Hierarchical Evidence-Driven Reasoning for Long Document Understanding

2026-07-05

AI Summary

This work addresses the limitations of existing retrieval-augmented generation (RAG) approaches in long-document understanding, which are prone to interference from semantically similar yet irrelevant passages and suffer from overreliance on initial retrieval results, often missing critical evidence and propagating cascading errors. To overcome these challenges, the authors propose HIEVI-RAG, a novel framework featuring a hierarchical evidence-driven mechanism. It orchestrates a four-stage process—multi-hop question decomposition, coarse-grained visual page retrieval, fine-grained cross-page verification, and memory-guided iterative generation—to enable iterative evidence accumulation and dynamic reasoning. The framework integrates a multimodal retriever, an EVIAGENT multi-page verification agent trained via GRPO, and a memory-augmented generation strategy. Evaluated on four benchmarks, HIEVI-RAG substantially outperforms current open-source methods, achieving an average accuracy improvement of 8.05%.

Abstract

Retrieval-Augmented Generation (RAG) streamlines long-document understanding by leveraging retrieval mechanisms to restrict input images to a highly curated subset. However, existing multimodal RAG pipelines primarily face two critical challenges: first, standard semantic similarity retrievers frequently fetch topically overlapping yet answer-void distractor pages that mislead downstream generation; second, rigid single-pass pipelines heavily depend on initial retrieval success, where any omission of core evidence inevitably causes cascading errors. To address these challenges, we introduce HIEVI-RAG, a hierarchical, evidence-driven multimodal RAG framework for closed-domain document understanding. HIEVI-RAG systematically factorizes complex queries into a cooperative four-stage pipeline: (1) hierarchical question decomposition to break multi-hop root queries into atomic child questions; (2) coarse visual page retrieval leveraging a multimodal retriever to fetch candidate pages based on semantic similarity; (3) fine-grained page verification via EVIAGENT, a specialized multi-page verifier trained with GRPO to execute cross-page reasoning over multi-image blocks; and (4) memory-guided iterative generation that leverages accumulated sub-question context to execute multi-round, dynamic reasoning over the prioritized sequence. Extensive evaluations across four benchmarks demonstrate the robust efficacy and synergy of our framework, which significantly outperforms existing open-source baselines and exceeds the strongest reported baseline by an average of 8.05% in accuracy.

Problem

• Retrieval-Augmented Generation
• long document understanding
• multimodal RAG
• evidence retrieval
• distractor pages

Innovation

• Hierarchical Reasoning
• Evidence-Driven Verification
• Multimodal RAG
• Iterative Generation
• Cross-Page Reasoning