직업 분류
소프트웨어 개발자
설명
Run a first-party reliability benchmark comparing two or more LLMs on hard, trap-laden tasks through one identical harness, graded by reading and scored pass^k. Use when the user wants to benchmark, compare, or A/B models for reliability (not raw capability),…
원문 언어: 영어
업데이트