LLM-Quiz-Generator: An Open-Source Large Model-Driven Tool for Automatic Educational Quiz Generation

LLM-Quiz-Generator is an open-source project that explores the use of open-source large language models to automatically generate educational quiz questions. It supports extracting key knowledge points from learning materials and generating question-answer pairs, providing an intelligent quiz generation solution for teachers and educators. This solves the time-consuming and labor-intensive problem of manual question creation, and it can be run locally with full control over the generation process.

Teachers manually designing high-quality quizzes need to cover core knowledge points, reasonable difficulty gradients, clever distractors, etc., which takes several hours. The development of personalized education requires generating customized quizzes for different groups, further increasing the workload. This has prompted the exploration of AI automated question creation possibilities.

System Architecture

Modular design, including input processing (multi-format support, text preprocessing), knowledge extraction (identifying concepts and relationships), question generation (questions and answers/distractors), and quality assessment modules.

Open-Source Model Support

Recommended models include Llama2/3, Mistral, Qwen, etc. Deployment methods include local (llama.cpp/Ollama), self-hosted API (vLLM/TGI), and cloud (Hugging Face).

Prompt Engineering Strategy

Use carefully designed prompt templates, such as knowledge extraction prompts (extracting concept names, definitions, details) and question generation prompts (specifying quantity, difficulty, question type, etc.).

Teacher Lesson Preparation Assistant

Quickly generate chapter quizzes, differentiated questions, current affairs-related questions, and practice question banks.

Online Education Platforms

Automatically generate module quizzes, adaptive tests, and unlimited practice questions, reducing costs.

Corporate Training

Generate assessment questions from documents, ensure content mastery, regular updates, and multi-language support.

Self-Learning Tools

Convert materials into quizzes, identify knowledge gaps, and generate targeted practice.

Advantages

• Open-source and localized: Data privacy, low cost, customizable, offline available;
• Educational professionalism: Supports Bloom's taxonomy, cognitive load theory;
• Flexible and scalable: Add question types, integrate new LLMs, custom scoring, export multiple formats.

Limitations

Quality control (unclear expression, distractor issues), subject limitations (mathematical derivation, etc.), model dependence (hallucinations, insufficient capabilities of small models).

Countermeasures

Human-machine collaboration (AI generation + manual review), quality checklists, prompt optimization, few-shot learning.

Tool Comparison

Feature	LLM-Quiz-Generator	Commercial AI Tools	Traditional Question Bank Systems
Open-source	✅	❌	Partial
Local run	✅	❌	✅
Customization	High	Low	Medium
Cost	Low	High	Medium
Question quality	Model-dependent	Relatively high	High
Usage threshold	Medium	Low	Low
Unique value: The combination of open-source + local + customizable, suitable for institutions with technical capabilities.

Quick Start

1. Clone the repository: git clone https://github.com/raquelvargas16/llm-quiz-generator
2. Install dependencies: pip install -r requirements.txt
3. Configure the model: Edit config.json to specify endpoints, models, etc.
4. Generate quizzes: Use Python code to load materials, generate, and export quizzes.

Future Directions

• Short-term: Improve prompt library, multi-input formats, quality assessment, user-friendly UI;
• Mid-term: Adaptive difficulty, similarity detection, learning analysis, multi-language enhancement;
• Long-term: End-to-end evaluation system, LMS integration, personalized paths, multi-modal questions.