AIF-C01 Temperature Practice Question

A data science team is using Amazon Bedrock to generate synthetic data for training a new model. They need to ensure the generated data is diverse and covers edge cases. Which THREE parameters should they adjust to maximize diversity? (Select THREE.)

⚠ Common exam trap

AWS often tests the misconception that decreasing top-p or fixing the seed increases diversity, when in fact both actions reduce randomness and limit the model's ability to generate varied outputs.

Answer choices

Why each option matters

Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.

Correct answer & explanation

✓

Increase the top-p value to 0.9

Option E (increase the temperature) is correct because temperature scales the logits before softmax, flattening the probability distribution so lower-probability tokens are sampled more often, which directly increases output diversity and helps surface edge cases. Option A (increase top-p to 0.9) is correct because top-p (nucleus) sampling retains the smallest set of tokens whose cumulative probability reaches 0.9, keeping a broad candidate pool instead of truncating to only the most likely tokens. Option B (increase top-k to 50) is correct because top-k sampling restricts sampling to the 50 highest-probability tokens, and a larger k widens that pool, allowing more varied token choices than a small k. Option C (decrease top-p to 0.1) is not correct because a narrow nucleus of only the top ~10% cumulative probability makes generation more deterministic and less diverse. Option D (set a fixed seed) is not correct because a fixed seed makes sampling reproducible, not more diverse, and can actually reduce variation across runs.

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

✓
Increase the top-p value to 0.9
Why this is correct
Raising top-p to 0.9 widens nucleus sampling, so the model draws from a larger cumulative probability mass rather than the most likely tokens alone. This directly satisfies the diversity and edge-case coverage requirement by admitting lower-probability token choices during synthetic data generation.
✓
Increase the top-k value to 50
Why this is correct
Increasing top-k to 50 lets sampling consider the fifty highest-probability tokens instead of a narrow set, broadening candidate selection. This directly serves the diversity and edge-case coverage goal by preventing the model from repeatedly emitting only the most probable tokens.
✗
Decrease the top-p value to 0.1
Why it's wrong here
Lowering top-p to 0.1 restricts sampling to the highest-probability tokens, sharply reducing diversity and suppressing edge cases. It is tempting because low top-p values are used when deterministic, focused output is wanted, such as factual question answering, not synthetic data generation.
✗
Set the seed to a fixed value
Why it's wrong here
A fixed seed makes generation reproducible, so repeated runs produce near-identical outputs and fail to broaden coverage. It is tempting because seeding is genuinely useful for debugging and benchmarking, where consistent, repeatable results matter more than diversity.
✓
Increase the temperature
Why this is correct
Higher temperature flattens the softmax probability distribution, giving lower-probability tokens greater chance of selection. This directly satisfies the diversity and edge-case coverage requirement, since synthetic outputs vary more instead of converging on the single most likely continuation.

Go deeper

AIF-C01 Temperature Practice Question

Why each option matters

Option-by-option breakdown

Related to this question