ExTest-Predict

How The AI Prediction Works

ExTest-Predict uses the GLM-4.6 AI language model to analyze your course materials and predict potential exam questions:

Content Analysis Phase

When you upload files, our system processes them through multiple analysis stages:

• Text Extraction - Pulls text from PDFs, images (OCR), audio (speech-to-text), and video
• Topic Identification - Identifies main topics, subtopics, and key concepts
• Pattern Recognition - Finds recurring themes, terminology, and relationships
• Emphasis Detection - Notes which topics are covered most extensively
• Concept Mapping - Understands how different ideas connect to each other

Question Generation Phase

Based on the analysis, the AI generates multiple-choice questions that:

• Test understanding of core concepts (not just memorization)
• Mirror the style and difficulty level of your materials
• Cover topics proportionally to how they appear in your content
• Include plausible distractor options (wrong answers that test understanding)
• Vary in difficulty from recall to application questions
• Incorporate your professor preferences - If you provided information about your professor's teaching style, the AI uses this to tailor questions to match their patterns

Likelihood Scoring System

Each question is assigned a likelihood score based on:

• 85%+ (High) - Topics covered extensively, central concepts, heavily emphasized in materials
• 60-84% (Medium) - Important topics mentioned multiple times but not the main focus
• 40-59% (Low) - Minor topics, peripheral details, mentioned briefly or in passing

Important: These scores represent how well questions match your course content and the AI's confidence in the prediction. They are NOT guarantees that questions will appear on the actual exam. Think of them as a study priority guide.

Professor Style Adaptation

The AI analyzes your professor's teaching patterns to generate questions similar to their style:

• Question phrasing and structure
• Types of concepts they test (definitions, applications, comparisons, etc.)
• Difficulty level progression
• Common themes they emphasize
• Distractor patterns (what makes wrong answers tempting)

Features & Capabilities

Multi-Format Upload

Upload in any format you have - PDFs, PowerPoint slides, images of notes, audio recordings, video lectures, or text files. The system handles all formats seamlessly.

Professor Preferences Input (NEW!)

Tell the AI about your professor's teaching style and preferences. Get personalized predictions based on what YOUR professor focuses on. Mention topics they emphasize, question types they prefer, and their testing patterns for even more accurate predictions.

Advanced AI Analysis

Powered by GLM-4.6, providing deep understanding of academic content across all subjects.

Smart Likelihood Scoring

Questions are ranked by likelihood so you know what to prioritize. Focus your study time on high-likelihood questions while still reviewing medium and low ones.

Interactive Quiz Mode

Don't just read - practice! Take an interactive quiz with immediate feedback. See explanations for each answer to learn from mistakes.

Progress Tracking

Track your quiz performance over time. See which topics you're strong in and which need more work. Retake quizzes as many times as you want.

Regenerate Anytime

Not satisfied with the questions? Click "Regenerate" to get a fresh set of predictions from your materials. Each generation may produce different questions.

Export (Coming Soon)

Soon you'll be able to export your questions to PDF for offline studying or printing.

Dark Mode Design

Easy on the eyes for late-night study sessions. Our professional dark theme reduces eye strain during long study periods.

Tips for Best Results

Upload Comprehensive Materials

Why: More diverse materials = better analysis

Include lecture notes, textbook chapters, slides, handouts, and any supplementary materials. The AI finds connections across different sources to identify truly important topics.

Use Current Semester Content

Why: Professors test what they recently taught

Upload materials from this semester or course. Old materials from previous semesters may not reflect current teaching focus.

Mix Different File Types

Why: Each format captures different information

• Slides - Show structure and main points
• Notes - Include details and explanations
• Recordings - Capture emphasis and context

Upload Multiple Files

Why: More data = better predictions

Don't just upload one slide deck. Upload all lectures, all notes, all materials for comprehensive coverage.

Use Professor Preferences (NEW!)

Why: Personalization = better accuracy for YOUR course

Take advantage of the Professor Preferences text box! Share insights about your professor's testing style:

• Topics they repeatedly emphasize in lectures
• Types of questions they love (case studies, calculations, definitions, comparisons)
• Specific theories, concepts, or authors they focus on
• Whether they prefer practical applications vs. theoretical understanding
• Any patterns you've noticed from previous exams or assignments

The more specific you are, the better the AI can tailor questions to match YOUR professor's style. This feature can significantly improve prediction accuracy for your specific course.

Review All Likelihood Levels

Why: Professors surprise us sometimes

Even low-likelihood questions can appear on exams. Review everything, but prioritize high and medium likelihood questions.

Use the Quiz Feature Repeatedly

Why: Active recall = better retention

Take quizzes multiple times. Each time you'll learn something new. Track your improvement and focus on weak areas.

