Case Study

Analyzing a Positional Encoding Modification

A team is working with a language model originally trained on text sequences up to 2048 tokens. To adapt it for documents up to 4096 tokens, an engineer modifies the positional encoding functions by decreasing the period of each function. Analyze the likely outcome of this specific modification. Will the model successfully handle the longer sequences? Explain your reasoning based on how the period of the encoding functions relates to the range of positions the model can represent.

0

1

Updated 2025-10-02

Contributors are:

Who are from:

Tags

Ch.3 Prompting - 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