Modern neural language models (LMs) are powerful tools for modeling human
sentence production and comprehension, and their internal representations are
remarkably well-aligned with representations of language in the human brain.
But to achieve these results, LMs must be trained in dist