Algorithm Design and Stronger Guarantees for the Improving Multi-Armed Bandits Problem