Study Introduces Selective Attack Method for Multi-Task Learning
A recent study introduces the Stealthy Multi-Task Attack (SMTA), a method designed to selectively target one task within a multi-task learning framework while preserving or enhancing the performance of other tasks.