Search results for ICML 2026
AI & Enterprise
KAIST develops red-teaming tech to find 7 times more diverse AI vulnerabilities
Korea Advanced Institute of Science and Technology said it developed a safety verification technology that can uncover about seven times more diverse hidden vulnerabilities in generative AI than existing methods. The team built a new red-teaming framework, Stable-GFlowNet, to verify large language model safety. Tests found 134 unique attack types versus 17 previously, while maintaining a 92 percent attack success rate. The work was selected as an ICML 2026 spotlight paper.
AI & Enterprise
KAIST develops tech to boost safety of custom AI while preserving performance
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.
AI & Enterprise
Team Naver showcases AI full-stack technology at ICML 2026
Team Naver said on Sunday it took part in the International Conference on Machine Learning (ICML) 2026, presenting research results ranging from advanced AI models to physical AI. The event was held at Seoul\'s COEX from July 6 to 11. It showcased work across AI safety, model and agent operations, and 3D spatial understanding. Highlights included red-teaming using Stable-GFlowNet, a model-merging method called SyMerge, and a Seoul World Model developed jointly with partners.
