# Meaning of "Dependencies in "Long-term Dependencies" in LSTM?

LSTM is used instead of simple RNNs because it tackles Long-term dependencies, and solve vanishing gradient problem. I am confused a bit what is meant by the term "dependencies"?