Introduction to Large Language Models Week 5 Answers
Are you looking for Introduction to Large Language Models Week 5 Answers. All weeks of Introduction To Internet Of Things available here.
Table of Contents
Introduction to Large Language Models Week 5 Answers (July-Dec 2025)
Course link: Click here
Question 1. Which of the following best explains the vanishing gradient problem in RNNs?
a) RNNs lack memory mechanisms for long-term dependencies.
b) Gradients grow too large during backpropagation.
c) Gradients shrink exponentially over long sequences.
d) RNNs cannot process variable-length sequences.
Question 2. In an attention mechanism, what does the softmax function ensure?
a) Normalization of decoder outputs
b) Stability of gradients during backpropagation
c) Values lie between -1 and 1
d) Attention weights sum to 1
Question 3. Which of the following is true about the difference between a standard RNN and an LSTM?
a) LSTM does not use any non-linear activation.
b) LSTM has a gating mechanism to control information flow.
c) RNNs have fewer parameters than LSTMs because they use convolution.
d) LSTMs cannot learn long-term dependencies.
Question 4. Which gate in an LSTM is responsible for deciding how much of the cell state to keep?
a) Forget gate
b) Input gate
c) Output gate
d) Cell candidate gate
Question 5. What improvement does attention bring to the basic Seq2Seq model?
a) Reduces training time
b) Removes the need for an encoder
c) Allows access to all encoder states during decoding
d) Reduces the number of model parameters
Question 6. Which of the following is a correct statement about the encoder-decoder architecture?
a) The encoder generates tokens one at a time.
b) The decoder summarizes the input sequence.
c) The decoder generates outputs based on encoder representations and its own prior outputs.
d) The encoder stores only the first token of the sequence.
Question 7. What is self-attention in Transformers used for?
a) To enable sequential computation
b) To attend to the previous layer’s output
c) To relate different positions in the same sequence
d) To enforce fixed-length output
Question 8. Why are RNNs preferred over fixed-window neural models?
a) They have a smaller parameter size.
b) They can process sequences of arbitrary length.
c) They eliminate the need for embedding layers.
d) None of the above.
Question 9. Given the following encoder and decoder hidden states, compute the attention scores. (Use dot product as the scoring function)
Encoder hidden states: h1 = [7, 3], h2 = [0, 2], h3 = [1, 4]
Decoder hidden state: s = [0.2, 1.5]
a) 0.42, 0.02, 0.56
b) 0.15, 0.53, 0.32
c) 0.64, 0.18, 0.18
d) 0.08, 0.91, 0.01
Click here for all nptel assignment answers
These are Introduction to Large Language Models Week 5 Answers