Responses are too long for your UI. What do you change first?
Why A is the answer
The prompt, to ask for the desired length
The prompt controls how the model responds, so stating a length is the first and cheapest fix.
In detail
- Changing the prompt can adjust a response toward the desired output.
- It is the cheapest change to try first.
- Set an output token limit too, as a backstop.
Use case
An illustrative scenario, not a real incident.
Replies overflow a small panel. Adding "answer in two sentences" to the prompt fixes it at once, with no change to the deployment.
See it in different ways
flowchart LR S0["Change the prompt"] --> S1["Check the reply"] --> S2["Add a token limit if needed"]
flowchart TD
Q{"Is the reply too long?"}
Q -->|"Yes"| A0["Ask for a length in the prompt"]
Q -->|"Still too long"| A1["Add an output token limit"]Key point: Try the prompt first.
Why the other options are wrong
- B. The grants on the Unity Catalog schema for the app
- Grants control access, not length.
- C. The cluster's machine type
- Machine type changes speed, not what the model writes.
- D. The embedding model
- The embedding model affects retrieval, not reply length.