ROUGE and BLEU Evaluation Metrics Quiz

Answers and explanations appear as you take the quiz.

1. Which metric was originally designed for machine translation and emphasizes precision?
   • BLEU
   • ROUGE
   • BERTScore
   • F1 only
2. What is ROUGE primarily oriented toward?
   • Precision
   • Recall
   • Latency
   • Model size
3. What does BLEU's brevity penalty prevent?
   • Producing overly long outputs
   • Using rare words
   • Gaming the score with very short outputs
   • Repeating the same word
4. What does ROUGE-L measure?
   • The number of letters in the output
   • Latency in milliseconds
   • The largest learning rate used
   • The longest common subsequence between candidate and reference
5. What is a key shared limitation of both BLEU and ROUGE?
   • They only measure surface n-gram overlap and miss paraphrase or meaning
   • They require GPUs to compute
   • They cannot handle English text
   • They need no reference text
6. How does BLEU prevent a system from gaming the score by repeating a high-frequency n-gram?
   • It ignores all repeated words
   • It clips each candidate n-gram count to its maximum count in the references
   • It only counts the first sentence
   • It penalizes any repetition to zero
7. When does BLEU’s brevity penalty reduce a candidate’s score?
   • When the candidate includes a unigram match
   • When the reference contains more than four words
   • When the candidate is shorter than the reference length
   • When the candidate uses a paraphrase
8. Why does BLEU take a geometric mean across n-gram precisions?
   • It converts precision into a confidence interval
   • It guarantees semantic equivalence
   • It removes the need for reference text
   • A low score at one n-gram order can substantially lower the combined score