Summary

This project delivers an AI-powered personalized exam system for adaptive question generation and recommendation. It targets educators and learners, leveraging LLMs, knowledge graphs, and Bayesian Knowledge Tracing for tailored learning experiences beyond static question banks.

How It Works

The system integrates the OpenPangu 7B LLM for NLP tasks, a Knowledge Graph for structured domain knowledge, and a RAG engine for precise question recommendation. Student mastery is tracked via Bayesian Knowledge Tracing (BKT), enabling adaptive question selection. The LLM also automates knowledge graph construction and updates, creating a dynamic learning environment.

Quick Start & Requirements

• Installation: Docker-based deployment is recommended; requires code transfer into a container and running python main.py.
• Prerequisites:
  • Hardware: Ascend 910B2 NPU (essential for AI acceleration), CPU (Kunpeng-920 recommended), 196GB storage (100GB+ free).
  • Software: Linux (Ubuntu 22.04.5 LTS), Python 3.11.12 (in container), Docker 18.09.0+, Ascend CANN 23.0.6.
  • Models: OpenPangu-Embedded-7B-V1.1 and BGE-small-zh-v1.5 (download instructions provided).
  • Data: JSON question bank for initial migration.
• Setup: Invol