ASI-Bench submission portal
Contribute a scientific task or upload a benchmark run bundle for the ASI-Bench evaluation of general intelligence, innovation, and autonomous execution. Both flows preserve provenance, pass structured checks, and stay traceable through official review.
Task contribution
A clear scientific question and expected output.
Ground truth, scorer thresholds, and runtime needs.
One same-model B1-B4 evaluation with timings and reproducibility evidence.
Submission workflows
Open the Guided Flow. If your Task already exists locally, CLI upload is explained before you begin.
Start nowRun the public framework, package a checksummed bundle, and confirm it for official scoring.
OpenCompare reviewed benchmark runs on the public leaderboard.
Open