Concept

Computational Infeasibility of Exhaustive Search in LLM Decoding

The complete search space for potential output sequences in LLMs grows exponentially with the length of the generated text. This rapid expansion makes a direct or exhaustive search for the optimal sequence computationally infeasible for all but the shortest outputs.

0

1

Updated 2025-10-07

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