Abstract:Retrieval-Augmented Generation (RAG) has emerged as a reliable external knowledge augmentation technique to mitigate hallucination issues and parameterized knowledge limitations in Large Language Models (LLMs). Existing Adaptive RAG (ARAG) systems struggle to effectively explore multiple retrieval sources due to their inability to select the right source at the right time. To address this, we propose a multi-source ARAG framework, termed MSPR, which synergizes reasoning and preference-driven retrieval to adaptive decide "when and what to retrieve" and "which retrieval source to use". To better adapt to retrieval sources of differing characteristics, we also employ retrieval action adjustment and answer feedback strategy. They enable our framework to fully explore the high-quality primary source while supplementing it with secondary sources at the right time. Extensive and multi-dimensional experiments conducted on three datasets demonstrate the superiority and effectiveness of MSPR.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that the existing Adaptive Retrieval - Augmented Generation (ARAG) systems are unable to effectively select appropriate information sources in multi - source retrieval, resulting in the inability to fully utilize high - quality information sources. Specifically: 1. **Challenges in multi - source retrieval**: Existing ARAG systems have difficulty effectively exploring multiple retrieval sources because they cannot select the right source at the right time. This leads to insufficient utilization of high - quality information sources and poor performance in complex tasks such as multi - hop question answering. 2. **Differences in characteristics of different information sources**: Different retrieval sources have significant differences in content quality, information scale, and knowledge density. For example, a high - quality but small - scale local corpus and a large - scale but lower - quality web browser. Directly introducing multi - source retrieval with these different characteristics may lead to blind selection of retrieval sources, hindering the priority and in - depth exploration of high - quality information sources. 3. **Limitations in the adaptive information collection process**: Even when basic retrieval preference descriptions are provided, existing methods are still difficult to select the optimal retrieval source and the appropriate timing during the adaptive process. To solve these problems, the paper proposes a multi - source ARAG framework (MSPR), which restricts the adaptive knowledge collection process by combining inference and preference - driven retrieval strategies, thereby achieving more effective multi - source retrieval and knowledge enhancement.

Towards Multi-Source Retrieval-Augmented Generation via Synergizing Reasoning and Preference-Driven Retrieval

A Multi-Source Retrieval Question Answering Framework Based on RAG

Multi-Source Knowledge Pruning for Retrieval-Augmented Generation: A Benchmark and Empirical Study

Understand What LLM Needs: Dual Preference Alignment for Retrieval-Augmented Generation

Retrieval-Augmented Generation for Large Language Models: A Survey

Searching for Best Practices in Retrieval-Augmented Generation

Retriever-and-Memory: Towards Adaptive Note-Enhanced Retrieval-Augmented Generation

R^2AG: Incorporating Retrieval Information into Retrieval Augmented Generation

CORAG: A Cost-Constrained Retrieval Optimization System for Retrieval-Augmented Generation

RichRAG: Crafting Rich Responses for Multi-faceted Queries in Retrieval-Augmented Generation

Speculative RAG: Enhancing Retrieval Augmented Generation through Drafting

Retrieval-Augmented Generation with Estimation of Source Reliability

IM-RAG: Multi-Round Retrieval-Augmented Generation Through Learning Inner Monologues

Retrieval-Augmented Generation for Natural Language Processing: A Survey

Corrective Retrieval Augmented Generation

Self-adaptive Multimodal Retrieval-Augmented Generation

Plan$\times$RAG: Planning-guided Retrieval Augmented Generation

Similarity is Not All You Need: Endowing Retrieval Augmented Generation with Multi Layered Thoughts

MemoRAG: Moving towards Next-Gen RAG Via Memory-Inspired Knowledge Discovery

LongRAG: A Dual-Perspective Retrieval-Augmented Generation Paradigm for Long-Context Question Answering