Learn Before
Concept

Node Expansion in Step-Level Search

Node expansion in step-level search involves using a Large Language Model to generate subsequent reasoning steps for a given partial path ai\mathbf{a}_{\le i}. The LLM produces one or more candidate next steps, denoted as ai+1(1),...,ai+1(M)a_{i+1}^{(1)}, ..., a_{i+1}^{(M)}. Each candidate step is then individually appended to the current path ai\mathbf{a}_{\le i} to create a new, extended potential partial path, formatted as ai+1=(a1,...,ai,ai+1(j))\mathbf{a}_{\le i + 1} = (a_1, ..., a_i, a_{i+1}^{(j)}), which is then evaluated or further expanded.

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