Iterator UDFs are preferred for large datasets because they allow processing data in chunks, which can help manage memory usage and improve performance. This is particularly important in big data scenarios where the entire dataset might not fit into memory.