Short Answer

Impact of Dataset Quality on Fine-Tuning

An AI development team is preparing a dataset to align a pre-trained language model to be a helpful and harmless assistant. Their dataset consists of thousands of prompt-response pairs. However, a quality check reveals that a significant portion of the responses are suboptimal: some are factually incorrect, others are rude, and a few are overly verbose. If the team proceeds to fine-tune the model on this raw dataset as is, what is the most likely outcome for the model's behavior after training? Explain your reasoning.

0

1

Updated 2025-10-05

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