Post

Perhaps a Langchain style sanitisation can be applied to remove all multi -, spaces etc. Can save a few tokens but more importantly improve inference