在 Amazon Bedrock AgentCore Evaluations 中使用决策模型
原标题：Use decision models with Amazon Bedrock AgentCore Evaluations
AI 摘要
AWS 博客介绍如何在 Amazon Bedrock AgentCore Evaluations 中通过代码型评估器接入决策模型（decision model），用于对智能体进行评分。文章对比了决策模型与 LLM-as-a-judge 的差异：决策模型针对固定选项返回校准后的概率分布，适合窄而可重复的判断；LLM 评估器则提供解释与深度分析。示例将 TypeSafe 的托管决策模型 Jev 和开源的 Strands Decider 2B 接入 AgentCore，构建了一个 HR 助手智能体的端到端评估流程。
正文节选
Use decision models with Amazon Bedrock AgentCore Evaluations Some of the questions you ask when evaluating an agent are narrow and repeatable. Did this response stay on policy? Is this tool call safe to approve? Does this answer match the retrieved context? Decision models answer exactly that kind of question. They take a structured state and a fixed set of allowed answers and return a calibrated probability over them, rather than free-form text whose confidence drifts from run to run. That mak