Moderate Evals Question 117 of 223

Why are BLEU and ROUGE weak metrics for open-ended LLM answers?

GenAI / LLM · Speak this in 60–90 seconds · Faridabad & Delhi NCR

PICTURE THIS: AN LLM TURN

Text inTokens
TransformerAttention
Text outNext token

Simple meaning

They reward n-gram overlap with a reference, but many correct phrasings exist.

1

WHY — Evals instead of guessing?

Why interviewers care about Evals:

They are checking judgment

on Evals.

A good answer names

the situation, the default choice, and one exception - that reads as experience.

Stay structured

Name the idea, why it exists, then one short example.

Close cleanly

End with when you use it and one common pitfall.

2

STEPS — What happens step by step?

Before you speak the answer, walk the interviewer through these steps:

  1. 1
    They reward n-gram overlap

    with a reference, but many correct phrasings exist.

  2. 2
    Why it exists

    A good paraphrase scores low

  3. 3
    a copied wrong sentence

    can score high.

  4. 4
    Prefer task-specific checks, rubrics,

    and human or judged scores.

  5. 5
    Common mistake

    What juniors usually get wrong.

  6. 6
    Close

    When you pick this over the alternative.

3

EXAMPLE — See it in action

Here's a short line you can speak, broken into clear beats:

Say this line
“A good paraphrase scores low”
Break into beats
Agoodparaphrasescoreslow
Speaking order
298740833747163290

Note: Adapt this scaffold to your own project — keep it under 60–90 seconds.

Key takeaway

They reward n-gram overlap with a reference, but many correct phrasings exist. A good paraphrase scores low

Chat with us