Plan*RAG: Efficient Test-Time Planning for Retrieval Augmented Generation

Overview

**Plan*RAG is a research paper (arXiv, October 28, 2024) by Prakhar Verma, Sukruta Prakash Midigeshi, Gaurav Sinha, Arno Solin, Nagarajan Natarajan, and Amit Sharma. It introduces a framework for structured multi-hop reasoning in retrieval-augmented generation (RAG) through test-time reasoning plan generation.

• Paper: https://arxiv.org/abs/2410.20753
• Category**: Agentic Search / RAG

Key Idea

Existing agentic RAG approaches such as ReAct maintain reasoning chains within the language model's context window, which often leads to plan fragmentation and execution failures. Plan*RAG's key insight is to isolate the reasoning plan as a directed acyclic graph (DAG) outside the LM's working memory. This enables:

1. Systematic exploration of reasoning paths; 2. Atomic subqueries enabling precise retrievals and grounding; 3. Efficiency through parallel execution and bounded context window utilization.

Results

Plan*RAG's modular design allows it to be integrated with existing RAG methods. On standard multi-hop reasoning benchmarks, it consistently achieves improvements over recently proposed methods such as RQ-RAG and Self-RAG, while maintaining comparable computational costs.

Original Abstract

> We introduce Plan*RAG, a novel framework that enables structured multi-hop reasoning in retrieval-augmented generation (RAG) through test-time reasoning plan generation. While existing approaches such as ReAct maintain reasoning chains within the language model's context window, we observe that this often leads to plan fragmentation and execution failures. Our key insight is that by isolating the reasoning plan as a directed acyclic graph (DAG) outside the LM's working memory, we can enable (1) systematic exploration of reasoning paths, (2) atomic subqueries enabling precise retrievals and grounding, and (3) efficiency through parallel execution and bounded context window utilization. Moreover, Plan*RAG's modular design allows it to be integrated with existing RAG methods, thus providing a practical solution to improve current RAG systems. On standard multi-hop reasoning benchmarks, Plan*RAG consistently achieves improvements over recently proposed methods such as RQ-RAG and Self-RAG, while maintaining comparable computational costs.

Context

Plan*RAG sits at the intersection of RAG and agentic search, addressing how retrieval, ranking, generation, and tool invocation should be orchestrated in the LLM era — in particular, how to allocate the *reasoning budget* and *action space* (whether to retrieve, how many times, and which tools to call). Related work includes agentic information retrieval, RAG-and-reasoning surveys, and reinforcement-learning-based search agents such as AceSearcher.