[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-llmgenai--LLMInterviewQuestions":3,"tool-llmgenai--LLMInterviewQuestions":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":76,"owner_location":79,"owner_email":78,"owner_twitter":78,"owner_website":80,"owner_url":81,"languages":78,"stars":82,"forks":83,"last_commit_at":84,"license":78,"difficulty_score":85,"env_os":86,"env_gpu":87,"env_ram":87,"env_deps":88,"category_tags":91,"github_topics":78,"view_count":23,"oss_zip_url":78,"oss_zip_packed_at":78,"status":16,"created_at":92,"updated_at":93,"faqs":94,"releases":95},2934,"llmgenai\u002FLLMInterviewQuestions","LLMInterviewQuestions","This repository contains LLM (Large language model) interview question asked in top companies like Google, Nvidia , Meta , Microsoft & fortune 500 companies.","LLMInterviewQuestions 是一个专为大型语言模型（LLM）求职者打造的开源面试题库，汇集了来自谷歌、英伟达、Meta、微软等顶尖科技公司及财富 500 强企业的真实面试题。面对当前 LLM 领域技术迭代快、面试考察维度广的痛点，它系统性地整理了超过 100 道高质量问题，帮助候选人打破信息差，高效备战技术面试。\n\n该资源特别适合正在寻求 LLM 工程师、算法研究员或 AI 应用开发岗位的开发者与研究人员使用。其核心亮点在于内容覆盖全面且紧贴实战，将复杂的知识体系划分为提示工程、检索增强生成（RAG）、向量数据库原理、模型微调、偏好对齐（RLHF\u002FDPO）、幻觉控制及智能体系统等 15 个关键类别。从基础的 Token 概念到生产级的文档切片策略，再到具体的架构设计案例，LLMInterviewQuestions 不仅提供了问题清单，更引导用户深入理解工业界解决实际问题的思路，是提升大模型领域专业竞争力的实用指南。","# 100+ LLM Interview Questions for Top Companies\n\nThis repository contains over 100+ interview questions for Large Language Models (LLM) used by top companies like Google, NVIDIA, Meta, Microsoft, and Fortune 500 companies. Explore questions curated with insights from real-world scenarios, organized into 15 categories to facilitate learning and preparation.\n\n---\n\n#### You're not alone—many learners have been reaching out for detailed explanations and resources to level up their prep.\n\n#### You can find answers here, visit [Mastering LLM](https:\u002F\u002Fwww.masteringllm.com\u002Fcourse\u002Fllm-interview-questions-and-answers?previouspage=allcourses&isenrolled=no#\u002Fhome).\n#### Use the code `LLM50` at checkout to get **50% off**\n\n---\n\n![Image Description](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fllmgenai_LLMInterviewQuestions_readme_b8aae1309e80.jpg)\n\n---\n## Table of Contents\n\n1. [Prompt Engineering & Basics of LLM](#prompt-engineering--basics-of-llm)\n2. [Retrieval Augmented Generation (RAG)](#retrieval-augmented-generation-rag)\n3. [Document digitization & Chunking](#document-digitization-&-chunking)\n4. [Embedding Models](#embedding-models)\n5. [Internal Working of Vector Databases](#internal-working-of-vector-databases)\n6. [Advanced Search Algorithms](#advanced-search-algorithms)\n7. [Language Models Internal Working](#language-models-internal-working)\n8. [Supervised Fine-Tuning of LLM](#supervised-fine-tuning-of-llm)\n9. [Preference Alignment (RLHF\u002FDPO)](#preference-alignment-rlhfdpo)\n10. [Evaluation of LLM System](#evaluation-of-llm-system)\n11. [Hallucination Control Techniques](#hallucination-control-techniques)\n12. [Deployment of LLM](#deployment-of-llm)\n13. [Agent-Based System](#agent-based-system)\n14. [Prompt Hacking](#prompt-hacking)\n15. [Miscellaneous](#miscellaneous)\n16. [Case Studies](#case-studies)\n\n---\n\n## Prompt Engineering & Basics of LLM\n\n- **What is the difference between Predictive\u002FDiscriminative AI and Generative AI?