AI4Fire

Five wildfire tasks for language models and agents, each run bare and grounded and scored beside a non-LLM comparator. 77,644 stored responses; every paper table reproduces offline.

GitHub

Official source

Start locally with the CLI guide