LLMs are few shot learners

LLMs are few shot learners 4-lms_are_few_shot_learners.pdf , 6-Language_Models_are_Few-Shot_Learners.pdf

• I skim through the paper. I especially get the idea of how gpt3 is trained with x-shot settings.

• Figure 2.1 : How gpt3 is trained with zero-shot, one-shot, few-shot settings:

  • 
  • Given:
    • Task description
    • One or more examples ( … → …)
    • Prompt (… → ?)