Treat external label IDs, model classes, and deterministic arguments as separate contracts

Sparse external label identifiers should never double as dense model-head positions. Normalize the catalog into explicit bidirectional ID/class mappings, construct sibling negatives in external-ID space, convert only at the loss boundary, and map predictions back before evaluation. Keep transcript-derived arguments in a separate deterministic resolver that returns empty values whenever evidence is incomplete or conflicting. Finally, report whether hard-negative examples actually activated; a configured objective with zero eligible pairs is not evidence of learning.

Applies when: a classifier emits stable deployed identifiers and also reconstructs structured arguments from noisy text or audio. Keywords: sparse labels class index hard negatives fail closed extraction metrics