Which change makes model output more deterministic?
Why C is the answer
Lowering the temperature
Lower temperature concentrates sampling on the most likely tokens.
In detail
- A lower temperature gives more consistent output.
- Low temperature favors the most likely tokens.
Use case
An illustrative scenario, not a real incident.
A compliance bot must answer the same way each time, so temperature is lowered.
See it in different ways
mindmap
root((LLM service))
Hardware
GPU memory
Throughput
Batching
Experience
Streaming
Conversation history
Packaging
NIMflowchart TD
Q{"What matters more?"}
Q -->|"Consistency"| A0["Lower temperature"]
Q -->|"Variety"| A1["Higher temperature"]Key point: Lower temperature is more deterministic.
Why the other options are wrong
- A. Raising the temperature
- That adds randomness.
- B. Raising the top-p value
- That widens the token range.
- D. Removing the system prompt
- That changes guidance, not randomness.