Interact-RAG: Reason and Interact with the Corpus, Beyond Black-Box Retrieval

TL;DR

Interact-RAG tackles the bottleneck of black-box retrieval in retrieval-augmented generation by equipping LLM agents with a Corpus Interaction Engine and a reasoning-enhanced workflow. The system trains an autonomous end-to-end agent via supervised fine-tuning on reasoned interaction trajectories, followed by reinforcement learning with Group Relative Policy Optimization, achieving significant gains across six RAG benchmarks, especially in multi-hop tasks. This approach improves retrieval efficiency and generalization to out-of-distribution data, demonstrating the value of integrating interactive retrieval with structured reasoning. Overall, Interact-RAG offers a transparent, capable paradigm for agent-driven information seeking in RAG pipelines.

Abstract

Retrieval-Augmented Generation (RAG) has significantly enhanced LLMs by incorporating external information. However, prevailing agentic RAG approaches are constrained by a critical limitation: they treat the retrieval process as a black-box querying operation. This confines agents' actions to query issuing, hindering its ability to tackle complex information-seeking tasks. To address this, we introduce Interact-RAG, a new paradigm that elevates the LLM agent from a passive query issuer into an active manipulator of the retrieval process. We dismantle the black-box with a Corpus Interaction Engine, equipping the agent with a set of action primitives for fine-grained control over information retrieval. To further empower the agent on the entire RAG pipeline, we first develop a reasoning-enhanced workflow, which enables both zero-shot execution and the synthesis of interaction trajectories. We then leverage this synthetic data to train a fully autonomous end-to-end agent via Supervised Fine-Tuning (SFT), followed by refinement with Reinforcement Learning (RL). Extensive experiments across six benchmarks demonstrate that Interact-RAG significantly outperforms other advanced methods, validating the efficacy of our reasoning-interaction strategy.

Figures (6)

• Figure 1: A brief demonstration of Interact-RAG. It empowers the agent with fine-grained control over the information-seeking process, leveraging a set of interactive actions. In contrast, conventional RAG is confined to ineffective loops of query issuing.
• Figure 2: An illustration of our reasoning-enhanced workflow.
• Figure 3: Comparison of the efficiency of information retrieval.
• Figure 4: Performance during RL training. Measured on a sampled subset.
• Figure 5: Action invocation status in different training stages. Measured on the Musique dataset.