Bidirectional Encoder Representations from Transformers, 340M-parameter encoder-only model revolutionizing NLP pre-training:contentReference[oaicite:4]{index=4}.