Zero-Shot, One-Shot, Few-Shot: How Many Examples a Task Really Needs
There are three ways to hand a task to a frozen model: describe it (zero-shot), show one worked example (one-shot), or show a handful (few-shot). This topic covers the GPT-3 origin of the idea, the surprising research on what examples actually teach (format as much as labels), the sensitivities that quietly break production prompts (order, label balance, selection), and a 2024–2026 decision playbook.
01.The Problem: Describe the Game, or Just Play It?
You have a frozen model (weights untouchable — the setup from Topic 157) and a task: classify support tickets as billing, bug, or praise.
How do you teach it? You have exactly three moves:
- Describe it in words.
- Show one worked example, then ask.
- Show several worked examples, then ask.
Which move works — and how many examples are actually worth their token cost?
This topic answers with numbers: from the model's own history (GPT-3 introduced "few-shot prompting" in 2020), from the research on what examples really teach (surprise: mostly format), and from the failure modes that silently wreck production classifiers (order, label balance, cost).
The one-line intuition to keep: words describe the game; demonstrations play it. The question is how many rounds you need before the model gets it.
Shot Spectrum and Its Sharp Edges 🎯
Shot Spectrum and Its Sharp Edges 🎯
Adding demonstrations is the most reliable zero-retraining capability lever — but examples influence through format and position as much as through their labels.
Unlock Topic #158: Zero-Shot, One-Shot, Few-Shot: How Many Examples a Task Really Needs
You are viewing a preview. The full in-depth technical walkthrough, worked derivations, and code notebooks for this concept, along with self-assessment quizzes, are available with Pro or Lifetime Access.
Failure modes, high-throughput bottlenecks, and real FAANG implementation decisions.
Interactive system topology diagrams, live parameter simulators, and downloadable SVG charts.
Staff-level multiple-choice quiz questions with instant feedback and answer explanations.
Firebase Google authentication automatically syncs your completed topics and quiz scores.
How clear and actionable was this distributed systems breakdown?