Request access to ALE reference (ground-truth) data
This dataset contains the ground-truth / reference outputs used to score the Agents Last Exam (ALE) benchmark. Access is gated to protect benchmark integrity. Approval is granted manually.
By requesting access you agree to ALL of the following. Misuse may result in access revocation and exclusion from the benchmark leaderboard.
Log in or Sign Up to review the conditions and access this dataset content.
Agents Last Exam v1.1 — Reference Data
150 active tasks. Use matching v1.1 task code, environment images, and data revisions.
Evaluator references are under tasks/<task_id>/<variant>/reference/. Access requires manual approval. Do not use references for model training, expose them to an agent during evaluation, or redistribute them. Stage references only after the solver stops.
Restore native filenames at the snapshot root with tar -xf _transport/v1.1/path-compatibility.tar -C ..
Logical data manifest SHA256: 05c8975a5c68309d724a3bdbefa7ef1675bada4f4bde809fbacf113d25bb24bf.
- Downloads last month
- 17,346