46. Which of the following is true about the “few-shot” learning ability of GPT models?
A) It can learn to perform a task accurately with very few examples
B) It can learn to perform a task without any examples
C) It can learn to perform a task with less data than other models
D) It cannot perform few-shot learning
47. Which of the following is an advantage of GPT models for natural language processing tasks?
A) They require very little computational resources
B) They do not require any pre-processing of input data
C) They can learn complex relationships between words and sentences
D) They have a limited vocabulary
48. What is the “prompt engineering” technique used in GPT models?
A) Creating a specific set of input data for the model to learn from
B) Manipulating the input text to guide the model’s generation
C) Controlling the model’s generation by adjusting the probability distribution of generated text
D) None of the above
49. Which of the following is a disadvantage of GPT models for natural language processing tasks?
A) They have a limited understanding of common sense knowledge
B) They require very little training data
C) They are highly interpretable
D) They cannot generate text in multiple languages
50. What is the primary advantage of GPT models over traditional rule-based approaches for natural language processing tasks?
A) They require less computational resources
B) They do not require any pre-processing of input data
C) They can learn to perform complex tasks without explicit instructions
D) They have a limited vocabulary