**\n- **What is LLM, and how are LLMs trained?**\n- **What is a token in the language model?**\n- **How to estimate the cost of running SaaS-based and Open Source LLM models?**\n- **Explain the Temperature parameter and how to set it.**\n- **What are different decoding strategies for picking output tokens?**\n- **What are different ways you can define stopping criteria in large language model?**\n- **How to use stop sequences in LLMs?**\n- **Explain the basic structure prompt engineering.**\n- **Explain in-context learning**\n- **Explain type of prompt engineering**\n- **What are some of the aspect to keep in mind while using few-shots prompting?**\n- **What are certain strategies to write good prompt?**\n- **What is hallucination, and how can it be controlled using prompt engineering?**\n- **How to improve the reasoning ability of LLM through prompt engineering?**\n- **How to improve LLM reasoning if your COT prompt fails?**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Retrieval Augmented Generation (RAG)\n\n- **how to increase accuracy, and reliability & make answers verifiable in LLM**\n- **How does RAG work?**\n- **What are some benefits of using the RAG system?**\n- **When should I use Fine-tuning instead of RAG?**\n- **What are the architecture patterns for customizing LLM with proprietary data?**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Document digitization & Chunking \n\n- **What is chunking, and why do we chunk our data?**\n- **What factors influence chunk size?**\n- **What are the different types of chunking methods?**\n- **How to find the ideal chunk size?**\n- **What is the best method to digitize and chunk complex documents like annual reports?**\n- **How to handle tables during chunking?**\n- **How do you handle very large table for better retrieval?**\n- **How to handle list item during chunking?**\n- **How do you build production grade document processing and indexing pipeline?**\n- **How to handle graphs & charts in RAG**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Embedding Models\n\n- **What are vector embeddings, and what is an embedding model?**\n- **How is an embedding model used in the context of LLM applications?**\n- **What is the difference between embedding short and long content?**\n- **How to benchmark embedding models on your data?**\n- **Suppose you are working with an open AI embedding model, after benchmarking accuracy is coming low, how would you further improve the accuracy of embedding the search model?**\n- **Walk me through steps of improving sentence transformer model used for embedding?**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Internal Working of Vector Databases\n\n- **What is a vector database?**\n- **How does a vector database differ from traditional databases?**\n- **How does a vector database work?**\n- **Explain difference between vector index, vector DB & vector plugins?**\n- **You are working on a project that involves a small dataset of customer reviews. Your task is to find similar reviews in the dataset. The priority is to achieve perfect accuracy in finding the most similar reviews, and the speed of the search is not a primary concern. Which search strategy would you choose and why?