Open elective · 60 marks · 2 hrs
Generative AI
Exam: 27 Nov 2026 · 14:30-16:30 · Mumbai University, B.E. AI & DS Sem III (NEP-2020).
Format note: Dec 2025 paper: 60 marks / 2 hours (QP 25503). May 2026 paper: 30 marks / 1 hour (QP 3001). Winter 2026 format is unconfirmed — the official timetable lists the 27 Nov 2026 paper only as 'O/E' (open elective).
Units & topics
| Unit | Title & topics |
|---|---|
| Unit 1 | Introduction to Generative AI, Language Models & LLM Architectures RNN — architecture & working, LLM — Large Language Models, Generative AI — definition & scope, Challenges in Generative AI, Importance in five real-world domains, Traditional vs modern language models, Transformer architecture, LSTM, Tokenization in transformers |
| Unit 2 | GPT, ChatGPT, Ethics & Generative AI Tools GPT architecture & working, GPT variants, Ethics in Generative AI, Generative AI tools / applications, Pre-training, Chat-GPT, Multi-head mechanism in GPT, Self-attention in GPT, GPT and Chat-GPT — definition |
| Unit 3 | Asked anyway — not in the syllabus CNN — architecture & working, Importance of data in training |
Question papers
Scanned originals as printed. Open a paper for preview, or download the PDF.
Dec 2025 · QP 25503
60 marks · 2 hrs · verified — scanned original
May 2026 · QP 3001
30 marks · 1 hr · verified — scanned original
Most-asked topics
How many questions each topic got in Dec 2025 and May 2026. History, not a prediction — nothing here is guaranteed to appear next.
| Topic | Dec 2025 | May 2026 | Total | Marks |
|---|---|---|---|---|
| GPT architecture & working | 1 | 2 | 3 | 17 |
| RNN — architecture & working | 1 | 1 | 2 | 13 |
| LLM — Large Language Models | 1 | 1 | 2 | 10 |
| Generative AI — definition & scope | 1 | 1 | 2 | 7 |
| GPT variants | 1 | 1 | 2 | 10 |
| Ethics in Generative AI | 1 | 1 | 2 | 10 |
| Generative AI tools / applications | 1 | 1 | 2 | 10 |
| Pre-training | 1 | 1 | 2 | 9 |
Question sets
Topic-wise question sets from the two papers — questions only.
Gen AI — Topic-wise Question Sets (Dec 2025 + May 2026)
4 pages · Scanned university question papers, grouped by topic · Questions only — no answers
All questions — Dec 2025 + May 2026
30 questions, quoted as printed. No answers are published on this site.
| Q | Topic | Question | Marks | Paper |
|---|---|---|---|---|
| Q1a | Generative AI — definition & scope | Define Generative AI. Explain its scope with suitable examples? | 5 | genai-dec2025 |
| Q1b | GPT variants | List any three GPT variants and mention one use case for each. | 5 | genai-dec2025 |
| Q1c | Importance of data in training | What is the importance of data in training the AI model? | 5 | genai-dec2025 |
| Q1d | Generative AI tools / applications | Explain "text-to-image" tools with an example. | 5 | genai-dec2025 |
| Q2a | CNN — architecture & working | Explain How CNN Works? | 8 | genai-dec2025 |
| Q2b | Pre-training | Explain pre-training and fine tuning and instruct tuning in gen-AI? | 7 | genai-dec2025 |
| Q3a | Multi-head mechanism in GPT | What is the multiread mechanism in GPT? (Printed as “multiread” — read as the multi-head attention mechanism.) | 7 | genai-dec2025 |
| Q3b | RNN — architecture & working | Explain the architecture of RNN & how it works? | 8 | genai-dec2025 |
| Q4a | Challenges in Generative AI | What are the challenges in Generative AI? | 8 | genai-dec2025 |
| Q4b | GPT architecture & working | Write a short note on the architecture and working of GPT models. | 7 | genai-dec2025 |
| Q5a | Chat-GPT | What is chat-GPT and how does it work? | 7 | genai-dec2025 |
| Q5b | LLM — Large Language Models | What is LLM and Types of LLM explain? | 8 | genai-dec2025 |
| Q6a | Ethics in Generative AI | Ethical concerns related to Generative AI. | 5 | genai-dec2025 |
| Q6b | Importance in five real-world domains | Explain the importance of Generative AI in any five real-world domains. | 5 | genai-dec2025 |
| Q6c | Traditional vs modern language models | Differentiate between traditional language models and modern language models. | 5 | genai-dec2025 |
| Q6d | Self-attention in GPT | Explain the importance and role of self-attention in GPT models. | 5 | genai-dec2025 |
| Q1a | Generative AI — definition & scope | Define Generative AI. | 2 | genai-may2026 |
| Q1b | GPT and Chat-GPT — definition | What is GPT and Chat-GPT? | 2 | genai-may2026 |
| Q1c | LLM — Large Language Models | Define LLM. | 2 | genai-may2026 |
| Q1d | Pre-training | What is pre-training? | 2 | genai-may2026 |
| Q1e | LSTM | What is LSTM? | 2 | genai-may2026 |
| Q1f | Tokenization in transformers | What is tokenization in transformer model? | 2 | genai-may2026 |
| Q2a | CNN — architecture & working | Explain the architecture of CNN & how it works? | 5 | genai-may2026 |
| Q2b | GPT architecture & working | Explain architecture and working of GPT models. | 5 | genai-may2026 |
| Q3a | Ethics in Generative AI | Explain ethical issues in Generative AI. | 5 | genai-may2026 |
| Q3b | GPT variants | Explain different GPT variants and their applications. | 5 | genai-may2026 |
| Q4a | Transformer architecture | Explain the architecture and working of Transformer models. | 5 | genai-may2026 |
| Q4b | Generative AI tools / applications | Explain any five Generative AI tools with examples. | 5 | genai-may2026 |
| Q5a | RNN — architecture & working | Explain RNN and its limitations. | 5 | genai-may2026 |
| Q5b | GPT architecture & working | Explain GPT architecture and working. (Same question as May Q2b — attempt only one.) | 5 | genai-may2026 |