Attention is All We Need

A groundbreaking paper that introduces the Transformer model, which relies entirely on attention mechanisms, discarding recurrence and convolutions entirely.