Why might the same prompt give different results?
Why A is the answer
Generative models are nondeterministic
Nondeterminism is a listed disadvantage. Lower temperature can reduce variation.
In detail
- Generative models are nondeterministic.
- Test with several runs, not just one.
Use case
An illustrative scenario, not a real incident.
A team sends one prompt twice and gets two different summaries. Lowering temperature reduces the difference.
See it in different ways
mindmap
root((Generative AI))
Token
Unit of text
Drives cost
Embedding
Meaning as numbers
Chunking
Focused passages
Hallucination
Confident and false
Temperature
VarietysequenceDiagram participant P0 as App participant P1 as Model P0->>P1: Prompt P1-->>P0: Result A P0->>P1: Same prompt P1-->>P0: Result B
Key point: Generative models are nondeterministic.
Why the other options are wrong
- B. AWS rotates the account credentials
- Credentials do not change output.
- C. Prompts expire after a few minutes and must be sent again to a new model
- Prompts do not expire.
- D. Each Region uses a different language
- Regions do not change the language.