oro-agents/smb-bench
Viewer • Updated • 666 • 115
SMB Bench
The economic value of agents will soon come from running whole businesses, not from finishing tasks.
Note Per-run results, statistics, recompute script and tool-call traces for every run (CC BY 4.0).
oro-agents/smb-bench-grader
Updated • 42Note Access-requested: grade-only regrade bundle, worlds and run evidence.
SMB Bench
📈Can AI learn from delayed consequences?
Note Report figures, leaderboard and a browser over every run.