GraphShieldMistral interactive policy-graph explorer
Explore five recorded Shieldstral classification scenarios across the
90-node hierarchy. Search categories, inspect exact policy queries,
compare validated and orphan matches, and review classified,
ambiguous, unresolved, and safe states.
This static Space visualizes recorded classifier outputs. It does not
execute the Apple MLX model in the browser. Run the linked model locally
to classify new documents and generate fresh graph exports.
GraphShieldMistral scenario clusters
The graph separates the published taxonomy structure from observed scenario classifications. Select a policy node to read the exact query used for that class. Select a scenario node to read the submitted document and its reconciled path.
Each superclass is a deterministic cluster of subcategories and leaves. Examples are safe, non-operational taxonomy exemplars. Supplied scenarios are separately grouped by their primary superclass.
Research-oriented demonstration.
The graph layer does not alter model predictions. It presents the
reconciled classifier output, including primary and secondary leaves,
raw broad-policy matches, orphan branches, and hierarchy consistency.
The recorded scenarios are examples rather than a complete safety benchmark.