AI Knowledge Base

在自然语言处理(NLP)中,意图指用户表达某种目的,想做或想要的事情。通过分析用户发出的信息,识别其背后的意图,从而回复相关的内容。例如，“订餐”、“查看天气”、“我想去巴黎”等都是有效的意图。 为了...

1. 引言 在人工智能领域，Multi Agent 系统正逐渐成为解决复杂问题、实现高效协作的关键技术。CrewAI 作为一款强大的多 Agent 协作工具，为开发者提供了便捷的方式来构建智能协作系统...

原文：https://arxiv.org/pdf/2412.15479 解读：此文章本身创新不多，应用意义也不大。不过，让我想起很久很久以前阅读的三篇高含金量的文章，阅读此文结合之前的三篇文章，希望给...

原文：https://arxiv.org/pdf/2309.04269 速读：《从稀疏到密集：使用链式密度提示的GPT-4摘要生成》 收录到：总结知识常用prompts 摘要 在自动文本摘要中，确...

发现一篇好玩的论文《Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs》，主题是分析o1 类推理模型 思考路径频...

速读 智能体记忆的挑战与Zep的创新 智能体（AI Agents）在复杂任务中面临记忆瓶颈。传统基于大型语言模型（LLM）的智能体受限于上下文窗口，难以有效整合长期对话历史和动态数据，限制了表现并易产...

《Titans: Learning to Memorize at Test Time》原文：https://arxiv.org/pdf/2501.00663v1 Titans 架构 非官方实现：htt...

OpenAI 的 Deep Research 工具横空出世后，各大厂商都推出了自研的 Deep Research 工具。所谓Deep Research，是和普通搜索比较的，简单的RAG检索生成一般只有...

摘要 尽管大型语言模型（LLMs）表现出色，但它们容易产生幻觉并生成事实不准确的信息。这一挑战促使人们努力进行属性文本生成，促使LLMs生成包含支持证据的内容。在本文中，我们提出了一种名为Think...

如果你的 RAG 应用未能达到预期效果，或许是时候重新审视你的分块策略了。更好的分块意味着更精准的检索，最终带来更高质量的回复。 然而，分块技术并非“一招鲜吃遍天”，没有哪一种方法是绝对最优的。你需要...

在人工智能浪潮的中心，AI Agent (智能体) 正以惊人的速度进化，如同科幻电影中走出的智能助手，悄然渗透进企业的每一个角落。它们不再是遥不可及的未来概念，而是企业提升效率、优化流程、决胜市场的秘...

The field of generative AI is currently evolving rapidly, with new frameworks and technologies emerging. Therefore, readers need to be aware that the content presented in this paper may be time-sensitive. In this paper, we will take an in-depth look at two of the leading frameworks for building LLM applications: LangChain and LangGr...

It's only February, and Deep Search is already looming as the new search standard for 2025. Giants like Google and OpenAI have unveiled their "Deep Research" products in an effort to capitalize on this wave of technology...

导读 本文档旨在通过一系列提示词示例（部分），帮助读者快速理解并掌握提示词工程的核心概念与应用。这些示例均源自于一篇关于提示工程技术系统综述的学术论文（《The Prompt Report: A Sy...

研究人员研究了一种“越狱攻击”技术——这种方法可以被用来绕过大语言模型 (LLM) 开发者设置的安全护栏。这项技术被称为“多样本越狱攻击”，在 Anthropic 自身模型以及其他 AI 公司生产的模...

The Python ecosystem has never been short of package management and environment management tools, from the classic pip and virtualenv to pip-tools and conda to the modern Poetry, PDM, and so on. Each tool has its area of specialization, but often...

As the capabilities of large-scale language models (LLMs) are rapidly evolving, traditional benchmark tests, such as MMLU, are showing limitations in distinguishing top models. Relying on knowledge quizzes or standardized tests alone, it has become difficult to fully measure the nuanced capabilities of models that are crucial in real-world interactions, such as...

In the age of AI-assisted programming, we want AI to generate code that is not just static text, but can be parsed, edited, previewed, and even executed. This demand has given rise to a new interaction paradigm, Artifact. In this article, we'll go from theoretical concepts to practical realities...

Recently, when I was choosing a data processing tool for the RAG knowledge base for my smart customer service project, I took a fresh look at the current mainstream document processing projects, including olmOCR, Marker, MinerU, Docling, Markitdown, Llamaparse...