Concept

Limitation of Relying on Human-Crafted Inputs for Synthetic Data Generation

A key drawback of generating fine-tuning data with an LLM is its dependence on human-created or collected inputs. These inputs may lack the diversity needed to ensure the model generalizes well to the broad range of real-world user queries, which are often not covered in existing NLP datasets.

0

1

Updated 2026-05-01

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