Question 37
Domain 1: Agent Architecture, Design, and DevelopmentAn agent is asked to answer the question: “If Tom has 3 apples and gives 1 to Sarah, how many does he have left?”. To ensure accurate reasoning, which technique would most effectively improve the agent’s response?
Correct answer: B
Explanation
Task decomposition improves reasoning by breaking a problem into smaller steps, such as identifying the starting amount, the transfer, and the remainder. For “Tom has 3 apples and gives 1 to Sarah,” the agent can compute 3 - 1 = 2, which reduces errors in multi-step thinking.
Why each option is right or wrong
A. Zero-shot Learning
B. Task Decomposition
Task decomposition is the best fit here because the agent must separate the problem into discrete substeps: identify the initial quantity (3), account for the transfer away (1), and then compute the remainder. In a simple subtraction word problem like this, breaking the prompt into those steps directly supports the correct arithmetic result of 2 apples and reduces the chance of misreading the question.
C. Random Sampling
D. Batch Normalization