**\n- **Explain vector search strategies like clustering and Locality-Sensitive Hashing.**\n- **How does clustering reduce search space? When does it fail and how can we mitigate these failures?**\n- **Explain Random projection index?**\n- **Explain Locality-sensitive hashing (LHS) indexing method?**\n- **Explain product quantization (PQ) indexing method?**\n- **Compare different Vector index and given a scenario, which vector index you would use for a project?**\n- **How would you decide ideal search similarity metrics for the use case?**\n- **Explain different types and challenges associated with filtering in vector DB?**\n- **How to decide the best vector database for your needs?**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Advanced Search Algorithms\n\n- **What are architecture patterns for information retrieval & semantic search?**\n- **Why it’s important to have very good search**\n- **How can you achieve efficient and accurate search results in large-scale datasets?**\n- **Consider a scenario where a client has already built a RAG-based system that is not giving accurate results, upon investigation you find out that the retrieval system is not accurate, what steps you will take to improve it?**\n- **Explain the keyword-based retrieval method**\n- **How to fine-tune re-ranking models?**\n- **Explain most common metric used in information retrieval and when it fails?**\n- **If you were to create an algorithm for a Quora-like question-answering system, with the objective of ensuring users find the most pertinent answers as quickly as possible, which evaluation metric would you choose to assess the effectiveness of your system?**\n- **I have a recommendation system, which metric should I use to evaluate the system?**\n- **Compare different information retrieval metrics and which one to use when?**\n- **How does hybrid search works?**\n- **If you have search results from multiple methods, how would you merge and homogenize the rankings into a single result set?**\n- **How to handle multi-hop\u002Fmultifaceted queries?**\n- **What are different techniques to be used to improved retrieval?**\n  \n\n[Back to Top](#table-of-contents)\n\n---\n\n## Language Models Internal Working\n\n- **Can you provide a detailed explanation of the concept of self-attention?**\n- **Explain the disadvantages of the self-attention mechanism and how can you overcome it.**\n- **What is positional encoding?**\n- **Explain Transformer architecture in detail.**\n- **What are some of the advantages of using a transformer instead of LSTM?**\n- **What is the difference between local attention and global attention?**\n- **What makes transformers heavy on computation and memory, and how can we address this?**\n- **How can you increase the context length of an LLM?**\n- **If I have a vocabulary of 100K words\u002Ftokens, how can I optimize transformer architecture?**\n- **A large vocabulary can cause computation issues and a small vocabulary can cause OOV issues, what approach you would use to find the best balance of vocabulary?**\n- **Explain different types of LLM architecture and which type of architecture is best for which task?