Skip to content

Training Language Models to Follow Instructions with Human Feedback

Authors: Ouyang et al. Year: 2022 ArXiv/Link: https://arxiv.org/abs/2203.02155

Summary

InstructGPT scaled RLHF to large language models, creating instruction-following models that better align with human intent.

Key Concepts

  • RLHF at scale
  • Instruction following
  • Reward models
  • Policy optimization
  • Alignment

Impact

Enabled practical alignment of large language models

Category

Alignment & RLHF