Treewalker repeatedly queries a model's ungrounded memory while analyzing token-level logprobs, branching into alternative token paths at decision thresholds, and aggregating appearance rates to map the model's association landscape for a brand.

flowchart TD
    A[Brand Query] --> B[Model Parametric Generation]
    B --> C[Analyze Token Logprobs]
    C --> D{Alt Token Prob > Threshold?}
    D -- Yes --> E[Branch Alternative Paths]
    D -- No --> F[Continue Primary Sequence]
    E --> G[Aggregate Over N Runs]
    F --> G
    G --> H[Association Map & Discovered Entities]