How to Build a Zero-Shot Classification Prompt in Python

Creates a prompt for zero-shot text classification by pairing input text with candidate labels and a hypothesis template.

Easy Python 3.9+ Aug 9, 2026 AI & LLM integration patterns 13 views 0 copies

zero-shot prompt classification nlp llm

Python code

21 lines

Python 3.9+

from typing import Dict, List


def build_zero_shot_prompt(
    text: str,
    candidate_labels: List[str],
    hypothesis_template: str = "This is about {}.",
) -> Dict[str, List[str]]:
    """Build a prompt ready for zero-shot classification."""
    return {
        "sequences": text,
        "candidate_labels": candidate_labels,
        "hypothesis_template": hypothesis_template,
    }


if __name__ == "__main__":
    sample_text = "A young child learns how to ride a bicycle in the park."
    labels = ["sports", "education", "nature"]
    prompt = build_zero_shot_prompt(sample_text, labels)
    print(prompt)

Output

stdout

{'sequences': 'A young child learns how to ride a bicycle in the park.', 'candidate_labels': ['sports', 'education', 'nature'], 'hypothesis_template': 'This is about {}.'}

How it works

The function bundles text, candidate labels, and a hypothesis template into a dictionary that mimics Hugging Face's zero-shot pipeline input format. The hypothesis template with {} lets the model form a natural-language premise like 'This is about sports.' Keeping the template constant for all labels ensures fair comparisons. The return type hints clarify the expected shape for callers and static checkers.

Common mistakes

Using a hypothesis template without the `{}` placeholder
Passing labels as a single string instead of a list
Forgetting to import Dict and List when using type hints

Variations

Use a dataclass to represent the prompt instead of a plain dictionary
Add a `multi_class` flag to control mutually exclusive vs independent classification

Real-world use cases

Classifying incoming support tickets into predefined categories before routing to a team.
Filtering user-generated content into topics for moderation or recommendation engines.
Pre-labeling dataset examples before sending them to a human-in-the-loop review.

How to Build a Zero-Shot Classification Prompt in Python

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from AI & LLM integration patterns

Tutorials

Quizzes

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from AI & LLM integration patterns

Keep learning

Tutorials

Quizzes