Techniques to Improve Generative AI Model Output →hardMultiple Choice

Generative AI Leader Practice Question: Techniques to Improve Generative AI Model Output

Exhibit

Refer to the exhibit.

```
# Model configuration before change
model = GenerativeModel("text-bison@002")
response = model.generate(
    prompt="Summarize the following article: ...",
    temperature=0.7,
    top_k=40,
    top_p=0.95
)
# After change
model = GenerativeModel("text-bison@002")
response = model.generate(
    prompt="Summarize the following article: ...",
    temperature=0.2,
    top_k=10,
    top_p=0.85
)
```

Refer to the exhibit. The team changed the generation parameters to reduce output variability. However, summaries now often repeat the same phrases. Which parameter change is most likely causing the repetition?

⚠ Common exam trap

Generative AI Leader often tests the confusion between temperature (sharpens distribution, can cause repetition at low values) and top_p/top_k (truncate candidates but preserve randomness), causing candidates to blame the wrong parameter.

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

✓

Reducing temperature from 0.7 to 0.2

Reducing temperature from 0.7 to 0.2 makes the model's token selection much more deterministic, favoring the highest-probability tokens. This low-entropy sampling often causes the model to repeat the same phrases across summaries because it consistently picks the most likely continuation.

Answer analysis

Option-by-option breakdown

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

✗
Reducing top_p from 0.95 to 0.85
Why it's wrong here
Lower top_p also reduces diversity but less impactful than temperature.
✓
Reducing temperature from 0.7 to 0.2
Why this is correct
Low temperature increases determinism and repetition.
✗
Using the same model text-bison@002
Why it's wrong here
Keeping the same model text-bison@002 changes nothing about decoding; the model version is fixed and identical before and after the parameter edit, so it cannot introduce repetition. Pinning a model version is correct when you need reproducible, stable behaviour across deployments.
✗
Reducing top_k from 40 to 10
Why it's wrong here
Reducing top_k from 40 to 10 restricts sampling to the ten highest-probability tokens, which can cause looping, but top_k is a hard cutoff that still samples randomly among those ten, so repetition is less pronounced than with greedy decoding. Top_k is correctly used to bound the candidate pool.

Go deeper

Generative AI Leader Practice Question: Techniques to Improve Generative AI Model Output

Exhibit

Why each option matters

Option-by-option breakdown

Related to this question