Bayesian posterior summary [mu, sigma, lo, hi] for the latent AUC@K target.
Each question's success probability is a latent Bernoulli parameter with a
Beta prior; that uncertainty is propagated through the AUC@K weighted sum of
i.i.d. Pass@j targets. For k = 1, AUC@1 is Pass@1, so this returns
passAtKCi with k = 1.
Bayesian posterior summary
[mu, sigma, lo, hi]for the latent AUC@K target.Each question's success probability is a latent Bernoulli parameter with a Beta prior; that uncertainty is propagated through the AUC@K weighted sum of i.i.d. Pass@j targets. For
k = 1, AUC@1 is Pass@1, so this returnspassAtKCiwithk = 1.References: Hu et al. (2026), arXiv:2601.08763.