KW — Parse failure rate

Parse failure rate. The fraction of model responses that cannot be parsed into a valid label; recorded as a metric rather than guessed, so unusable output is visible (a model with rate ≥0.98 produced no usable result).

Source: PUMA project documentation + raw benchmark results · Traceability: ANXAG-007,ANXAG-008,ANXAG-009 · Confidence: verified-at-primary-source