返回全部动态
在 Amazon SageMaker AI 上用多轮 RL 微调搜索智能体
原标题：Fine-tune a search agent with multi-turn RL on Amazon SageMaker AI
AI 摘要
AWS 在 Amazon SageMaker AI 上推出多轮强化学习（MTRL）能力，用于微调搜索智能体。文章以 Qwen3.6-27B 为例，结合 BM25 与向量检索工具，在 FRAMES、BRIGHT 和 Enterprise RAG 数据集上训练，使小模型获得接近前沿模型的多轮检索可靠性，同时保持更低延迟与成本。
以上摘要由 AI 生成，可能存在误差。事实请以原文为准。
正文节选
Fine-tune a search agent with multi-turn RL on Amazon SageMaker AI Search agents powered by large language models (LLMs) are transforming how enterprises retrieve information. Rather than requiring users to craft the perfect query, a search agent autonomously decides what to search for, which retrieval strategy to use, and when to stop searching. It does this across multiple rounds of interaction, refining its approach based on what it has already retrieved. However, getting this multi-step beha
发布时间：2026-10-02 23:44
抓取时间：2026-10-03 00:41
来源机构：AWS