The acquisition algorithm uses a structured two-stage selection process designed to balance breadth across thematic categories with aggressive optimization of high-performing hypotheses. In Stage 1, it samples a Rank Factor Class using a freshness-based diversity weight \((1 / (1 + \text{rounds\_done}))^{\text{diversity\_weight}}\), which systematically penalizes heavily tested categories and forces exploration of under-tested mechanisms. In Stage 2, it narrows to the claims within that selected class and applies exploration-exploitation policies—such as Thompson Sampling across Beta distribution parameters, Upper Confidence Bound (UCB), or epsilon-greedy selection—to select the exact claim tested in the subsequent round.