Explanation results

Implemented explainers return a structured result object containing explanation outputs and evaluation information.

Typical result content includes:

• local feature or region contributions;
• top-ranked influential components;
• the model output being explained;
• explanation-quality metrics;
• data required by supported plots.

The exact fields depend on the modality. Consult the generated API reference for the authoritative Python interface.

Interpretation boundary

A high-ranked feature is associated with changes in the local surrogate approximation. It is not automatically a causal factor, a globally important feature, or evidence of the model's internal reasoning process.

The metric catalogue and recommended acceptance thresholds are under construction.