KAIST said a research team developed a learning framework designed to improve the safety of customised AI trained on personal or corporate documents while keeping task performance. The Buffer-and-Reinforce approach applies a temporary buffering module during fine-tuning to block malicious data from affecting the core model, then removes it and adds a safety reinforcement module using QR decomposition. In tests, the share of dangerous answers fell to about 8 percent from about 18 percent.

A machine-learning-based methodology developed by South Korean researchers to assess and predict frequency availability has been adopted as a new reference approach in international radio spectrum management. The Electronics and Telecommunications Research Institute said an ITU-R Study Group 1 meeting in Geneva approved a new report on spectrum availability assessment and forecasting. The methodology draws on ETRI research using mobile network data and will be issued as an official ITU-R report after final editing.