• Vedant Fuley
• V Jogendra Rao
• Harsha Nishad
• Sourabh Dewangan

The increasing demand for scalable and consistent evaluation of open-ended responses in educational environments necessitates robust AI-assisted assessment systems. This paper introduces the Human Answer Reconciliation Pipeline (HARP), a modular and extensible framework designed to automate the grading of text-based student responses using Large Language Models (LLMs). Unlike traditional systems that rely solely on reference answers, HARP supports both reference-based and standalone answer evaluation modes, enabling flexible assessment across diverse pedagogical contexts. The system features a FastAPI-based backend with customizable scoring formats (e.g., 0–10, 0–5, percentage), batch evaluation capabilities, and caching for performance optimization. The Model Context Protocol (MCP) Server is central to HARP's architecture, which enables modular LLM integration, file-based request handling, and standardized evaluation flows. Additionally, HARP employs techniques such as Reference Answer Guided (RAG) evaluation and zero/few-shot prompting strategies for enhanced scoring accuracy and interpretability. Designed for deployment in academic institutions, e-learning platforms, and AI research settings, HARP bridges the gap between manual grading and intelligent automation, delivering transparent, adaptable, and pedagogically aligned assessments.

Automated Answer Evaluation, Human Answer Reconciliation Pipeline (HARP), Large Language Models (LLM), FastAPI, Model Context Protocol (MCP), Educational Technology, RAG Evaluation, NLP in Education, Batch Scoring System, AI-based Grading.

"HARP: Human Answer Reconciliation Pipeline ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 4, page no.k344-k355, April-2025, Available :http://www.jetir.org/papers/JETIR2504A42.pdf

"HARP: Human Answer Reconciliation Pipeline ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 4, page no. ppk344-k355, April-2025, Available at : http://www.jetir.org/papers/JETIR2504A42.pdf