Short Answer

Analyzing the Weak-to-Strong Objective Function

Consider the training objective for a powerful language model being fine-tuned using labels generated by a less powerful model: maximize Σ log Pr(weak_model_label | input). Explain how this mathematical objective frames the fine-tuning process as a form of knowledge transfer, identifying which model acts as the 'teacher' and which acts as the 'student'.

0

1

Updated 2025-10-06

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

Analysis in Bloom's Taxonomy

Cognitive Psychology

Psychology

Social Science

Empirical Science

Science