# Towards AI Search Paradigm

**Authors:** Yuchen Li, Hengyi Cai, Rui Kong, Xinran Chen, Jiamin Chen, Jun Yang, Haojie Zhang, Jiayi Li, Jiayi Wu, Yiqun Chen, Changle Qu, Keyi Kong, Wenwen Ye, Lixin Su, Xinyu Ma, Long Xia, Daiting Shi, Jiashu Zhao, Haoyi Xiong, Shuaiqiang Wang  
**arXiv:** [2506.17188](https://arxiv.org/abs/2506.17188)  
**Date:** 2025-06-20  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2506.17188v1/

## Key points

- Authors introduce a comprehensive blueprint for next-generation search systems
- Proposed AI Search Paradigm utilizes a modular architecture with four LLM-powered agents: Master, Planner, Executor, and Writer
- Agents collaborate through coordinated workflows to assess query complexity, decompose problems into executable plans, and manage tool usage, task execution, and content synthesis
- Key methodologies presented include task planning and tool integration strategies aligned with retrieval-augmented generation techniques and efficient LLM inference methods
- Comprehensive approach spans algorithmic techniques and infrastructure-level optimizations to inform the development of trustworthy adaptive and scalable AI search systems

## Summary

In their paper "Towards AI Search Paradigm," authors Yuchen Li, Hengyi Cai, Rui Kong, Xinran Chen, Jiamin Chen, Jun Yang, Haojie Zhang, Jiayi Li, Jiayi Wu, Yiqun Chen, Changle Qu, Keyi Kong,
Wenwen Ye, Lixin Su, Xinyu Ma,
Long Xia,
Daiting Shi,
Jiashu Zhao,
Haoyi Xiong,
Shuaiqiang Wang and Dawei Yin introduce a comprehensive blueprint for next-generation search systems that aim to emulate human information processing and decision-making. The proposed AI Search Paradigm utilizes a modular architecture consisting of four LLM-powered agents - Master,
Planner,
Executor and Writer - which dynamically adapt to a wide range of information needs ranging from simple factual queries to complex multi-stage reasoning tasks. These agents collaborate through coordinated workflows to assess query complexity effectively decompose problems into executable plans and manage tool usage,
task execution and content synthesis. The authors systematically present key methodologies for realizing this paradigm including task planning and tool integration strategies aligned with robust retrieval-augmented generation techniques and efficient LLM inference methods. This comprehensive approach spans both algorithmic techniques and infrastructure-level optimizations with the goal of informing the development of trustworthy adaptive and scalable AI search systems.