New method helps AI detect unfamiliar inputs while learning new tasks

Original: TOOD: Task-Aware Out-of-Distribution Score Calibration for Continual Learners

Source: arxiv.org ↗