Why is there no single score for a good LLM output, and how do you choose a metric?
LLM & AILevel: IntermediateEvaluation
Why is there no single score for a good LLM output, and how do you choose a metric?
Hints
- Identify the stage: pretraining, adaptation, generation or evaluation.
- Name the inputs, constraints and a case where the approach fails.
Answer
Language is open-ended. The same correct answer can be written a hundred ways, so no one number captures "good". Instead you match the metric to the task: word overlap for translation, execution tests for code, a rubric judge for chat.
Source and licence: Skillumen — Awesome AI Engineer Interview Questions · CC0-1.0