A Systematic Framework for Enterprise Knowledge Retrieval: Leveraging LLM-Generated Metadata to Enhance RAG Systems
This post summarizes the paper "A Systematic Framework for Enterprise Knowledge Retrieval: Leveraging LLM-Generated Metadata to Enhance RAG Systems" (arXiv:2512.05411), by Pranav Pushkar Mishra, Kranti Prakash Yeole, Ramyashree Keshavamurthy, Mokshit Bharat Surana, and Fatemeh Sarayloo (Dec 2025).
Key points
- Problem: In enterprise settings, efficiently retrieving relevant information from large, complex knowledge bases is essential for operational productivity and informed decision-making. Standard RAG pipelines often lack rich semantic structure at the document-segment level.
- Approach: A structured pipeline uses LLMs to dynamically generate meaningful metadata for document segments, substantially improving their semantic representations and retrieval accuracy.
- Experimental design: A controlled 3x3 matrix compares three chunking strategies (semantic, recursive, naive) with three embedding techniques (content-only, TF-IDF weighted, prefix-fusion), with ablation analysis isolating each component's contribution.
- Metadata-enriched approaches consistently outperform content-only baselines.
- Recursive chunking + TF-IDF weighted embeddings: 82.5% precision.
- Naive chunking + prefix-fusion: strongest ranking quality, NDCG 0.813.
- Ground truth generated via cross-encoder reranking (silver standard); statistical significance confirmed with Bonferroni-corrected paired t-tests.
- Metadata enrichment maintains sub-30 ms P95 latency.
- The work sits at the intersection of RAG and large-scale search/retrieval, addressing how to reallocate responsibilities across retrieval, ranking, and generation in the LLM era.
- For engineers: metadata enrichment is a pluggable module — it can be added to existing RAG stacks without changing the core retriever, with quantifiable precision/NDCG gains and negligible latency cost.
- For researchers: the 3x3 factorial design with ablations and corrected significance testing offers a reproducible evaluation template for chunking/embedding studies.
- Related literature includes surveys on RAG and reasoning (arXiv:2504.15909), agentic information retrieval (arXiv:2410.09713), and RAG in vision (arXiv:2503.18016).
Headline results
Original abstract
> In enterprise settings, efficiently retrieving relevant information from large and complex knowledge bases is essential for operational productivity and informed decision-making. This research presents a systematic empirical framework for metadata enrichment using large language models (LLMs) to enhance document retrieval in Retrieval-Augmented Generation (RAG) systems. Our approach employs a structured pipeline that dynamically generates meaningful metadata for document segments, substantially improving their semantic representations and retrieval accuracy. Through a controlled 3 X 3 experimental matrix, we compare three chunking strategies -- semantic, recursive, and naive -- and evaluate their interactions with three embedding techniques -- content-only, TF-IDF weighted, and prefix-fusion -- isolating the contribution of each component through ablation analysis. The results demonstrate that metadata-enriched approaches consistently outperform content-only baselines, with recursive chunking paired with TF-IDF weighted embeddings yielding 82.5% precision and naive chunking with prefix-fusion achieving the strongest ranking quality (NDCG 0.813). Our evaluation employs cross-encoder reranking for silver-standard ground truth generation, with statistical significance confirmed via Bonferroni-corrected paired t-tests. These findings confirm that metadata enrichment improves vector space organization and retrieval effectiveness while maintaining sub-30 ms P95 latency, providing a quantitative decision framework for deploying high-performance, scalable RAG systems in enterprise settings.