「completion-gate」の検索結果

Independent LLM judge evaluates task completion separately from the executing agent, catching false success claims by reviewing task goal, actions taken, final state, and evidence. Produces PASS/FAIL with confidence score and reasoning.

日本語の概要は準備中です。原文の説明を表示しています。