Researchers find language models do a poor job of following basic instructions

Researchers investigating whether a machine learning language model could understand basic sets of instructions found it performed poorly.