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

TL;DR

Interact-RAG introduces an active retrieval manipulation paradigm for LLMs, enabling fine-grained control and reasoning in information-seeking tasks, leading to superior performance over existing black-box retrieval methods.

Contribution

This work presents a novel interaction framework with a Corpus Interaction Engine, allowing LLMs to actively manipulate retrieval, and develops a reasoning-enhanced workflow for improved RAG performance.

Findings

Outperforms existing methods on six benchmarks

Enables zero-shot execution and interaction trajectory synthesis

Improves retrieval accuracy and task success rates

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.