What are the strengths and weaknesses associated with CBOW and Skip-gram models in word embedding tasks?

CBOW:
Strengths: Efficient training, especially with small datasets; tends to perform well with frequent words.
Weaknesses: May not capture rare words effectively; context aggregation may lead to loss of fine-grained information.
Skip-gram:
Strengths: Better at capturing rare words and their semantics; context independence allows for more nuanced embeddings.
Weaknesses: Slower training, especially with large datasets; may struggle with frequent words due to their context diversity.