Few-Shot Prompting Task

From GM-RKB
(Redirected from Few-shot prompting)
Jump to navigation Jump to search

A Few-Shot Prompting Task is a NLP task that require an input of exemplar prompts.



Referneces

2023

  • chat
    • Few-shot prompting is a technique used in the context of machine learning, particularly with large-scale language models like GPT-3 and GPT-4. It involves providing a model with a few examples or instances (called "prompts") to demonstrate the desired task or format of the response. The model then generalizes from these examples to perform the task with new input data.

      The idea behind few-shot prompting is that these models have already learned a lot of knowledge during their pre-training, and providing a few examples helps them understand the user's intent and adapt to a wide range of tasks without the need for extensive fine-tuning.

2022