CS182 Special Participation: LLM Analysis Website

This website documents student experiences using various Large Language Models (LLMs) to solve homework problems for CS182: Deep Learning.

📊 Overview

• 110 Special Participation A posts - Students used LLMs for non-coding homework problems
• 90 Special Participation B posts - Students used LLMs for coding homework problems
• 13+ LLMs analyzed including DeepSeek, Gemini, Grok, Mistral, ChatGPT, Claude, and more
• Searchable and filterable by student name, LLM model, homework assignment, and keywords

🎯 Purpose

This website serves to:

1. Document student insights on LLM behaviors and capabilities
2. Provide a searchable resource for understanding different LLMs' strengths and weaknesses
3. Credit students for their contributions with links to their profiles/work
4. Create a knowledge base for future students and research

✨ Features

Search & Filter

• Full-text search across all submissions
• Filter by participation type (A or B)
• Filter by LLM model
• Filter by homework assignment
• Filter by student name

LLM Analysis

• Comparative analysis of different LLMs
• Common strengths and weaknesses for each model
• Behavior patterns observed by students
• Statistics on usage and performance

Submissions

• Complete student submissions with attribution
• Links to chat transcripts, Google Docs, and GitHub repos
• View counts and engagement metrics
• Staff comments and endorsements
• Expandable content for easy browsing

🚀 Deployment

Option 1: Simple HTTP Server (Local Testing)

cd website
python3 -m http.server 8000

Then open http://localhost:8000 in your browser.

Option 2: Static Hosting (Production)

The website is a static site with no backend dependencies. You can deploy it to any static hosting service:

GitHub Pages

# Push the website folder to a GitHub repository
git add website/
git commit -m "Add LLM analysis website"
git push

# Enable GitHub Pages in repository settings
# Point to the website folder

Netlify

# Drag and drop the website folder to Netlify
# Or use Netlify CLI:
netlify deploy --prod --dir=website

Vercel

# Install Vercel CLI
npm i -g vercel

# Deploy
cd website
vercel --prod

eecs182.org Integration

To integrate with the main eecs182.org website:

1. Copy the entire website directory to the eecs182.org server
2. Rename it if needed (e.g., llm-participation)
3. Link to it from the main site: https://eecs182.org/llm-participation/

The website is self-contained with:

• All data in JSON format in the data/ folder
• No external dependencies
• No build process required
• Works with or without a domain name

📁 File Structure

website/
├── index.html          # Main HTML page
├── styles.css          # All CSS styles
├── app.js              # JavaScript application
├── data/               # JSON data files
│   ├── participation_a.json
│   ├── participation_b.json
│   ├── insights_a.json
│   ├── insights_b.json
│   └── statistics.json
└── README.md           # This file

🔧 Technical Details

Technologies Used

• Pure HTML/CSS/JavaScript - No frameworks required
• Responsive Design - Works on desktop and mobile
• Client-side Search - Fast filtering without backend
• JSON Data - Easy to update and maintain

Browser Support

• Chrome/Edge (latest)
• Firefox (latest)
• Safari (latest)
• Mobile browsers

Performance

• Static files only - fast load times
• Client-side rendering - no server load
• Optimized for 200+ submissions
• Lazy loading for content expansion

📝 Data Sources

All data is sourced from Ed Discussion posts (https://edstem.org/us/courses/84647):

• Downloaded on December 8, 2025
• 558 total posts analyzed
• 200 Special Participation posts extracted
• Automated parsing with manual verification

🎓 Student Credits

All submissions are attributed to students with:

• Full name displayed
• Links to external resources (chat logs, documents, repos)
• View counts and engagement metrics
• Staff endorsements highlighted

Students who included personal websites or GitHub profiles have them linked directly in their submissions.

📈 Statistics

Special Participation A (Non-coding problems)

• Top LLMs: DeepSeek (16), Mistral (10), Gemini (9)
• Most tested: HW3, HW4, HW2
• Common themes: Accuracy, Hallucinations, Explanation Quality

Special Participation B (Coding problems)

• Top LLMs: Gemini (14), DeepSeek (8), Grok (8)
• Most tested: HW4, HW3, HW2
• Common themes: Error patterns, Iterative problem-solving

🔄 Updating Data

To update the website with new posts:

1. Download new Ed posts:

python download_ed_posts.py

2. Parse and analyze:

python parse_participation_posts.py
python analyze_insights.py

3. Copy new data:

cp website_data/*.json website/data/

4. Deploy updated website

🐛 Troubleshooting

Issue: Data not loading

• Check browser console for errors
• Ensure JSON files are in data/ folder
• Verify JSON files are valid (use a JSON validator)
• Check CORS settings if hosting locally

Issue: Search not working

• Clear browser cache
• Check JavaScript console for errors
• Ensure app.js is loaded correctly

Issue: Links broken

• Verify external links are still valid
• Check for URL encoding issues
• Ensure HTTPS/HTTP protocol is correct

📧 Contact

For questions about this website or the CS182 course, visit:

• Course website: https://eecs182.org
• Ed Discussion: https://edstem.org/us/courses/84647

📄 License

This website and data are for educational purposes as part of CS182: Deep Learning at UC Berkeley. Student contributions remain the intellectual property of the respective authors.

---

Built with ❤️ for CS182 students and future deep learning researchers.