Read Explanations Carefully

Why: Understand the "why" not just the answer

The AI provides explanations for each question. Read them even for questions you get right - you'll learn the reasoning behind the answer.

Don't Rely Exclusively on Predictions

Why: AI can't see everything

Use this as a study tool, not a replacement for learning. The AI predicts based on your materials, but your professor may test things not explicitly covered. Always study your complete course materials.

Frequently Asked Questions

Is this cheating?

No! This is a study tool, similar to having a smart friend quiz you based on your notes. You're still learning the material yourself - we're just helping you practice more efficiently. It's no different than using practice problems from a textbook or study guide.

How accurate are the predictions?

The likelihood scores indicate how well questions match your course content and emphasis, NOT a guarantee they'll appear on the exam. Think of it as a sophisticated study guide that helps you focus on what matters most. Many students find these predictions helpful, but results vary by course and professor.

What happens to my uploaded files?

Your files are processed by our AI to generate questions, then stored securely for your session. They are only used to create your personalized question predictions. We do not share your files with anyone or use them for any other purpose.

Can I use this for any subject?

Yes! The AI works across all subjects - STEM, humanities, arts, business, medicine, law, etc. As long as you have course materials in supported formats, ExTest-Predict can generate practice questions.

How many files can I upload?

As many as you need! Each file must be under 50MB. Most users upload 5-15 files per session including slides, notes, and recordings. There's no limit on the total number of files.

Is my data private and secure?

Yes. Your files are processed securely using industry-standard encryption. We do not share your data with third parties or use it to train our models. Your academic materials remain private.

Why are some questions marked "low likelihood"?

Low likelihood means the topic was covered briefly or wasn't emphasized in your materials. However, professors sometimes test these topics too! We recommend reviewing all questions but prioritizing high and medium likelihood ones.

Can I regenerate questions?

Yes! Click the "Regenerate" button after viewing your results. You'll get a fresh set of predictions. Each regeneration may produce different questions from the same materials.

Does this work for online classes?

Absolutely! Upload screenshots, lecture notes, or recordings from online courses just like you would for in-person classes.

What if my files aren't supported?

If you have a file format we don't support, try converting it. For example, convert Word docs to PDF, or download online slides as PDF. Most formats can be converted easily.

What should I write in the Professor Preferences box?

Share anything you've noticed about your professor's teaching style! Examples: "They love questions about photosynthesis and cellular respiration" or "They focus on practical applications instead of theory" or "They always ask about specific case studies from lectures." The more specific, the better the predictions will match your professor's style.

Is the Professor Preferences feature required?

No, it's completely optional! You can leave it blank and still get great predictions based on your uploaded materials alone. However, providing professor preferences can help tailor the questions to be even more accurate for YOUR specific course and professor.

Technical Specifications

Supported File Formats

Documents: PDF, TXT, JSON

Presentations: PPT, PPTX

Images: JPG, JPEG, PNG, GIF

Audio: MP3, WAV

Video: MP4, AVI, MOV

File Size Limit

Maximum 50MB per file. Most lecture slides and notes are well under this limit. For very large video files, consider breaking them into smaller segments.

AI Model

Powered by GLM-4.6 (General Language Model 4.6), an advanced language model developed for complex reasoning and content understanding.

Browser Compatibility

Works on all modern browsers including Chrome, Firefox, Safari, and Edge. For the best experience, we recommend using the latest version of Chrome or Firefox.

Processing Time

Small uploads (1-3 files, <10MB): 10-20 seconds

Medium uploads (4-10 files, 10-50MB): 20-40 seconds

Large uploads (10+ files, >50MB): 40-60 seconds

Times may vary based on file complexity and server load.

Session Management

Your questions are stored for the current session. If you close the browser or refresh, you'll need to regenerate. We're working on persistent storage for future updates.

Internet Requirements

Stable internet connection required for uploading files and generating questions. Once questions are generated, quizzes work offline for the current session.

Example Use Cases

MIDTERM PREPARATION

Upload all materials from the first half of the semester. Generate 15-20 questions to identify key themes. Take quizzes multiple times. Focus extra time on high-likelihood topics.

FINAL EXAM REVIEW

Upload comprehensive materials including all lectures, notes, and textbook chapters. Generate 20 questions for full exam simulation. Take timed quizzes to practice pacing. Review low-likelihood questions too - finals often cover cumulative content.

QUICK BEFORE-CLASS REVIEW

Upload just the recent lecture slides. Generate 5 questions for a quick 10-minute review session before class. Helps reinforce what you just learned.

GROUP STUDY

Use predictions to create group study sessions. Each person takes the quiz independently, then discuss answers and explanations together. Great for identifying knowledge gaps.

TUTORING SESSION PREP

Generate questions before meeting with a tutor or study group. Use the predictions to guide your session - focus on topics you got wrong or were unsure about.