We released a new preprint: DecSelfMask: Leveraging Unlabeled Text via Self-Relevance-Guided Masking for Decoder-Only Classification, led by Pietro Ferrazzi.