Training language models to follow instructions with human feedback

Open paper

Benchmark	Model	Rank	Results
question-answering-on-timequestions	InstructGPT	#8	P@1: 22.4
question-answering-on-tiq	InstructGpt	#5	P@1: 23.6