**\n\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Supervised Fine-Tuning of LLM\n\n- **What is fine-tuning, and why is it needed?**\n- **Which scenario do we need to fine-tune LLM?**\n- **How to make the decision of fine-tuning?**\n- **How do you improve the model to answer only if there is sufficient context for doing so?**\n- **How to create fine-tuning datasets for Q&A?**\n- **How to set hyperparameters for fine-tuning?**\n- **How to estimate infrastructure requirements for fine-tuning LLM?**\n- **How do you fine-tune LLM on consumer hardware?**\n- **What are the different categories of the PEFT method?**\n- **What is catastrophic forgetting in LLMs?**\n- **What are different re-parameterized methods for fine-tuning?**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Preference Alignment (RLHF\u002FDPO)\n\n- **At which stage you will decide to go for the Preference alignment type of method rather than SFT?**\n- **What is RLHF, and how is it used?**\n- **What is the reward hacking issue in RLHF?**\n- **Explain different preference alignment methods.**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Evaluation of LLM System\n\n- **How do you evaluate the best LLM model for your use case?**\n- **How to evaluate RAG-based systems?**\n- **What are different metrics for evaluating LLMs?**\n- **Explain the Chain of Verification.**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Hallucination Control Techniques\n\n- **What are different forms of hallucinations?**\n- **How to control hallucinations at various levels?**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Deployment of LLM\n\n- **Why does quantization not decrease the accuracy of LLM?**\n- **What are the techniques by which you can optimize the inference of LLM for higher throughput?**\n- **How to accelerate response time of model without attention approximation like group query attention?**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Agent-Based System\n\n- **Explain the basic concepts of an agent and the types of strategies available to implement agents**\n- **Why do we need agents and what are some common strategies to implement agents?**\n- **Explain ReAct prompting with a code example and its advantages**\n- **Explain Plan and Execute prompting strategy**\n- **Explain OpenAI functions strategy with code examples**\n- **Explain the difference between OpenAI functions vs LangChain Agents**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Prompt Hacking\n\n- **What is prompt hacking and why should we bother about it?**\n- **What are the different types of prompt hacking?**\n- **What are the different defense tactics from prompt hacking?**\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Miscellaneous\n\n- **How to optimize cost of overall LLM System?**\n- **What are mixture of expert models (MoE)?**\n- **How to build production grade RAG system, explain each component in detail ?**\n- **What is FP8 variable and what are its advantages of it**\n- **How to train LLM with low precision training without compromising on accuracy ?**\n- **How to calculate size of KV cache**\n- **Explain dimension of each layer in multi headed transformation attention block**\n- **How do you make sure that attention layer focuses on the right part of the input?