Best Practices
Document Upload
Optimal Preparation
- Structure documents with clear headings
- Use consistent formatting
- Add metadata (title, author, date)
- Avoid watermarks and background images
Batch Upload
Upload related documents together (e.g., all chapters of a textbook). This makes later RAG exams easier.
Question Creation
Topic Formulation
- Specific rather than general
- Provide context
- Keep Bloom's level in mind
Examples
Good:
- "Python Lists – Methods append(), extend(), insert()"
- "Algorithms – Time Complexity of Sorting Methods"
Poor:
- "Python" (too broad)
- "Programming" (too general)
Quality Control
- Always review generated questions
- Pay attention to confidence scores
- Adjust difficulty levels accordingly
- Use source references for verification
RAG Exams
- Select 3–5 relevant documents (optimal)
- Give a specific focus
- Too many documents lead to lower quality
ChatBot Usage
Begin with overview questions and deepen progressively:
User: "What is Heapsort?"
Bot: [Explains Heapsort]
User: "How does that differ from Quicksort?"
Bot: [Compares both algorithms]
User: "Which is more efficient for large datasets?"
Bot: [Analyzes complexity]
Use Review Queue Effectively
- Review questions promptly after generation — context is fresher
- Use filter options (status, difficulty, question type) to keep the queue manageable
- Rejecting is better than allowing bad questions: Quality over quantity
- If many questions are rejected: Adjust prompt, improve source document, or refine topic
- Approve only questions you would ask yourself
Exam Composer
- Plan exam structure before selecting questions: How many questions? Which types? What difficulty distribution?
- Mix question types (Multiple Choice + open-ended) for varied exams
- Export a test version and read it fully through before creating the final version
- Check automatic numbering and formatting in the exported document