職業分類
ソフトウェア開発者
説明
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),…
原文の言語: 英語
更新