Concept

Penalty Functions Based on Hidden States

Instead of only evaluating the final generated text, a penalty function can be designed to operate on the internal hidden states of a large language model. This approach allows for the assessment and penalization of undesirable properties at the level of the model's internal representations during the generation process.

0

1

Updated 2026-05-05

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