Task quality reports

5 tasks · Zero-Trust Physics v0.15.3 · Protocol Company

TaskKindReference rewardCheats above 0Pass@1
g3-budget-grasp-least-grip-at-6.1-t4Budgeted search10Frontier model tool pass@1 62%
g3-chain-column-feeds-line-fastest-flow-t3Chained design10Frontier model tool pass@1 50%
g3-limit-chamber-highest-reservoir-t2Design to a limit10Frontier model tool pass@1 50%
g3-forensics8-14Forensics10Frontier model tool pass@1 38%
g3-budget-hammer-flow-to-wall-yield-t5Budgeted search10Frontier model tool pass@1 62%