**\n\n\n[Back to Top](#table-of-contents)\n\n---\n\n## Case Studies\n\n- **Case Study 1**: LLM Chat Assistant with dynamic context based on query\n- **Case Study 2**: Prompting Techniques\n\n[Back to Top](#table-of-contents)\n\n---\n\n*For answers for those questions please, visit [Mastering LLM](https:\u002F\u002Fwww.masteringllm.com\u002Fcourse\u002Fllm-interview-questions-and-answers?previouspage=allcourses&isenrolled=no#\u002Fhome).*\n\n","# 100+ 家顶级公司的大语言模型面试题\n\n本仓库包含100多道由谷歌、英伟达、Meta、微软以及财富500强等顶级公司使用的大语言模型（LLM）相关面试题。这些题目基于真实场景精心挑选，分为15个类别，便于学习和备考。\n\n---\n\n#### 你并不孤单——许多学习者都在寻求详细的解答和资源，以提升自己的备考水平。\n\n#### 你可以在这里找到答案，访问 [Mastering LLM](https:\u002F\u002Fwww.masteringllm.com\u002Fcourse\u002Fllm-interview-questions-and-answers?previouspage=allcourses&isenrolled=no#\u002Fhome)。\n#### 结账时使用优惠码 `LLM50`，即可享受**50%折扣**。\n\n---\n\n![图片描述](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fllmgenai_LLMInterviewQuestions_readme_b8aae1309e80.jpg)\n\n---\n## 目录\n\n1. [提示工程与大语言模型基础](#prompt-engineering--basics-of-llm)\n2. [检索增强生成（RAG）](#retrieval-augmented-generation-rag)\n3. [文档数字化与分块](#document-digitization-&-chunking)\n4. [嵌入模型](#embedding-models)\n5. [向量数据库的内部工作机制](#internal-working-of-vector-databases)\n6. [高级搜索算法](#advanced-search-algorithms)\n7. [语言模型的内部工作机制](#language-models-internal-working)\n8. [大语言模型的监督微调](#supervised-fine-tuning-of-llm)\n9. [偏好对齐（RLHF\u002FDPO）](#preference-alignment-rlhfdpo)\n10. [大语言模型系统的评估](#evaluation-of-llm-system)\n11. [幻觉控制技术](#hallucination-control-techniques)\n12. [大语言模型的部署](#deployment-of-llm)\n13. [基于智能体的系统](#agent-based-system)\n14. [提示词攻击](#prompt-hacking)\n15. [杂项](#miscellaneous)\n16. [案例研究](#case-studies)\n\n---\n\n## 提示工程与大语言模型基础\n\n- **预测\u002F判别式AI与生成式AI有什么区别？**\n- **什么是大语言模型？大语言模型是如何训练的？**\n- **语言模型中的“token”是什么？**\n- **如何估算运行基于SaaS的大语言模型和开源大语言模型的成本？**\n- **请解释“温度”参数及其设置方法。**\n- **在选择输出token时有哪些不同的解码策略？**\n- **在大语言模型中，可以采用哪些方式定义停止条件？**\n- **如何在大语言模型中使用停止序列？**\n- **请解释提示工程的基本结构。**\n- **请解释上下文学习。**\n- **请说明提示工程的类型。**\n- **在使用少样本提示时，需要注意哪些方面？**\n- **有哪些编写优质提示的策略？**\n- **什么是幻觉？如何通过提示工程来控制幻觉？**\n- **如何通过提示工程提升大语言模型的推理能力？**\n- **如果您的COT提示失败，如何进一步提升大语言模型的推理能力？**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 检索增强生成（RAG）\n\n- **如何提高大语言模型的准确性、可靠性，并使回答可验证？**\n- **RAG是如何工作的？**\n- **使用RAG系统有哪些优势？**\n- **什么情况下应该使用微调而不是RAG？**\n- **利用专有数据定制大语言模型的架构模式有哪些？**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 文档数字化与分块\n\n- **什么是分块？我们为什么要对数据进行分块？**\n- **影响分块大小的因素有哪些？**\n- **分块的方法有哪些类型？**\n- **如何确定理想的分块大小？**\n- **对于年度报告等复杂文档，最佳的数字化和分块方法是什么？**\n- **在分块过程中如何处理表格？**\n- **如何处理非常大的表格以提高检索效果？**\n- **在分块过程中如何处理列表项？**\n- **如何构建生产级的文档处理和索引管道？**\n- **在RAG中如何处理图表？**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 嵌入模型\n\n- **什么是向量嵌入？什么是嵌入模型？**\n- **嵌入模型在大语言模型应用中是如何使用的？**\n- **短文本和长文本的嵌入有何不同？**\n- **如何在你的数据上对嵌入模型进行基准测试？**\n- **假设你正在使用OpenAI的嵌入模型，但在基准测试后发现准确率较低，你将如何进一步提高嵌入搜索模型的准确率？**\n- **请详细说明如何改进用于嵌入的Sentence Transformer模型？**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 向量数据库的内部工作机制\n\n- **什么是向量数据库？**\n- **向量数据库与传统数据库有何不同？**\n- **向量数据库是如何工作的？**\n- **请解释向量索引、向量数据库和向量插件之间的区别。**\n- **你正在做一个涉及少量客户评论数据集的项目。你的任务是在数据集中找到相似的评论。优先目标是完美地找到最相似的评论，而搜索速度并不是主要考虑因素。