Loading…
Glossary term
Giving a model a handful of examples in the prompt so it can perform a task.
Few-shot prompting puts a handful of worked examples into the prompt before the real question. The model infers the pattern from them and applies it, with no training involved. It is a form of in-context learning.
A few-shot prompt
Classify the sentiment.
"I loved it" -> positive
"Terrible service" -> negative
"It was fine" -> neutral
"Absolutely fantastic" ->A few good examples can raise accuracy sharply and cost far less than fine-tuning.