KAIST said on Jan. 27 that a research team led by Hyunwoo Kim (김현우) jointly with a Korea University team developed a technique to transfer learned knowledge between different AI models. The team proposed TransMiter, a transferable adaptation method that reuses adaptation experience across models regardless of structure or size. It transfers know-how based on prediction results, avoiding repeated, time-consuming training and limiting slowdowns.

SK Telecom said its research on an AI recommendation model that understands personal behavioral context and explains why it makes recommendations was recognised at the international AI conference AAAI 2026 in Singapore. The paper was selected for an on-site presentation, given to about the top 4 percent of submissions, and included a poster presentation. The research covers the company’s LLM-based One Model version 4.0, including applying inference-capability reinforcement learning in a service environment.