AI

N【Vidar论文研读分享】一种更安全的LLM指令层次结构

今天是世界读书日,我们一起来简短地读一篇轻松惬意的LLM安全相关的文章。来自OpenAI团队,发布于4月19日,非常的热乎。Intro现代大语言模型 (LLM) 不再单纯...

N华中科技大学 | 大模型中‘故障词元’的检测和分类

原文标题:Glitch Tokens in Large Language Models: Categorization Taxonomy and Effective Detection原文作者:Yuxi Li*, Yi Liu*, Gelei Deng, Ying Zhan...

Visual Analysis of Jailbreak Attacks Against Large Language Models

Abstract 抽象 The proliferation of large language models (LLMs) has underscored concerns regarding their security vulnerabilities, notably against ...

Garak – A Generative AI Red-teaming Tool

Exploring 'Red Teaming' for LLMs, we combine technical insights and real-world scout experience to enhance cyber defenses against new vulnerabiliti...

Microsoft’s ‘AI Watchdog’ defends against new LLM jailbreak method

(Credit: monticellllo – stock.adobe.com) (图片来源:monticellllo – stock.adobe.com) Microsoft has discovered a new method to jailbreak larg...

Many-shot jailbreaking

Read the paper 阅读白皮书 We investigated a “jailbreaking” technique — a method that can be used to evade the safety guardrails put in place by t...

AI PC之NPU漫谈

点击蓝字 关注我们AI PC之NPU漫谈前言AI的发展日新月异,形成了一个时代的强大潮流。然而,真正推动这一势头的,是AI算力的持续扩张,特别是向边缘计算设备的...

[IEEE TPAMI 2024] 将CLIP模型转换为端到端文本识别器

本文简要介绍TPAMI 2024录用论文“Turning a CLIP Model into a Scene Text Spotter”的主要工作。这篇文章介绍了一种新方法FastTCM,专注于直接将CLIP 模型用...

利用知识图谱提高RAG应用的准确性

Graph retrieval-augmented generation (GraphRAG) 正在逐渐流行,并成为传统向量搜索检索方法的有力补充。这种方法利用了图数据库的结构化特性,将数据组织...

《大语言模型(LLM)攻防实战手册》第一章:提示词注入(LLM01)- 间接提示词注入

前言 这一节内容比较多,原因是因为间接提示词注入衍生内容比较多,为了内容的完整性,所以文字较多各位看官见谅。另外文章欢迎转载,只是希望能加个微信公众...
1 2 3 24