Companion repository to our González-Márquez et al. (2026) paper "Cropping outperforms dropout as an augmentation strategy for self-supervised training of text embeddings" (TMLR) - View it on GitHub