Short Answer

Correcting a Step in a Generation Process

In a text generation process, a fast draft model proposes the token sequence ['over', 'the', 'lazy']. A more accurate evaluation model accepts the first two tokens, 'over' and 'the', but rejects the third token, 'lazy'. To continue generating the text, which model should be used to determine the very next token, and based on what preceding text sequence?

0

1

Updated 2025-10-04

Contributors are:

Who are from:

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