Scaling Retrieval-Augmented Reasoning with Parallel Search and Explicit Merging

2026-05-13

AI Summary

This work addresses the limitations of existing retrieval-augmented reasoning methods, which often generate only a single query per reasoning step, resulting in insufficient information coverage and low signal-to-noise ratios that hinder both accuracy and efficiency. To overcome these challenges, the authors propose MultiSearch, a novel framework that concurrently generates multi-perspective queries at each reasoning step to broaden information coverage and incorporates an explicit information fusion mechanism to distill retrieved content. Furthermore, the framework employs a multi-process reward–based reinforcement learning strategy to jointly optimize query generation and information integration. Experimental results demonstrate that MultiSearch significantly outperforms current baselines across seven benchmark datasets, effectively enhancing retrieval signal-to-noise ratios and improving reasoning performance on question-answering tasks.

Abstract

Deep search agents have proven effective in enhancing LLMs by retrieving external knowledge during multi-step reasoning. However, existing methods often generate a single query for retrieval at each reasoning step, limiting information coverage and introducing high noise. This may result in low signal-to-noise ratios (SNR) during search, degrading reasoning accuracy and leading to unnecessary reasoning steps. In this paper, we introduce MultiSearch, an RL-based framework that addresses these limitations through multi-query retrieval and explicit merging of retrieved information. At each reasoning step, MultiSearch generates queries from multiple perspectives and retrieves external information in parallel, expanding the scope of relevant information and mitigating the reliance on any single retrieval result. Then, the agent consolidates and refines retrieved information at the merging process, improving the SNR and ensuring more accurate reasoning. Additionally, we propose a reinforcement learning framework with a multi-process reward design to optimize agents for both multi-query retrieval and information consolidation. Extensive experiments on seven benchmarks demonstrate that MultiSearch outperforms baseline methods, enhancing the SNR of retrieval and improving reasoning performance in question-answering tasks.

Problem

• retrieval-augmented reasoning
• multi-query retrieval
• signal-to-noise ratio
• reasoning accuracy
• external knowledge retrieval

Innovation

• multi-query retrieval
• explicit merging
• reinforcement learning
• retrieval-augmented reasoning
• parallel search