AI & MLRecommendedAIML-EVAL-005
Generation Metrics
BLEU, ROUGE, and their limits: overlap scores for generated text.
- Difficulty
- 3 / 5
- Importance
- 4 / 5
- Study time
- 1–2 h
- Kind of work
- Mixed
Why it matters
The learner can compute a tiny overlap score and state two failures of overlap-as-quality.