The engine runs a structured model-based evaluation pass to determine whether live research findings corroborate or refute recalled claims, updating the claim's status while attaching retrieved source citations.

flowchart LR
    A[Research Findings] --> C{LLM Evaluator}
    B[Parametric Claims] --> C
    C -->|Corroborates| D[Status: Confirmed]
    C -->|Refutes| E[Status: Contradicted]
    C -->|Unmatched Finding| F[Status: Memory Gap]
    D --- G[(Attach Citations)]
    E --- G