Input files (the materials each task hands to the agent at run start) for the
Agents Last Exam (ALE) benchmark. Browsable per-task directory layout.
Task Card Metadata
One row per task: titles, prompts, taxonomy, input-file descriptors
Open
Task Input Data
The input/ files each task hands the agent at… See the full description on the dataset page:
https://huggingface.co/datasets/agents-last-exam/agents-last-exam-data.