Essay

Critique of the Last-Token Reward Calculation Method

A common method for calculating a reward score from a language model involves applying a linear transformation to the hidden state vector corresponding only to the final token of a given text sequence. Critically evaluate this approach. Discuss one significant advantage and one significant disadvantage of using only the final token's representation for this purpose.

0

1

Updated 2025-10-03

Contributors are:

Who are from:

Tags

Ch.4 Alignment - Foundations of Large Language Models

Foundations of Large Language Models

Foundations of Large Language Models Course

Computing Sciences

Evaluation in Bloom's Taxonomy

Cognitive Psychology

Psychology

Social Science

Empirical Science

Science