Which of the following statements best describes the role of attention mechanisms in neural networks?

Attention mechanisms let a neural network decide which parts of the input are most important for a given task, so it can give more weight to those parts and less to irrelevant ones. Instead of treating all input tokens equally, attention assigns a score to each token and uses these scores to mix the tokens into a context vector that reflects the most useful information. For example, in translating a sentence, attention helps the model look at the correct word in the source sentence when predicting each word of the output. This selective weighting improves accuracy and makes the network more efficient and interpretable.