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

Paper: https://arxiv.org/abs/2410.20753 Authors: Prakhar Verma, Sukruta Prakash Midigeshi, Gaurav Sinha, Arno Solin, Nagarajan Natarajan, Amit Sharma Published: October 28, 2024 Category: Agentic Search

What is Plan*RAG?

Plan*RAG is 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, the authors observe that this often leads to plan fragmentation and execution failures.

Key Insight

By isolating the reasoning plan as a directed acyclic graph (DAG) outside the LM's working memory, Plan*RAG 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

Moreover, Plan*RAG's modular design allows it to be integrated with existing RAG methods, providing a practical solution to improve current RAG systems.

Results

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.

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.

Why It Matters

Plan*RAG represents the shift in the RAG and Agentic Search space from one-shot retrieval toward an iterative, plannable, and verifiable knowledge access process. Treating the reasoning plan as an explicit, structured artifact (a DAG) rather than implicit in-context state improves robustness for multi-hop question answering, while parallel execution keeps latency and cost in check.

Related Work

• RQ-RAG and Self-RAG (baselines compared against)
• ReAct (in-context reasoning chains approach)
• Related entries: Agentic Information Retrieval (arXiv:2410.09713), Synergizing RAG and Reasoning survey (arXiv:2504.15909), AceSearcher (arXiv:2509.24193)