Learn Before
Best-of-N Sampling
Best-of-N sampling is a method used to improve model outputs by generating multiple (N) potential solutions for a given input. From this set of N outputs, the best one is selected using a specific mechanism, such as a verifier model or an evaluation function. This technique leverages a model's ability to generate diverse responses to find a higher-quality solution than what might be produced in a single attempt.
0
1
Tags
Ch.5 Inference - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Related
Self-Consistency as a Minimum Bayes Risk Search Process
Framing Answer Selection as a Search Problem
An LLM generates five different step-by-step solutions to a complex algebra problem. A separate verification model then evaluates each solution by checking if the final answer is correct and if each intermediate step logically follows from the previous one. The solution with the highest score from the verifier is chosen as the final output. Match the components of this process, when framed as a search problem, to their correct descriptions.
Analyzing Code Generation as a Search Problem
Best-of-N Sampling
Best-of-N Sampling (Parallel Scaling)
Search Algorithm for Solution Selection
Learn After
Evaluating a Chatbot Response Generation Strategy
A team is building a system to generate highly accurate legal document summaries. They decide to generate 20 different summary options for each document and then use a sophisticated, but slow, verification model to score and select the best one. Which of the following best describes the primary trade-off inherent in this strategy?
Deconstructing a Response Generation and Selection Process