SESKStuff Everyone Should Know
SESK: Stuff Everyone Should Know · sesk.info/ai-few-shot-no-weight-updates

When you put a few examples in a prompt (few-shot prompting), what happens to the model's weights?

Nothing. The examples work purely through the text.

Start playing free

  1. They are fine-tuned on the examples
  2. They stay exactly the same ✔
  3. They are saved for future users

Why: The 2020 GPT-3 paper showed big models can do new tasks from a few examples in the prompt, with no gradient updates or fine-tuning. Start a new chat and the examples are gone.

Huh, didn't know that: A prompt with no examples is called zero-shot; with exactly one, it is one-shot.

The dad joke
Few-shot prompting is like showing my teenager how to load the dishwasher three times. Learned for the moment, gone tomorrow.
Like this card?
Play the game

References

Answer: They stay exactly the same. The 2020 GPT-3 paper showed big models can do new tasks from a few examples in the prompt, with no gradient updates or fine-tuning. Start a new chat and the examples are gone.

  1. Brown et al., Language Models are Few-Shot Learners (arXiv 2005.14165) arxiv.org/abs/2005.14165
    GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction
    Checked 2026-10-10.
  2. Google Machine Learning Glossary: Generative AI (few-shot prompting) developers.google.com/machine-learning/glossary/generative
    A prompt that contains more than one (a "few") example demonstrating how the large language model should respond.
    Checked 2026-10-10.

Think this answer is wrong? Tap "Challenge this answer" on the card in the game and tell us why.

Test yourself in the game

Artificial IntelligenceWhat is a common use of the system prompt in a chat model API?Artificial IntelligenceWhat is chain-of-thought prompting?Artificial IntelligenceHow do AI, machine learning and deep learning fit together?

That's stuff everyone should know.