FrugalRAG: Learning to retrieve and reason for multi-hop QA

2025-07-10

AI Summary

To address the high retrieval cost and inefficient reasoning in multi-hop question answering over large-scale unstructured documents, this paper proposes a lightweight, retrieval-frugal framework. Building upon the standard ReAct paradigm, it replaces large-scale fine-tuning with meticulous prompt engineering and integrates minimal supervised signals (only 1,000 samples) with reinforcement learning to dynamically optimize retrieval paths. The core contribution is a principled trade-off between “fewer retrievals” and “higher accuracy”: on benchmarks such as HotPotQA, our approach reduces retrieval calls by 47% while achieving accuracy competitive with state-of-the-art methods—all using the same base model. Crucially, it incurs negligible training overhead, requires no domain-specific fine-tuning, and significantly improves the efficiency–accuracy balance of RAG systems.

Technology Category

• Search and Optimization: Learning to Search
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Natural Language Processing: Question Answering

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

We consider the problem of answering complex questions, given access to a large unstructured document corpus. The de facto approach to solving the problem is to leverage language models that (iteratively) retrieve and reason through the retrieved documents, until the model has sufficient information to generate an answer. Attempts at improving this approach focus on retrieval-augmented generation (RAG) metrics such as accuracy and recall and can be categorized into two types: (a) fine-tuning on large question answering (QA) datasets augmented with chain-of-thought traces, and (b) leveraging RL-based fine-tuning techniques that rely on question-document relevance signals. However, efficiency in the number of retrieval searches is an equally important metric, which has received less attention. In this work, we show that: (1) Large-scale fine-tuning is not needed to improve RAG metrics, contrary to popular claims in recent literature. Specifically, a standard ReAct pipeline with improved prompts can outperform state-of-the-art methods on benchmarks such as HotPotQA. (2) Supervised and RL-based fine-tuning can help RAG from the perspective of frugality, i.e., the latency due to number of searches at inference time. For example, we show that we can achieve competitive RAG metrics at nearly half the cost (in terms of number of searches) on popular RAG benchmarks, using the same base model, and at a small training cost (1000 examples).

Problem

• Improving retrieval-augmented generation for multi-hop QA
• Reducing retrieval searches while maintaining accuracy
• Enhancing efficiency and frugality in RAG systems

Innovation

• Improved prompts enhance ReAct pipeline performance
• Supervised fine-tuning reduces retrieval search costs
• RL-based fine-tuning optimizes frugality in RAG