Reviews: Improved Regret Bounds for Bandit Combinatorial Optimization