond simple rule-based explanations, hoping the workshop would accelerate this synergistic integration.

“

Beyond simple rule-based explanations, implementing advanced, multimodal XAI requires unrestricted convergence and cooperation among the top experts in each field from sub-projects one through four. I hope this workshop will accelerate this synergistic integration.

The workshop featured presentations of 45 cutting-edge research papers. These covered new algorithms overcoming the limitations of existing XAI methods, techniques for providing interpretability for LLMs, and multimodal AI explanation methods. Intense Q&A sessions followed each presentation, fostering deep academic consensus. The Best Paper Award was presented to a study titled 'Interpretable Motion-Attentive Maps: Spatio-Temporally Localizing Concepts in Video Diffusion Transformers,' authored by researchers from KAIST and Yonsei University, recognizing advancements in visualizing concepts within video data.

Additionally, lawyer Won-hee Cho delivered a special lecture, examining the current state of legal tech domestically and internationally and the application of AI in the LEET (Law School Admission Test). He proposed continuous collaboration between AI researchers and legal experts to ensure transparency within the legal domain, drawing significant attention. The workshop also included brief introductions to specific outcome models derived from close collaboration within and between sub-projects, further fueling discussions. The third sub-project, led by Sogang University, exemplified interdisciplinary research, with professors from computer science, electronic engineering, and law collaborating on a pipeline for diagnostic assistance in speech disorders and threat classification for legal applications.

“

Through the intense discussions, we reconfirmed that close exchanges among researchers with diverse expertise are essential for the responsible application of AI technology in real-world social contexts. The collaborative ecosystem strengthened by the presentation of 45 papers and active idea sharing among the four sub-projects will serve as a solid foundation for opening an era of trustworthy artificial intelligence.