A metric measuring how well a language model predicts text, with lower values indicating better performance. Perplexity is a technical quality metric primarily useful for model comparison rather than legal analysis.
See: Benchmark; Evaluation (evals)