Quiz Space

Large Language Models · End Term · 21 Dec 2025 · September 2025 term · Set 1

Question 5: A research team wants to classify scientific abstracts in…

Question 5

+2 marksOne correct option

A research team wants to classify scientific abstracts into multiple topics simultaneously (e.g., ''ML'', ''biology'', ''statistics''), where each abstract may belong to more than one topic. They fine-tune BERT for this task. Which modification is MOST appropriate?

  1. A

    Replace the [CLS] vector with an average of the top-4 attention heads

  2. B

    Feed the [CLS] embedding into a linear layer with a sigmoid activation per label

  3. C

    Use token embeddings individually and classify each token

  4. D

    Use the [SEP] token embedding for multi-label prediction

Show answer

Correct answer

  • B

    Feed the [CLS] embedding into a linear layer with a sigmoid activation per label

See the whole paperTake it as a mock test

Question 5 of 24 in the IIT Madras BS Large Language Models (LLM) End Term paper sat on 21 Dec 2025, in the September 2025 term (Large Language Models 21 Dec 25). It carries 2 marks.

More questions from this paper

  1. Q1What happens when you use Batch Normalization with a batch size of 1?
  2. Q2A language model outputs the following logits for the next token: {cat: 3.2, dog: 2.9, bird: 1.1, fish: 0.5, snake: -0.…
  3. Q3Figure question
  4. Q4Figure question
  5. Q6A company wants to automatically correct noisy OCR text extracted from scanned documents. The text contains spelling mi…
  6. Q7Figure question
  7. Q8During autoregressive inference, key-value (KV) caching is used to avoid recomputing keys and values for previously gen…
  8. Q9Figure question
  9. Q10Figure question
  10. Q11If a model uses Absolute Positional Embeddings (APE), increasing the sequence length from 100 to 1000 requires:
  11. Q12Figure question
  12. Q13Figure question
  13. Q14Which of the following methods uses/use the concept: "The farther apart two tokens are, the less they should attend to …
  14. Q15You trained a model with max sequence length 512. At inference, you need to process sequence length 4096 without retrai…
  15. Q16Figure question
  16. Q17Figure question
  17. Q18Figure question
  18. Q19Figure question
  19. Q20Figure question
  20. Q21Figure question
  21. Q22A GPT-style causal language model is trained using the next-token prediction objective: During inference, the model mus…
  22. Q23Under which of the following decoding settings can the model produce different outputs across multiple runs on the same…
  23. Q24Figure question