Your Agent Aced the Task. Will It Do It Again?

Selection record

AdmittedSum of both 129 ≥ twice the threshold 120

Source tier

Official, first-hand; this tier's threshold is 60

Pre-filter

passed：AI Agent一致性与评测方法技术文章

Why it was chosen

Quantified Pass^k and consistency-gap results, together with diagnosis and improvement methods, provide grounds to reconsider agent evaluation criteria.

A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.