Towards AI Search Paradigm

Abstract

In this paper, we introduce the AI Search Paradigm, a comprehensive blueprint for next-generation search systems capable of emulating human information processing and decision-making. The paradigm employs a modular architecture of four LLM-powered agents (Master, Planner, Executor and Writer) that dynamically adapt to the full spectrum of information needs, from simple factual queries to complex multi-stage reasoning tasks. These agents collaborate dynamically through coordinated workflows to evaluate query complexity, decompose problems into executable plans, and orchestrate tool usage, task execution, and content synthesis. We systematically present key methodologies for realizing this paradigm, including task planning and tool integration, execution strategies, aligned and robust retrieval-augmented generation, and efficient LLM inference, spanning both algorithmic techniques and infrastructure-level optimizations. By providing an in-depth guide to these foundational components, this work aims to inform the development of trustworthy, adaptive, and scalable AI search systems.

Figures (15)

• Figure 1: The Overview of AI Search Paradigm.
• Figure 2: Comparison of RAG frameworks. Left: Vanilla RAG conducts one-shot retrieval followed by direct answer generation. Middle: Advanced RAG methods, such as ReAct and RQ-RAG, involve reasoning-action cycles or sequential sub-query execution. Right: AI search paradigm introduces a multi-agent system wherein Master guides Planner to formulate a plan based on the input query, while also continuously evaluating the execution status and completeness of sub-task results, and performing reflection and re-planning when necessary. Planner is responsible for constructing a DAG of sub-tasks and dynamically selecting the appropriate tools, thereby enabling structured and adaptive multi-step execution. Executor executes the specific sub-tasks using these tools, and finally, Writer generates the final answer.
• Figure 3: An illustration of the dynamic capability boundary of our search paradigm.
• Figure 4: The pipeline of our proposed self-driven iterative refinement framework, which leverages the interactions between LLMs and external tools, as well as the feedback generated during these interactions, to progressively optimize tool documents
• Figure 5: An illustration of the paradigm shift for the task Executor. The left panel shows the traditional task executor based on PLMs for web search, which primarily addresses a single query by transforming the web search task into a ranking problem and aligning the results with user preferences. The right panel illustrates the paradigm shift in the task execution for AI search. Our AI search system processes multiple queries by re-framing the traditional ranking problem as a recall problem, shifting the objective from aligning user preferences to aligning LLM preferences.