LLM 研究员
我要你充当大型语言模型(Large Language Model)研究领域的专家。请仔细阅读用户提供的论文、文本或概念性术语,然后回答他们提出的问题。在作答时,请确保不遗漏任何重要细节。基于你的理解,你还应说明该概念背后的原理、实现过程及目的。如有可能,你可利用网络搜索查找有关该概念或其推理过程的额外信息。在呈现信息时,凡有可用的论文引用或链接,均须一并提供。
Persistent Retrieval for Conversational Agents: A Study of Retrieval Efficiency
Karthik Ravula · Sathwik Kowtha · Rishabh Verma · Kala Mehak Jain · Gunjit Mittal
对话式智能体可以直接检索信息,也可以将检索委托给子智能体。这一设计选择影响检索历史的维护方式、每次模型调用接收的上下文,以及后续问题中需要重新检索的信息。比较三种设计:主智能体中基于技能的检索、生成发现响应的非持久化检索子智能体,以及本文的持久化检索子智能体——跨轮次维护检索历史并将检索到的证据直接返回主智能体,还允许主智能体向同一子智能体请求修正。在企业问题(独立提问、共享对话内提问、依赖式后续提问)上评估三种设计。在五种多轮设置中,相比基于技能的检索,记录到的智能体-模型 token 少 43.3%,估计 API 成本低 45.0%,平均响应时间缩短 9.1%;相比非持久化检索,每题响应时间减少 15.6 秒,但 token 多 65.1%,估计 API 成本高 29.1%。该方法使检索子智能体从生成综合响应转向维护检索状态并提供证据,供主智能体直接评估并用于答案生成。
尚无解读。
arXiv:2611.07899v1 Announce Type: cross Abstract: Conversational agents can retrieve information directly or delegate retrieval to a subagent. This design choice affects how retrieval history is maintained, what context each model call receives, and what information must be re-retrieved in follow-up questions. We compare three designs: skill-based retrieval in the main agent, a non-persistent retrieval subagent that generates a findings response, and our persistent retrieval subagent, which maintains retrieval history across turns and returns retrieved evidence directly to the main agent. Our approach also allows the main agent to request corrections from the same subagent. We evaluate the three designs on enterprise questions posed independently, within shared conversations, and as dependent follow-ups. Across five multi-turn settings, our approach uses 43.3% fewer recorded agent-model tokens and incurs 45.0% lower estimated API cost than skill-based retrieval, while reducing average response time by 9.1%. Compared with non-persistent retrieval, it reduces response time by 15.6 seconds per question, while using 65.1% more tokens and incurring 29.1% higher estimated API cost. Our approach shifts the retrieval subagent from producing synthesized responses to maintaining retrieval state and supplying evidence that the main agent can directly evaluate and use for answer generation.