Can I read Dataset Labeling for Hallucination Detection on EtoBox?

Dataset Labeling for Hallucination Detection by vishwasvm437 is a document available to read on EtoBox.

What is Dataset Labeling for Hallucination Detection about?

The document outlines a dataset labeling project aimed at improving hallucination detection in language models. It defines hallucinations as factual inaccuracies or fabrications in model outputs compared to provided context and establishes a taxonomy for labeling these inaccuracies. The labeling process involves assigning one of three labels (Hallucination, Not Hallucination, Ambiguous) and providing reasoning from a predefined list, with examples illustrating the criteria for labeling.

Author

vishwasvm437

Language

EN