Task Catalog

This page is the searchable public catalog of benchmark tasks.

Public Catalog

The current view is generated from the pinned 60-task ASI-Bench seed42 benchmark set. It includes only final public task definitions in that set and supports search and domain filtering.

Data Source

Task membership is fixed by a checked-in benchmark-set manifest. Public metadata is synchronized from apexin-ai/ASI-Bench, and each detail page identifies the exact public source commit and links only to tracked public files.