Which metric is the classic choice for machine translation?
Why C is the answer
BLEU
BLEU compares output with reference translations.
In detail
- BLEU compares output with reference translations.
- It needs reference text.
Use case
An illustrative scenario, not a real incident.
A localization team scores translations with BLEU against reference translations.
See it in different ways
flowchart TD
A{"What are you evaluating?"} -->|"Translation"| B["BLEU"]
A -->|"Summaries"| C["ROUGE"]
A -->|"RAG answers"| D["Faithfulness and relevance"]
A -->|"Open-ended quality"| E["Human or LLM judged rubric"]mindmap
root((Evaluation))
Translation
BLEU
Summaries
ROUGE
RAG
Faithfulness
Errors
Slice by topicflowchart TD
Q{"What is the task?"}
Q -->|"Translation"| A0["BLEU"]
Q -->|"Summarization"| A1["ROUGE"]Key point: BLEU suits translation.
Why the other options are wrong
- A. Throughput per GPU
- That measures speed.
- B. Cross-entropy on the padding tokens
- Padding is excluded from loss.
- D. Mean squared error
- That is for numeric targets.