Learn Before
Concept

Scaling the Search Space in Search Scaling

Scaling the search space is a key aspect of search scaling that involves expanding the set of candidate output sequences considered during the decoding process to improve the chances of discovering higher-quality outputs. A simple example of this is increasing the beam width in beam search, which allows more candidate sequences to be explored in parallel at each decoding step. This is particularly useful in tasks where the optimal solution is not immediately apparent from local, step-by-step decisions.

0

1

Updated 2026-05-06

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