Zero-Shot Prompting
Definition pending
We've cataloged "Zero-Shot Prompting" but haven't written a full definition yet. Definitions are hand-curated rather than auto-generated, so it takes time to cover every term.
Want this one prioritized? Email us and we'll bump it.
Practical example
Zero-shot means asking the model to do a task with no examples — just the instruction. "Classify this email as spam or not spam: [email text]." Zero-shot is the default for most production LLM apps because it's cheapest (fewest prompt tokens). It works well for simple tasks (classification, extraction, summarization) but poorly for nuanced tasks (legal analysis, creative writing).
Workflow example
Default to zero-shot, escalate to few-shot only if needed: (1) build your prompt with zero-shot, (2) run on 100 test examples, measure accuracy, (3) if accuracy > 90%, ship it — zero-shot is sufficient, (4) if 70–90%, try 3-shot and re-measure, (5) if <70% even with few-shot, the task is too hard for this model — consider fine-tuning or a larger model. Log accuracy and shot count per endpoint.