Practice¶
Full mock exams matching the format of the real paper: 24 multiple-choice questions across the four sections, 4 marks each, 90 minutes, no negative marking.
How to take a mock
- Set a 90-minute timer. No pausing, no reference material.
- Answer every question — no negative marking means blanks are strictly worse than guesses.
- Use scratch paper for the numerical questions (PQ math, top-k∩top-p, Levenshtein). Don't trust intuition on arithmetic.
- Flag hard questions and return to them after Q24.
- Grade honestly. For every wrong answer, reread the corresponding Learn guide block before your next mock.
Available mocks¶
| Mock | Coverage | Difficulty | Length |
|---|---|---|---|
| Mock Exam 1 | All 4 sections, 24 questions | Matches sample paper | 90 min |
Section weighting (as tested)¶
The exam is roughly balanced across four sections. Each has 6 questions × 4 marks = 24 marks:
| Section | Coverage | Lectures |
|---|---|---|
| 1 | ML foundations & model architecture | 1, 2 |
| 2 | LLM decoding, APIs, tooling, evaluation, safety | 4, 5, 6 |
| 3 | Prompting techniques & NLP metrics | 7, 8, 5 |
| 4 | Vector indexing & RAG security | 11, 12 |
Where's the sample paper?
The official upGrad × IIT Kharagpur sample paper is copyrighted and cannot be redistributed on this site. If you're enrolled in the course, download it from your course dashboard. This mock exam is an original replica in the same format and coverage.
After you finish a mock¶
Grade the whole thing before looking at any explanations. Then for each question:
- Got it right, confident: Move on.
- Got it right, guessed: Read the answer explanation — that's a knowledge gap disguised as a win.
- Got it wrong: Read the answer explanation, then open the linked Learn guide section and reread it.
Track your weakest section. Spend the next study block there.