FREESON: Retriever-Free Retrieval-Augmented Reasoning via Corpus-Traversing MCTS

2025-05-22

AI Summary

Existing retrieval-augmented reasoning approaches rely on standalone retrievers, leading to a representation bottleneck—where retrieval embeddings poorly align with generative requirements—and increased system overhead. This paper proposes the first retriever-agnostic, end-to-end reasoning framework, enabling Large Reasoning Models (LRMs) to autonomously traverse corpora and localize answers. Key contributions include: (1) CT-MCTS, a corpus-structure-aware Monte Carlo Tree Search algorithm that formulates retrieval as hierarchical path search over document graphs; (2) an LRM-driven retrieval mechanism that unifies retrieval and generation within a single model; and (3) a path-guided answer localization paradigm that grounds predictions in traversal trajectories. Evaluated on five open-domain question answering benchmarks, our method achieves average improvements of +14.4% in Exact Match (EM) and F1 scores, outperforming the strongest baselines by 3% on PopQA and 2WikiMultihopQA.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Knowledge Representation and Reasoning: Common-Sense Reasoning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• 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

Large Reasoning Models (LRMs) have demonstrated remarkable capabilities in multi-step reasoning and calling search engines at appropriate steps. However, existing retrieval-augmented reasoning approaches rely on separate retrieval models, limiting the LRM's role in retrieval to deciding when to retrieve and how to query. This separation not only increases hardware and operational costs but also leads to errors in the retrieval process due to the representation bottleneck, a phenomenon where the retriever's embedding space is not expressive enough to meet the generator's requirements. To address this, we shift our perspective from sequence-to-sequence matching to locating the answer-containing paths within the corpus, and propose a novel framework called FREESON (Retriever-FREE Retrieval-Augmented ReaSONing). This framework enables LRMs to retrieve relevant knowledge on their own by acting as both a generator and retriever. To achieve this, we introduce a variant of the MCTS algorithm specialized for the retrieval task, which we call CT-MCTS (Corpus-Traversing Monte Carlo Tree Search). In this algorithm, LRMs traverse through the corpus toward answer-containing regions. Our results on five open-domain QA benchmarks, including single-hop and multi-hop questions, show that FREESON achieves an average improvement of 14.4% in EM and F1 over four multi-step reasoning models with a separate retriever, and it also performs comparably to the strongest baseline, surpassing it by 3% on PopQA and 2WikiMultihopQA.

Problem

• Eliminates need for separate retrieval models in reasoning tasks
• Addresses representation bottleneck in retrieval-augmented reasoning
• Enables models to autonomously retrieve and reason without retrievers

Innovation

• Combines generator and retriever roles in LRMs
• Uses Corpus-Traversing MCTS for retrieval
• Eliminates separate retrieval models