AUC@K point estimate: average over questions of the trapezoidal area under the Pass@1..Pass@k curve. For k = 1, AUC@1 = Pass@1.
k = 1
References: Hu et al. (2026), arXiv:2601.08763.
M x N binary matrix with entries in {0, 1}.
M x N
Maximum sampling budget with 1 <= k <= N.
1 <= k <= N
The average AUC@K score across all M questions.
M
AUC@K point estimate: average over questions of the trapezoidal area under the Pass@1..Pass@k curve. For
k = 1, AUC@1 = Pass@1.References: Hu et al. (2026), arXiv:2601.08763.