#evaluation

Hugging Face's new benchmark judges how AI agents work, not just whether they finish

A new evaluation tool from Hugging Face scores AI agents on the path they take through a task, catching failure modes like looping and tool misuse that standard leaderboards miss.