Bird
0
0

Why does beam search decoding not guarantee finding the globally optimal sequence?

hard📝 Conceptual Q10 of 15
NLP - Text Generation
Why does beam search decoding not guarantee finding the globally optimal sequence?
ABecause it always selects the longest sequence
BBecause it uses random sampling instead of scoring
CBecause it only keeps a fixed number of candidates, potentially discarding the best path early
DBecause it ignores the model's probability scores
Step-by-Step Solution
Solution:

  1. Step 1: Understand beam search limitation

    Beam search keeps only a fixed number of top candidates at each step, which may discard the best sequence early.

  2. Step 2: Contrast with exhaustive search

    Unlike exhaustive search, beam search trades completeness for efficiency, so global optimum is not guaranteed.

  3. Final Answer:

    Because it only keeps a fixed number of candidates, potentially discarding the best path early -> Option C

  4. Quick Check:

    Beam search limits candidates, so no global guarantee [OK]
Quick Trick: Beam search prunes candidates, so best path may be lost [OK]
Common Mistakes:
MISTAKES
  • Thinking beam search uses random sampling
  • Assuming beam search always picks longest sequence
  • Believing beam search ignores scores

Want More Practice?

15+ quiz questions · All difficulty levels · Free

Free Signup - Practice All Questions
More NLP Quizzes
Sentiment Analysis Advanced - Hybrid approaches - Quiz 14medium Sentiment Analysis Advanced - Why advanced sentiment handles nuance - Quiz 7medium Sequence Models for NLP - Embedding layer usage - Quiz 10hard Sequence Models for NLP - Bidirectional LSTM - Quiz 7medium Text Generation - Why text generation creates content - Quiz 8hard Text Generation - Temperature and sampling - Quiz 10hard Text Similarity and Search - Why similarity measures find related text - Quiz 10hard Text Similarity and Search - Why similarity measures find related text - Quiz 3easy Topic Modeling - LDA with scikit-learn - Quiz 11easy Topic Modeling - Why topic modeling discovers themes - Quiz 1easy