Published onJanuary 27, 2022Aligning language models to follow instructionsGPT-3.5openairesearchGPT-3.5, trained language models that are much better at following user intentions than GPT-3 while also making them more truthful and less toxic.