Case Study

Debugging an Input Composition Method

A researcher is adapting a large language model for a new task. They are constructing the input sequence by first taking the user's text and converting it into its standard, fixed token embeddings. They then append a set of newly introduced, trainable vectors to the end of this sequence. Despite training, the model's performance is poor. Based on the standard method for composing inputs in this context, identify the fundamental error in the researcher's approach and explain the correct procedure.

0

1

Updated 2025-10-09

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

Application in Bloom's Taxonomy

Cognitive Psychology

Psychology

Social Science

Empirical Science

Science