A Correspondence Theorem in the Maximum Entropy Framework — LessWrong