Quiz Space

Introduction to Deep Learning and Generative AI · End Term · 10 May 2026 · January 2026 term · Set 1

Question 6: Consider an attention mechanism defined as:\ Attention(Q,…

Question 6

+4 marksOne correct option

Consider an attention mechanism defined as:
Attention(Q, K, V) = softmax(QKT) V
Which of the following is the most likely consequence of removing the scaling factor in this attention computation?

  1. A

    The attention scores become smaller, leading to uniform attention weights.

  2. B

    The dot product values grow large, causing the softmax to produce extremely peaked distributions and unstable gradients.

  3. C

    The model becomes invariant to the dimensionality of key vectors.

  4. D

    The attention mechanism works as usual.

Show answer

Correct answer

  • B

    The dot product values grow large, causing the softmax to produce extremely peaked distributions and unstable gradients.

See the whole paperTake it as a mock test

Question 6 of 24 in the IIT Madras BS Introduction to Deep Learning and Generative AI (Deep Learning and GenAI) End Term paper sat on 10 May 2026, in the January 2026 term (Introduction To Deep Learning And Generative Ai 06 May 26 (Session 2)). It carries 4 marks.

More questions from this paper

  1. Q1Which of the following is the MOST correct way to run a trained model on test data?
  2. Q2Consider the sentence:\ "Data science is fun"\ In a bigram language model, which probability expression correctly repre…
  3. Q3We are given the Q, K, V matrices to compute the scaled dot product attention matrix for a transformer.\ Keeping everyt…
  4. Q4Consider the following code representing a component of a traditional encoder decoder model which uses Bahdanau attenti…
  5. Q5What is the output of the following code?
  6. Q7Figure question
  7. Q8What is the primary capability that the Position-wise Feed-Forward Network (FFN) provides to the Transformer architectu…
  8. Q9Figure question
  9. Q10A transformer model has the following embedding layer: Vocabulary size = 50,000• Embedding dimension = 768• The embeddi…
  10. Q11Consider the following code: Which of the following describes the output of decoded_texts?
  11. Q12We use the following LORA code to finetune BERT model: But we notice that the code does not work as intended. Which of …
  12. Q13Which of the following are data augmentation techniques?
  13. Q14Consider the following optimization algorithms used for training neural networks: SGD with Momentum• RMSProp• Adam• Whi…
  14. Q15Select the correct statement(s) regarding a CNN:
  15. Q16Which of the following situations may lead to a mode collapse while training a GAN?
  16. Q17Consider the evaluation metrics commonly used for generative models: Fréchet Inception Distance (FID), Inception Score …
  17. Q18A text-to-image generative model is evaluated on three metrics with the following results: FID = 8.5• Inception Score (…
  18. Q19Consider an encoder-decoder architecture trained using maximum likelihood for sequence generation.\ Which of the follow…
  19. Q20You want to reduce parameters in multi-head attention from 4,096 to 1,024. Which of the following changes achieve(s) th…
  20. Q21Consider the following single-layer RNN:
  21. Q22Consider the following PyTorch model: Note: Do not round intermediate values.• Final answer can be rounded to 3 decimal…
  22. Q23Figure question
  23. Q24A model is fine-tuned with the following configuration: dataset_size = 2048• batch_size = 16• epochs = 2• Assuming the …