What AI says about LLMMaps

LLMMaps is a visualization tool and pipeline for stratified evaluation of large language models, described in its accompanying paper. It provides processed datasets and scripts to reproduce the visualizations, including a workflow to convert raw QA data into visualization-ready formats and a tree_vis.py script that consumes data files (.qar, .qaa, .hir) in a data directory. The repository includes example workflows (e.g., ChatGPT on PubMedQA) and aims to help researchers compare LLMs across stratified dimensions by visualizing model behavior.