Scholarlink

<p>The increase in popularity of Large Language Models (LLMs) has given rise to concerns about their vulnerabilities and security risks that make them open to exploitation and misuse. Their widespread adoption prompts the need for further exploration and mitigation of these potential dangers. Here, we adopt a comprehensive framework for examining LLM vulnerabilities centered on three fundamental criteria: reliability, resiliency, and responsibility. These criteria serve as the foundation for ensuring LLMs are reliable in real-world applications, resilient in the face of potential threats, and are responsibly deployed to uphold ethical and societal standards. By utilizing this framework, we aim to provide a structured and systematic approach for understanding, developing, and deploying trustworthy LLMs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Reliability, Resiliency, and Responsibility: A Framework for Addressing Security Concerns in Large Language Models

  • 期刊论文

  • 发表日期：

    2025年3月5日

查看全文
  • Linsey Passarella,
  • Edmon Begoli,
  • Colin Smith,
  • Amir Sadovnik

摘要

The increase in popularity of Large Language Models (LLMs) has given rise to concerns about their vulnerabilities and security risks that make them open to exploitation and misuse. Their widespread adoption prompts the need for further exploration and mitigation of these potential dangers. Here, we adopt a comprehensive framework for examining LLM vulnerabilities centered on three fundamental criteria: reliability, resiliency, and responsibility. These criteria serve as the foundation for ensuring LLMs are reliable in real-world applications, resilient in the face of potential threats, and are responsibly deployed to uphold ethical and societal standards. By utilizing this framework, we aim to provide a structured and systematic approach for understanding, developing, and deploying trustworthy LLMs.