Can I read Exposing LLM Vulnerabilities to Bias on EtoBox?

Exposing LLM Vulnerabilities to Bias by jcalberto.guerrero is a document available to read on EtoBox.

What is Exposing LLM Vulnerabilities to Bias about?

This document discusses the vulnerabilities of large language models (LLMs) to malicious prompts disguised as scientific language, revealing that these models can be manipulated to produce biased and toxic outputs. Experiments demonstrate that LLMs, including state-of-the-art models, can be persuaded to generate harmful stereotypes by misinterpreting social science studies, raising ethical concerns about their deployment. The findings highlight the need for improved safety measures and careful consideration

Author

jcalberto.guerrero

Language

EN