Selecting a Generation Strategy for a Real-Time Application
A company is building a live chatbot for customer support. The chatbot must provide helpful and coherent responses to user queries in real-time. The development team is considering two fundamental strategies for generating the chatbot's replies:
0
1
Tags
Ch.5 Inference - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Evaluation in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
Formula for a Candidate Solution Set
A developer is building a text generation system to find the single most probable 20-word summary of a given document. Their proposed method is to generate every possible 20-word sequence, calculate the probability of each one, and then select the sequence with the highest probability. Which statement best analyzes the fundamental limitation of this approach?
Selecting a Generation Strategy for a Real-Time Application
The Role of Search Algorithms in Text Generation
An engineer is developing a system to generate a 10-word sentence. They are comparing two approaches:
- Approach 1: Systematically generate every possible 10-word sequence from the system's vocabulary, calculate a probability score for each, and select the one with the highest score.
- Approach 2: Generate the first word, then, given that first word, generate the most probable second word, and so on, making a locally optimal choice at each of the 10 steps.
Which statement best analyzes the relationship between these two approaches?
Trade-offs of Greedy Search in Inference