What is Zero-shot?
It is the ability of artificial intelligence to perform a task it has never seen an example of, using only a description.
Overview
Zero-shot is when an artificial intelligence model generates an answer by understanding only the given instruction on a subject for which it has not been previously trained or specifically sampled. The model uses its general knowledge to provide a logical solution to a question or task it has never encountered before. This demonstrates the model's capacity to understand concepts rather than just memorizing them.
How it works
By using the model's general capabilities, you define a complex task with a clear prompt. If the model is proficient in the subject, it produces the correct result directly without needing extra training or examples (few-shot).
Where it is used
It is used in LLM-based applications, rapid prototyping processes, and systems that respond to users' instant questions.
Commonly confused with
It can be confused with few-shot learning; while you give the model a few examples in few-shot, you do not give any examples in zero-shot.
Frequently asked questions
Does it always work correctly?
No, if the model does not have basic knowledge about that subject in its training data, it may make mistakes.
Why is it called 'zero-shot'?
It means 'zero shot' in English; that is, the model has not been given any trial attempts or examples to hit the target.
Related terms
Related tools
This explanation was written in plain language for TreScout and machine-translated from the Turkish original · the Turkish version prevails. If something looks wrong or missing, write to hello@trescout.com. Read in Turkish →