你会选择哪种搜索策略？为什么？**\n- **请解释聚类和局部敏感哈希等向量搜索策略。**\n- **聚类如何缩小搜索空间？它在什么情况下会失效？我们又该如何缓解这些失效问题？**\n- **请解释随机投影索引。**\n- **请解释局部敏感哈希（LHS）索引方法。**\n- **请解释乘积量化（PQ）索引方法。**\n- **比较不同的向量索引，并根据具体场景，说明你会为某个项目选择哪种向量索引。**\n- **如何为特定用例决定理想的搜索相似度指标？**\n- **请解释向量数据库中过滤的不同类型及其挑战。**\n- **如何为你的需求选择最佳的向量数据库？**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 高级搜索算法\n\n- **信息检索与语义搜索的架构模式有哪些？**\n- **为什么拥有非常优秀的搜索功能非常重要？**\n- **在大规模数据集中，如何实现高效且准确的搜索结果？**\n- **假设客户已经构建了一个基于RAG的系统，但结果不够准确。经过调查发现，检索系统的准确性存在问题。你会采取哪些步骤来改进它？**\n- **请解释基于关键词的检索方法。**\n- **如何微调重排序模型？**\n- **请说明信息检索中最常用的指标及其失效的情况。**\n- **如果你要为一个类似Quora的问答系统设计算法，目标是让用户尽快找到最相关的答案，你会选择哪种评估指标来衡量系统的有效性？**\n- **我有一个推荐系统，应该使用什么指标来评估这个系统？**\n- **比较不同的信息检索指标，并说明在什么情况下使用哪一种？**\n- **混合搜索是如何工作的？**\n- **如果有来自多种方法的搜索结果，你将如何将这些排名合并并统一成一个单一的结果集？**\n- **如何处理多跳或多维度查询？**\n- **有哪些技术可以用来提升检索效果？**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 语言模型内部工作机制\n\n- **请详细解释自注意力机制的概念。**\n- **请说明自注意力机制的缺点以及如何克服这些缺点。**\n- **什么是位置编码？**\n- **请详细解释Transformer架构。**\n- **与LSTM相比，使用Transformer有哪些优势？**\n- **局部注意力和全局注意力有什么区别？**\n- **为什么Transformer对计算和内存的需求很高？我们该如何解决这个问题？**\n- **如何增加大语言模型的上下文长度？**\n- **如果我的词汇表有10万个词\u002F标记，我该如何优化Transformer架构？**\n- **过大的词汇表会导致计算问题，而过小的词汇表则会引发OOV问题。你认为应该如何找到词汇表的最佳平衡点？**\n- **请介绍不同类型的LLM架构，并说明哪种架构最适合哪种任务？**\n\n\n[返回顶部](#table-of-contents)\n\n---\n\n## LLM的监督微调\n\n- **什么是微调？为什么需要进行微调？**\n- **在哪些场景下我们需要对LLM进行微调？**\n- **如何决定是否需要进行微调？**\n- **如何改进模型，使其仅在有足够的上下文时才回答问题？**\n- **如何为问答任务创建微调数据集？**\n- **如何设置微调的超参数？**\n- **如何估算微调LLM所需的基础设施资源？**\n- **如何在消费级硬件上对LLM进行微调？**\n- **PEFT方法有哪些不同的类别？**\n- **什么是LLM中的灾难性遗忘？**\n- **有哪些不同的再参数化微调方法？**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 偏好对齐（RLHF\u002FDPO）\n\n- **在什么阶段你会选择偏好对齐方法而不是SFT？**\n- **什么是RLHF？它是如何使用的？**\n- **RLHF中存在什么样的奖励黑客问题？**\n- **请解释不同的偏好对齐方法。**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## LLM系统的评估\n\n- **如何评估最适合你应用场景的LLM模型？**\n- **如何评估基于RAG的系统？**\n- **评估LLM有哪些不同的指标？**\n- **请解释验证链方法。**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 幻觉控制技术\n\n- **幻觉有哪些不同的形式？**\n- **如何在各个层面控制幻觉？**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## LLM的部署\n\n- **为什么量化不会降低LLM的准确性？**\n- **有哪些技术可以优化LLM的推理过程以提高吞吐量？**\n- **如何在不使用分组查询注意力等近似方法的情况下加快模型响应速度？**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 基于智能体的系统\n\n- **请解释智能体的基本概念以及可用于实现智能体的不同策略。**\n- **我们为什么需要智能体？常见的智能体实现策略有哪些？**\n- **请结合代码示例解释ReAct提示法及其优势。**\n- **请解释计划与执行提示策略。**\n- **请结合代码示例解释OpenAI函数策略。**\n- **请说明OpenAI函数与LangChain智能体之间的区别。**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 提示词攻击\n\n- **什么是提示词攻击？为什么我们需要关注它？**\n- **提示词攻击有哪些不同的类型？**\n- **有哪些防御提示词攻击的策略？**\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 其他\n\n- **如何优化整个LLM系统的成本？**\n- **什么是专家混合模型（MoE）？**\n- **如何构建生产级别的RAG系统？请详细解释每个组件。**\n- **什么是FP8变量？它有哪些优势？**\n- **如何在不牺牲准确性的前提下，使用低精度训练来训练LLM？**\n- **如何计算KV缓存的大小？**\n- **请解释多头变换注意力模块中每一层的维度。**\n- **如何确保注意力层专注于输入的正确部分？**\n\n\n[返回顶部](#table-of-contents)\n\n---\n\n## 案例研究\n\n- **案例研究1**：根据查询动态调整上下文的LLM聊天助手\n- **案例研究2**：提示词技巧\n\n[返回顶部](#table-of-contents)\n\n---\n\n*如需这些问题的答案，请访问[Mastering LLM](https:\u002F\u002Fwww.masteringllm.com\u002Fcourse\u002Fllm-interview-questions-and-answers?previouspage=allcourses&isenrolled=no#\u002Fhome)。*","# LLMInterviewQuestions 快速上手指南\n\n**工具简介**：`LLMInterviewQuestions` 并非一个需要安装运行的软件库，而是一个汇集了 100+ 大语言模型（LLM）面试真题的知识库。