1Mediumgeneral⏱ 45sWhat advantage does the Transformer architecture's self-attention mechanism provide over older RNN-based models?📎▼AIt processes tokens strictly one at a time, just like RNNsBIt lets the model directly attend to relationships between all tokens in a sequence in parallel, capturing long-range dependencies more effectively and enabling much faster parallelized trainingCIt eliminates the need for tokenizationDIt only works for translation tasks