Imagine you are developing a new machine translation system and you want to implement a Transformer model. Which of the following components, introduced by the key contributors, is essential for allowing the model to focus on different parts of the input sequence when generating output?

The key component is the attention mechanism, especially multi‑head self‑attention, which lets the model assign different weights to different tokens of the input while producing each output word. By computing a weighted sum of all input positions, the model can “look” at the most relevant parts of the source sentence for each generated word. This is why the Transformer can handle long sentences and capture long‑range dependencies. For example, when translating “The cat sat on the mat,” the attention for the word “sat” will focus mainly on the word “cat” and “sat” itself, while the word “mat” will attend to “on” and “the. ” This ability to focus on the right parts is what makes Transformers effective for machine translation.