Prompt Strategy Classifier

Problem Context

You are building an AI assistant configuration system. The system needs to automatically select the most appropriate prompting strategy for a given task based on specific characteristics. The three strategies available are:

• Zero-Shot: The model answers with no examples provided. Use when the task is straightforward, well-defined, and the model is likely to have strong prior knowledge.
• Few-Shot: The model is given k example input-output pairs before the actual query. Use when the task involves a specific format, pattern recognition, or domain-specific output style.
• Chain-of-Thought (CoT): The model is guided to reason step by step before answering. Use when the task requires multi-step reasoning, math, logic, or complex decision-making.

Task

Given a list of task descriptors, classify each task into one of the three prompting strategies: "zero-shot", "few-shot", or "chain-of-thought".

Each task is represented as a list of integer feature flags:

• features[0]: 1 if the task requires multi-step reasoning, else 0
• features[1]: 1 if the task involves a specific output format or style, else 0
• features[2]: 1 if the task is a well-known, general-knowledge query, else 0
• features[3]: 1 if the task involves mathematical or logical computation, else 0

Classification Rules (in priority order):

1. If features[0] == 1 OR features[3] == 1 → "chain-of-thought"
2. Else if features[1] == 1 → "few-shot"
3. Else → "zero-shot"

Input Format

• tasks: a list of n tasks, where each task is a list of exactly 4 integers (each 0 or 1)

Output Format

Return a list of n strings, where each string is the recommended prompting strategy for the corresponding task.

Examples

Example 1

Input: tasks = [[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0]] Output: ["chain-of-thought", "few-shot", "zero-shot"] Explanation:

• Task 1: features[0] == 1 → multi-step reasoning required → "chain-of-thought"
• Task 2: features[1] == 1 → specific format needed → "few-shot"
• Task 3: , no other flags → general knowledge →