它涵盖了从提示工程、RAG、向量数据库到模型微调、对齐及部署等 15 个核心领域，专为准备谷歌、英伟达、Meta 等顶尖科技公司面试的开发者设计。\n\n由于本项目本质为文档资源，无需复杂的环境配置，以下是针对中国开发者的阅读与使用指南。\n\n## 环境准备\n\n本项目无需特定的操作系统或硬件依赖，只需具备以下基础环境即可浏览和学习：\n\n- **操作系统**：Windows, macOS, 或 Linux 均可。\n- **前置依赖**：\n  - 现代浏览器（推荐 Chrome, Edge 或 Firefox）。\n  - GitHub 账号（可选，用于 Star 项目或追踪更新）。\n  - 网络环境：由于仓库托管于 GitHub，国内访问可能受限，建议配置科学上网环境或使用 GitHub 加速镜像服务。\n\n## 获取与访问步骤\n\n你可以通过以下两种方式获取内容：\n\n### 方式一：在线直接浏览（推荐）\n直接访问 GitHub 仓库页面查看目录和问题列表：\n1. 打开浏览器访问项目地址：`https:\u002F\u002Fgithub.com\u002FDSCKG\u002FLLMInterviewQuestions` (注：此处为示例地址，实际请以源仓库为准)。\n2. 利用页面右侧的 **Table of Contents** 快速跳转至感兴趣的分类（如 RAG、Embedding Models 等）。\n\n### 方式二：克隆到本地（便于离线查阅与笔记）\n如果你希望将问题列表下载到本地进行标记或二次整理，可使用以下命令：\n\n```bash\n# 克隆仓库\ngit clone https:\u002F\u002Fgithub.com\u002FDSCKG\u002FLLMInterviewQuestions.git\n\n# 进入目录\ncd LLMInterviewQuestions\n\n# 使用 Markdown 编辑器（如 VS Code）打开\ncode .\n```\n\n*注：若下载速度慢，可尝试使用国内镜像源克隆（需替换具体镜像地址）：*\n```bash\ngit clone https:\u002F\u002Fgitee.com\u002Fmirror\u002FLLMInterviewQuestions.git\n```\n\n## 基本使用\n\n本项目的核心用法是**按主题模块进行针对性复习**。以下是最高效的使用路径：\n\n### 1. 定位知识盲区\n根据 `Table of Contents` 找到你薄弱的环节。例如，若你对检索增强生成（RAG）不熟悉，直接跳转至 [Retrieval Augmented Generation (RAG)](#retrieval-augmented-generation-rag) 章节。\n\n### 2. 自测与思考\n阅读该章节下的具体问题，尝试在不看答案的情况下口头或书面回答。\n**示例场景**：\n> **问题**：*How does RAG work?* (RAG 是如何工作的？)\n> **自测动作**：尝试画出 RAG 的流程图，解释文档切片、向量化、检索和生成的全过程。\n\n### 3. 深入进阶学习\n对于难以回答的问题（如 *Explain different preference alignment methods*），该项目提供了官方配套的课程链接作为详细解答来源。\n- 访问 [Mastering LLM](https:\u002F\u002Fwww.masteringllm.com\u002Fcourse\u002Fllm-interview-questions-and-answers) 获取深度解析。\n- *提示：结账时使用代码 `LLM50` 可享受优惠。*\n\n### 4. 实战案例研究\n完成理论基础后，务必阅读底部的 **Case Studies** 部分，通过 \"LLM Chat Assistant with dynamic context\" 等真实案例，将理论知识转化为解决实际问题的思路。\n\n---\n*本指南旨在帮助开发者高效利用该开源题库。具体的面试题答案与深度技术解析，请参考项目提供的官方课程资源。*","某大型金融科技公司的高级算法工程师李明，正紧急准备一家头部科技巨头的 LLM 架构师岗位面试，需要在短时间内系统梳理从 RAG 优化到模型微调的全链路知识。\n\n### 没有 LLMInterviewQuestions 时\n- **复习范围模糊**：面对海量的 LLM 论文和博客，难以确定谷歌、英伟达等大厂究竟关注哪些核心考点，导致复习精力分散。\n- **实战细节缺失**：虽然懂理论，但遇到“如何处理年报中的复杂表格分块”或\"RAG 系统中幻觉控制的具体策略”等生产级问题时，缺乏具体的解决思路。\n- **知识体系碎片化**：对 RLHF、DPO 偏好对齐及向量数据库内部原理等进阶内容理解零散，无法在面试中构建逻辑严密的回答框架。\n- **模拟演练不足**：缺少涵盖提示词攻击、智能体系统设计等前沿话题的高质量模拟题，难以应对面试官的深度追问。\n\n### 使用 LLMInterviewQuestions 后\n- **考点精准锁定**：直接获取按 15 个类别整理的 100+ 道真题，清晰掌握大厂在 Prompt 工程、部署及评估等环节的真实考察重点。\n- **场景化深度解析**：通过“文档数字化与分块”等章节，迅速掌握了处理复杂图表和长列表的工业界最佳实践，能从容回答生产环境难题。\n- **结构化知识重构**：借助从嵌入模型原理到监督微调的完整大纲，将碎片知识串联成体系，能够条理清晰地阐述技术选型背后的权衡。\n- **前沿话题全覆盖**：提前演练了关于 Agent 系统和提示词黑客防御等高阶问题，在面试中展现出超越预期的技术视野和落地能力。\n\nLLMInterviewQuestions 将原本漫无目的的泛读转化为针对大厂实战需求的精准突击，显著提升了候选人的技术自信与面试通过率。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fllmgenai_LLMInterviewQuestions_3c532327.png","llmgenai","Mastering LLM","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fllmgenai_fc90dc3a.jpg",null,"India","https:\u002F\u002Fwww.masteringllm.com\u002F","https:\u002F\u002Fgithub.com\u002Fllmgenai",1716,363,"2026-04-03T06:19:36",1,"","未说明",{"notes":89,"python":87,"dependencies":90},"该仓库仅包含大语言模型（LLM）相关的面试问题列表和目录，不包含任何可执行的代码、模型文件或安装脚本。因此，它没有特定的运行环境、GPU、内存或依赖库需求。用户只需通过浏览器阅读 Markdown 文档，或访问文中提到的外部课程链接获取答案。",[],[26,13,54],"2026-03-27T02:49:30.150509","2026-04-06T06:52:05.452306",[],[]]