Link: http://arxiv.org/abs/2412.02113v1
PDF Link: http://arxiv.org/pdf/2412.02113v1
Summary: In recent years, Large Language Models (LLMs) have garnered considerableattention for their remarkable abilities in natural language processing tasks.
However, their widespread adoption has raised concerns pertaining to trust andsafety.
This systematic review investigates the current research landscape ontrust and safety in LLMs, with a particular focus on the novel application ofLLMs within the field of Trust and Safety itself.
We delve into thecomplexities of utilizing LLMs in domains where maintaining trust and safety isparamount, offering a consolidated perspective on this emerging trend.
\ By synthesizing findings from various studies, we identify key challenges andpotential solutions, aiming to benefit researchers and practitioners seeking tounderstand the nuanced interplay between LLMs and Trust and Safety.
This review provides insights on best practices for using LLMs in Trust andSafety, and explores emerging risks such as prompt injection and jailbreakattacks.
Ultimately, this study contributes to a deeper understanding of howLLMs can be effectively and responsibly utilized to enhance trust and safety inthe digital realm.
Published on arXiv on: 2024-12-03T03:10:12Z