This is the official implementation of WALM (Word Agreement with Language Model), as proposed in: LLM Reading Tea Leaves: Automatically Evaluating Topic Models with Large Language Models (paper link).