Machine Assistant with Reliable Knowledge: Enhancing Student Learning via RAG-based Retrieval

TL;DR

MARK is a retrieval-augmented question-answering system that improves student learning by providing accurate, contextually grounded responses through a hybrid retrieval strategy and adaptive feedback mechanisms.

Contribution

This paper introduces MARK, a novel RAG-based system with hybrid retrieval and feedback loops for reliable, adaptive educational support and domain-specific technical assistance.

Findings

Effective in classroom settings for student queries

Handles routine, context-sensitive tasks in applied domains

Improves retrieval robustness with hybrid search strategy

Abstract

We present Machine Assistant with Reliable Knowledge (MARK), a retrieval-augmented question-answering system designed to support student learning through accurate and contextually grounded responses. The system is built on a retrieval-augmented generation (RAG) framework, which integrates a curated knowledge base to ensure factual consistency. To enhance retrieval effectiveness across diverse question types, we implement a hybrid search strategy that combines dense vector similarity with sparse keyword-based retrieval. This dual-retrieval mechanism improves robustness for both general and domain-specific queries. The system includes a feedback loop in which students can rate responses and instructors can review and revise them. Instructor corrections are incorporated into the retrieval corpus, enabling adaptive refinement over time. The system was deployed in a classroom setting as a substitute for traditional office hours, where it successfully addressed a broad range of student queries. It was also used to provide technical support by integrating with a customer-specific knowledge base, demonstrating its ability to handle routine, context-sensitive tasks in applied domains. MARK is publicly accessible at https://app.eduquery.ai.