hardMultiple Choice
AIF-C01 Practice Question: An AI practitioner is fine-tuning an Amazon Titan…
An AI practitioner is fine-tuning an Amazon Titan Text model on a dataset of customer support conversations to improve response accuracy. After training, the model's perplexity on the validation set is low, but during inference, the model frequently generates off-topic or nonsensical responses to real customer queries. What is the most likely cause?
⚠ Common exam trap
The AIF-C01 exam often tests the distinction between overfitting (Option C) and distribution mismatch (Option D), where candidates mistakenly attribute low validation perplexity with good generalization, but the trap is that overfitting would still show high perplexity on a representative validation set, whereas here the validation set itself is the problem.
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
✓
The validation set does not represent the distribution of real customer queries
The core issue is a mismatch between the validation set and real-world inference data. Low perplexity on the validation set indicates the model fits that specific distribution well, but if the validation set does not reflect the diversity, phrasing, or intent of actual customer queries, the model will generate off-topic or nonsensical responses during inference. This is a classic case of distribution shift, where the model has not generalized to the true target domain.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
The model's context window is too small for the inference queries
Why it's wrong here
A small context window truncates long inputs, which would cut off prompt content, but it cannot make a model ignore its fine-tuning and emit off-topic text on short queries. Context length matters when prompts exceed the token limit; here the symptom is a generalisation failure, not truncation.
- ✗
The temperature during inference is set too low
Why it's wrong here
Low temperature sharpens the probability distribution, making output more deterministic and repetitive, so it cannot produce the incoherent, off-topic responses described. Raising temperature is the lever for creative variation; the reported symptom instead indicates the model memorised the training set rather than generalising.
- ✗
The fine-tuning dataset is too small, causing overfitting
Why it's wrong here
A small dataset would inflate validation perplexity, not lower it; low perplexity here points to train/validation leakage or distribution mismatch, not overfitting. Dataset size is tempting because overfitting is the usual suspect, yet overfitting raises validation loss, so it cannot explain the observed pattern.
- ✓
The validation set does not represent the distribution of real customer queries
Why this is correct
Low validation perplexity with nonsensical live responses indicates the validation set shares the training distribution but not the real query distribution, so the model never learned the actual input patterns. The gap is distributional mismatch, not overfitting to the validation set itself.
Go deeper
Related to this question
About these practice questions
Courseiva writes every AIF-C01 question from scratch — 862 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. Learn why practice questions differ from exam dumps →
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This AIF-C01 practice question is part of Courseiva's free Amazon Web Services certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AIF-C01 exam.