"Cross-Architectural Positive Pairs improve the effectiveness of Self-Supervised Learning. (arXiv:2301.12025v1 [cs.CV])" — A novel self-supervised learning approach that leverages Transformer and CNN simultaneously to overcome the issues with existing self-supervised techniques which have extreme computational requirements and suffer a substantial drop in performance with a reduction in batch size or pretraining epochs.