#beam-size

Question 1: Single choice

You raise the beam size of a translation decoder from $k$ to $2k$ (output vocabulary $\mathcal{Y}$ , maximum output length $T'$ ). What happens?

Decoding work roughly doubles; the output is usually as likely or likelier, but that is not guaranteed1
Decoding work roughly doubles, and the output is guaranteed to be at least as likely as before2
Decoding work grows by a factor of $|\mathcal{Y}|$ , because each beam is expanded over the vocabulary3
Decoding work stays the same, because only the final ranking of the candidates changes4