Subject

    Computer Science

    Topic

    Artificial Intelligence

    What is the main role of self-attention in a transformer?

    a)
    Relate each token to relevant tokens in the context
    b)
    Assign fuzzy memberships only
    c)
    Mutate model weights randomly
    d)
    Remove token order from training data

    Explanation

    Self-attention computes context-dependent mixtures so token representations incorporate relevant information from other positions.

    Find Tutors

    View all Tutors