flowchart TD
    subgraph S1[Stage 1: Class Sampling]
        A[Rank Factor Classes] --> B[Calculate Inverse Freshness]
        B --> C[Sample Under-Tested Class]
    end

    subgraph S2[Stage 2: Intra-Class Bayesian Selection]
        C --> D{Acquisition Policy}
        D -->|Thompson Sampling| E[Select Best Factor/Claim]
        D -->|Upper Confidence Bound| E
        D -->|Epsilon-Greedy| E
    end

    subgraph Eval[Evaluation Loop]
        E --> F[Generate Snippet Edit]
        F --> G[LLM Competitor Re-ranking]
        G --> H[Update Posterior Probability]
        H -.->|Feedback Loop| A
    end