RobustDistiller enhances noise-invariant speech features through multi-task learning, improving model robustness and compression efficiency.