BLEU and ROUGE
Older text metrics that score a generated text by its word overlap with a reference text: BLEU for translation, ROUGE for summaries.
Both count shared word sequences: BLEU leans toward precision, ROUGE toward recall. They are cheap and deterministic, but they reward copying the wording of the reference, so a correct answer in different words scores low and a fluent wrong one can score high. For open-ended LLM output they have largely been replaced by judge models and task-specific checks, though they still appear in papers and vendor reports.
Related terms
- Evals: A repeatable test set that scores an AI feature, so changes can be measured instead of guessed.
- LLM as judge: Using a model with a rubric to grade another model’s output.
- Precision (AI metric): The share of returned results that are actually correct: of everything the system flagged or retrieved, how much was right.
- Recall (AI metric): The share of all correct items that the system actually found: of everything it should have caught, how much it did.
