Learn Before
Debugging a Generative Model's Selection Process
Based on the principle of sorting candidates by their likelihood, which part of the model's selection process is most likely failing? Explain your reasoning.
0
1
Tags
Ch.5 Inference - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Analysis in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
Pruning and K-Best Output in Beam Search
A language model is generating the next word in a sequence. After considering several possibilities, it has calculated the following probabilities for five potential next tokens. Arrange these tokens in the correct order, from the most probable to the least probable, as would be done during the ranking stage of a search process.
Debugging a Generative Model's Selection Process
A language model is generating a sequence and has produced the following potential next tokens with their associated probabilities:
{'on': 0.32, 'cute': 0.34, 'sick': 0.21, '.': 0.01, 'are': 0.12}. Which of the following lists represents the correctly sorted output of the ranking stage for these candidates?