When comparing contextualized word embeddings such as BERT versus static Word$2$Vec embeddings, which of the following statement$($s$)$ is$/$are true? $($Select all answers that apply$)$

Group of answer choices

In contrst to contextualized word embeddings, Word$2$Vec learns a single fixed embedding per word type.

A Word$2$Vec model that produces $300-$d word embeddings has fewer parameters than an BERT$-$like model $($e$.$g$.,$ ELMo$)$ that produces $300-$d embeddings.

Unlike Word$2$Vec, BERT is capable of computing representations for word types that are unseen during training.

Contextualized word embeddings transfer better to downstream tasks than static word embeddings.