Value-Based vs Policy-Based Methods

Value-based methods learn values to derive a policy; policy-based methods optimize the policy directly.

How do value-based methods differ from policy-based methods?