Iterative method improves reasoning across multiple retrieved documents
RCER: Reconstructing Cross-Evidence Relations for Iterative Retrieval-Augmented Generation
Information Retrieval
Summary
Answering complex questions often needs combining information from many documents. Existing methods struggle to link related facts and trace how ideas connect across multiple sources. The authors propose a new approach called RCER that better identifies connections between entities and reasoning steps through multiple rounds of gathering information. This leads to more accurate and coherent answers, especially in complicated stories or implicit questions. Tests show RCER outperforms previous best methods on several question-answering tasks.
What this means in practice
- •For ai product developers: Improve complex question answering in AI assistants by better connecting information across multiple knowledge sources.$Commercial implications: Enables more accurate and coherent answers in AI-driven customer support and virtual assistants, making products more effective.
- •For knowledge management teams: Enhance systems that integrate scattered documents by accurately linking related facts and reasoning paths for better insights.
Authors
Linyao Zheng, Xuhang Shi, Zhifang Mao, Xiuquan Hou, Sai Zhou
Abstract
Iterative retrieval-augmented generation (RAG) accumulates knowledge from retrieved evidence to answer complex questions, but existing methods struggle to reconstruct entity correspondences and inferential dependencies across retrieved passages as new information emerges. We propose RCER, an iterative RAG method that reconstructs these cross-evidence relations directly from accumulated source passages. Inter-round relational evidence acquisition identifies entity correspondences and relational dependencies to drive dynamic problem decomposition and iterative retrieval. Reasoning context reconstruction independently constructs a coherent reasoning context through joint selection of complementary passages and derivation of cross-evidence inferences. Cross-evidence compositional reasoning leverages these inferences to establish entity correspondences and compose relations across selected passages for answer derivation. Experiments show that RCER outperforms the best baselines by 2.83 EM and 3.06 F1 points on average across three multi-hop QA benchmarks and achieves superior performance on narrative and implicit reasoning tasks.