Apache Cassandra’s basic DML operations do not use Paxos. It is eventually consistent. Only for operations such as INSERT … IF NOT EXISTS or UPDATE … IF EXISTS… (and a few others like these which need read-modify-write atomicity) they use Paxos… they refer to these as light-weight transactions, and it requires about 4 network round-trips.
YugabyteDB uses Raft in a fundamental manner. So the key difference between the two systems is not about Raft vs. Paxos… but the scope of operations for which a distributed consensus protocol (like Raft or Paxos) is used.
To read about other differences/advantages, see here: Compare Apache Cassandra with YugabyteDB | YugabyteDB Docs