Shot learning

As explained in Grokking AI Applications

A prompting approach that provides zero or more examples to guide the output behavior of a language model.

Shot learning guides a language model by showing examples of the output pattern you want, rather than fine-tuning its behavior. Zero-shot prompting gives instructions without examples; one-shot gives one example; few-shot gives several. section 3.1

Choose the approach according to how much guidance the task needs:

• Zero-shot: describe the task and the expected features without an example. section 3.1
• One-shot: include one desired response, such as a sample concise summary or title. section 3.1
• Few-shot: provide multiple responses to demonstrate variation and formatting nuances. section 3.1

Prompting a title or classification pattern

A prompt might say, “A good title would be something like ‘X’,” or show that input “X” should produce output “Y” rather than “Z.” These examples make the intended answer type easier to emulate. section 3.1

Figure 3.3 The more examples you give the LLM, the easier the climb.

The figure reinforces how adding examples can make the desired output easier for an LLM to follow. section 3.1

How Grokking AI Applications covers it

How deep it goes

Distinguishes zero-, one-, and few-shot prompting but stays at prompt design rather than model internals.

Examples

Uses sample titles, output formats, alternatives, and good-versus-bad response pairs.

What it assumes

Assumes the ability to describe a task and its expected features, while teaching example-based prompting from basic distinctions.

What you can do after

Apply via prompt design by selecting zero, one, or several examples and observing their effect on output.

How it connects

Connects prompting with language models, fine-tuning, and GenAI applications.

liveBookmanning.com

Read Grokking AI Applications on liveBook Get the book at manning.com

Related concepts

• zero-shot promptingincludes · Zero-shot prompting asks a model to perform a task through instructions or a task description without providing task-specific examples or demonstrations.
• few-shot learningincludes · Few-shot learning is the adaptation or performance of a model on a task using only a small number of task-specific examples.
• one-shot learningincludes · One-shot learning is a machine-learning setting in which a model generalizes to recognize or perform a task from a single labeled or demonstrated example.

This explainer was generated by Manning's knowledge graph from the text of Grokking AI Applications. The section links point to the parts of the book it draws on.