Socio-Culturally Aware Evaluation Framework for LLM-Based Content Moderation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Kumar, Shanu, Kholkar, Gauri, Mendke, Saish, Sadana, Anubhav, Agrawal, Parag, Dandapat, Sandipan
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866929636688003072
author Kumar, Shanu
Kholkar, Gauri
Mendke, Saish
Sadana, Anubhav
Agrawal, Parag
Dandapat, Sandipan
author_facet Kumar, Shanu
Kholkar, Gauri
Mendke, Saish
Sadana, Anubhav
Agrawal, Parag
Dandapat, Sandipan
contents With the growth of social media and large language models, content moderation has become crucial. Many existing datasets lack adequate representation of different groups, resulting in unreliable assessments. To tackle this, we propose a socio-culturally aware evaluation framework for LLM-driven content moderation and introduce a scalable method for creating diverse datasets using persona-based generation. Our analysis reveals that these datasets provide broader perspectives and pose greater challenges for LLMs than diversity-focused generation methods without personas. This challenge is especially pronounced in smaller LLMs, emphasizing the difficulties they encounter in moderating such diverse content.
format Preprint
id arxiv_https___arxiv_org_abs_2412_13578
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Socio-Culturally Aware Evaluation Framework for LLM-Based Content Moderation
Kumar, Shanu
Kholkar, Gauri
Mendke, Saish
Sadana, Anubhav
Agrawal, Parag
Dandapat, Sandipan
Computation and Language
Artificial Intelligence
With the growth of social media and large language models, content moderation has become crucial. Many existing datasets lack adequate representation of different groups, resulting in unreliable assessments. To tackle this, we propose a socio-culturally aware evaluation framework for LLM-driven content moderation and introduce a scalable method for creating diverse datasets using persona-based generation. Our analysis reveals that these datasets provide broader perspectives and pose greater challenges for LLMs than diversity-focused generation methods without personas. This challenge is especially pronounced in smaller LLMs, emphasizing the difficulties they encounter in moderating such diverse content.
title Socio-Culturally Aware Evaluation Framework for LLM-Based Content Moderation
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2412.13578