Branching paths represent alternative token candidates that meet a specific probability threshold during generation, which the system explores through recursive prompt completions to reveal latent association trees.

flowchart LR
    A["Prompt: Brand Association"] --> B["Token: Custom (85%)"]
    A --> C["Branch 1: Sports (12%)"]
    A --> D["Branch 2: Team (8%)"]
    B --> B1["Jerseys (92%)"]
    C --> C1["Apparel (88%)"]
    D --> D1["Uniforms (90%)"]