Create a poster for the Neurips 2026 conference presentation

Created using ChatSlide

Dynamic subgroups enhance machine learning by addressing the limitations of global representations. DynaSub innovatively learns and adapts through GMM subgrouping, optimizing both representation and clustering. Its effectiveness is evidenced by impressive out-of-distribution detection accuracy, achieving 86%-98% class-OOD accuracy and significantly reducing false positive rates. Future developments will focus on addressing current limitations and extending model capabilities.