GPT-4 学习利用人类反馈进行总结

We’ve applied reinforcement learning from human feedback to train language models that are better at summarization.