Few-Shot and Zero-Shot Learning
Techniques where models perform tasks with few or no examples. Few-shot provides a handful of examples; zero-shot requires none, relying purely on instructions.
Few-shot and zero-shot learning represent a major shift from traditional ML, which required thousands of labeled examples. Large language models can often perform tasks with minimal or no training data.
Zero-Shot Learning: The model receives only a task description or instruction. Example: “Translate to French: Hello world” — the model has never seen this specific translation pair but understands the task from the instruction alone.
Few-Shot Learning: The prompt includes 1-5 examples before the actual query. These in-context examples guide the model’s output format and style. Few-shot prompting is a core technique in prompt engineering.
One-Shot Learning: A special case of few-shot where exactly one example is provided.
These techniques work because LLMs encode vast world knowledge during pre-training. They’re most effective with larger models and are central to modern AI applications where collecting training data is impractical.