PRODUCTION SCENARIO
A copywriter uses a foundation model on Amazon Bedrock to draft taglines. Running the same prompt repeatedly returns nearly identical wording, and the copywriter wants a wider range of drafts.
Which inference parameter change will meet this requirement?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Increase the Top P value in the request.
Top P sets the share of the probability distribution the model samples from, and a higher value enlarges that pool so the model considers less likely tokens, which produces more varied drafts. Decreasing the temperature does the opposite: it steepens the distribution and makes responses more deterministic. Raising the maximum response length allows longer taglines, not different ones. A stop sequence says where to stop generating, not which tokens to choose.