International Journal For Multidisciplinary Research

E-ISSN: 2582-2160   •   Impact Factor: 9.24

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Call for Paper Volume 8, Issue 4 (July-August 2026) Submit your research before last 3 days of August to publish your research paper in the issue of July-August.
Submit Research Paper Join as a Reviewer

Plagiarism is checked by the leading plagiarism checker TurnItIn

Call for Paper

Volume 8 Issue 4
July-August 2026

Submit your research paper

Indexing Partners
Academia Advanced Sciences Index Bielefeld Academic Search Engine CiteSeer DRJI Google Scholar Independent Search Engine & Directory Network ISI (International Scientific Indexing) Issuu Mendeley Research Networks
RefSeek ResearcherId - Thomson Reuters ResearchGate Scirus Scribd Semantic Scholar UTeM - Universiti Teknikal Malaysia Melaka Wiki for Call for Papers WorldCat Zenodo

LATTICE: Controlled Benchmarking of Hybrid Vector-SQL Retrieval for Enterprise Knowledge Bases Planning Across Tables, Passages, and Provenance for Auditable RAG

Author(s) Sandeep Nutakki
Country United States
Abstract Enterprise knowledge bases increasingly combine unstructured documents, relational records, metric tables, policy repositories, and operational event histories. Retrieval-augmented generation (RAG) systems usually treat this evidence as a passage-ranking problem, while relational systems preserve precise entity, row, and aggregate semantics but lack narrative recall. We present LATTICE, a hybrid vector-relational retrieval architecture that cross-links text chunks to typed enterprise entities and compiles natural-language questions into joint vector-SQL execution plans. In a controlled benchmark of 2,100 questions across three enterprise-style schemas with 33, 62, and 48 tables, LATTICE achieved 91.7% Hit@5 retrieval accuracy (95% CI: 90.5-92.9), 94.2% compute-answer accuracy on numeric and aggregation questions (95% CI: 92.1-96.0), and 98.1% citation precision using row-level provenance (95% CI: 97.2-98.9). For mixed vector-relational queries, early SQL pruning reduced median retrieval latency from 3.20 s for dense-only RAG to 0.41 s, a 7.8x speedup (p < 0.001). Ablations show that entity-card embeddings account for 7.2 percentage points of Hit@5 gain, the planner accounts for 16.9 points of compute accuracy, and row-level provenance accounts for 15.7 points of citation precision. Within this benchmark scope, the results support treating structured data as an execution substrate for enterprise RAG rather than only as text to embed.
Keywords hybrid retrieval, retrieval-augmented generation, vector search, relational databases, enterprise knowledge bases, text-to-SQL, provenance, explainable AI, benchmarking
Field Engineering
Published In Volume 8, Issue 3, May-June 2026
Published On 2026-06-08
DOI https://doi.org/10.36948/ijfmr.2026.v08i03.80826
View / Download PDF File

Share this



E-ISSN 2582-2160
CrossRef

CrossRef DOI prefix of IJFMR is 10.36948/ijfmr

Downloads
Research Paper Format Copyright Permission Form and Undertaking Form Cover Page Vol 8 Isu 4Cover Page Vol 8 Isu 3Cover Page Vol 8 Isu 2

All research papers published on this website are licensed under Creative Commons Attribution-ShareAlike 4.0 International License, and all rights belong to their respective authors/researchers.

CC-BY-SA Open Access

About IJFMR
Fees & Payment
Current Issue
Publication Archive 		Submit Research Paper
Track Submission Status
Publication Guidelines
Publication Ethics
Peer Review & Plagiarism 		Join as a Reviewer
Editors & Reviewers
Reviewer Referral Program
Get Reviewer Membership Certi. 		Website/Journal Policies
Usage Policy
Content Policies
Privacy Policy
Contact Us +91-9687-828-838 editor@ijfmr.com
Powered by Sky Research Publication and Journals