GRACE RAG Governed Retrieval Architecture for Canonical Evidence Synthesis Enabling Lightweight Deployment in Closed Domain Institutional Settings

Jul 16, 2026

Retrieval-Augmented Generation (RAG) is commonly used for institutional question answering, yet entity-dense corpora often cause vector-only retrieval to return fragmented evidence and increase dependence on inference-time reasoning. GRACE-RAG proposes a retrieval-governed, graph-augmented architecture that externalizes structural reasoning into a structured retrieval layer via offline knowledge construction. Experiments on Mistral 24B, GPT OSS 120B, and Gemini 2.5 Flash show up to 20% quality gains, with improved completeness, depth, and anticipatory coverage while reducing latency and avoiding proprietary dependencies.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages15

Reading Time38 min

Introduction

Related Work

Vector-Based Retrieval-Augmented Generation

Graph-Based Retrieval-Augmented Generation

GRACE-RAG: Governed Retrieval Architecture for Canonical Evidence Synthesis, Enabling Lightweight Deployment in Closed-Domain Institutional Settings Asit Desai National Payments Corporation of India Aman Kumar National Payments Corporation of India arXiv :2607 .000 13v 1 [ cs .IR] 8 May 2026 Prashant Devadiga National Payments Corporation of India Abstract Retrieval-Augmented Generation (RAG) systems are widely used in institutional question answering settings where responses must be grounded in authoritative documentation (Gao et al., 2023) . In entity-dense domains where relevant information is distributed across heterogeneous documents, vector-only retrieval often produces fragmented evidence and increases dependence on inference-time reasoning (Zhao et al., 2024) . This paper introduces GRACE-RAG, a retrieval-governed, graph-augmented RAG architecture that externalizes structural reasoning from the generative stage to a structured retrieval layer, resolving structural ambiguity offline, enabling deployment on self-hosted lightweight models calibrated to closed-domain institutional vocabulary. Experiments across three model capacities: Mistral 24B, GPT OSS 120B, and Gemini 2.5 Flash show consistent improvements in completeness, depth, and anticipatory coverage, with overall quality gains of up to 20% under mid-scale models, indicating that retrieval architecture governs structural quality over model scale, reducing computational and latency footprint without dependence on proprietary systems. 1 1 Introduction Institutional question answering systems operate under constraints that differ substantially from open-domain conversational assistants (Peng et al., 2024; Lund, 2025) . Queries in such environments frequently reference domain-specific entities, operational limits, eligibility rules, or conditional workflows whose relevant information is distributed across heterogeneous documents (Xu et al., 2024) . Retrieval-Augmented Generation (RAG) grounds language model outputs in external knowledge sources (Gao et al., 2023), but in entity-dense institutional corpora, semantic proximity alone is insufficient as queries may span multiple documents through implicit relational dependencies (Lewis et al., 2020) . To address these limitations, many systems introduce prompt-level orchestration or agentbased control flow (Gupta et al., 2024), increasing latency and computational cost through reliance on proprietary models that prioritize cross-domain generalization at the expense of domain-specific terminology precision (Arslan, 2024) . This work adopts an alternative perspective: structural ambiguity should be resolved prior to generation, and the language model should be restricted to synthesizing evidence rather than performing implicit structural reasoning (Cheng et al., 2025) . We therefore introduce GRACE-RAG, a retrieval-governed RAG architecture that externalizes entity normalization, relationship modeling, and semantic boundary alignment into an offline 1Preprint structure-manufacturing pipeline. During online inference, hybrid retrieval operates over dual embedding surfaces, content chunks and relationship summaries, allowing relational hypotheses to be ranked and validated before generation (Gupta et al., 2024; Wan et al., 2025) . The contributions of this paper are threefold: • We introduce GRACE-RAG, a retrieval-governed RAG architecture that decouples structural reasoning from generation through offline knowledge construction and bounded hybrid retrieval. • We present a dual-surface retrieval mechanism in which relationship summaries are embedded and indexed independently, enabling relational validation and graph-guided expansion without uncontrolled traversal. • We empirically demonstrate that governed retrieval with canonical evidence structuring delivers consistent structural quality gains, enabling a practical shift to selfhosted lightweight models with measurably reduced computa

FAQs

What problem does GRACE-RAG address in institutional question answering?

In entity-dense institutional corpora, vector-only retrieval can produce fragmented evidence, forcing the language model to rely on inference-time structural reasoning.

How does GRACE-RAG differ from prompt-level orchestration or agent-based control?

It resolves structural ambiguity before generation by externalizing entity normalization, relationship modeling, and boundary alignment into an offline structure-manufacturing pipeline.

What benefits do experiments report across different model capacities?

Across Mistral 24B, GPT OSS 120B, and Gemini 2.5 Flash, governed retrieval improves completeness, depth, and anticipatory coverage, yielding up to 20% overall quality gains while lowering computational and latency footprint.

Jul 16, 2026

Upload by Emma Wilson