What type of study is this?

This is a Quantitative Study study (also classified as: Experimental Study).

September 22, 2025Open Access

eSapiens's DEREK Module: Deep Extraction & Reasoning Engine for Knowledge with LLMs

ISIsaac Shi ZLZeyuan LiChina Mobile (China)FLFan LiuUniversity of Science and Technology of China

Key Points

DEREK module improves Recall@50 by approximately 1 percentage point using 1000-token chunks, facilitating better retrieval.
The combination of hybrid and rerank methods increases Precision@10 by about 7 percentage points in document QA.
A LangGraph verifier ensures citation accuracy, reducing unsupported statements to less than 3% for higher reliability.
The system operates securely with end-to-end TLS 1.3 and AES-256 encryption, making it suitable for high-stakes domains.

Abstract

We present the DEREK (Deep Extraction the verifier raises TRACe Utilization above 0.50 and limits unsupported statements to less than 3%. All components run in containers, enforce end-to-end TLS 1.3 and AES-256. These results demonstrate that the DEREK module delivers accurate, traceable, and production-ready document QA with minimal operational overhead. The module is designed to meet enterprise demands for secure, auditable, and context-faithful retrieval, providing a reliable baseline for high-stakes domains such as legal and finance.

KI fragen

Bookmark

View Full Paper