While Thompson sampling balances trade-offs probabilistically, the Content Optimizer also supports Upper Confidence Bound (UCB) to deterministically reward high-uncertainty claims and the epsilon-greedy method to mix random exploration with greedy exploitation, both operating downstream of the same diversity-weighting stage.