[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"site-home-live-entry-setting":3,"site-header-contests":8,"site-header-camps":49,"study-material-detail-study-29":104},{"enabled":4,"title":5,"statusText":5,"buttonText":5,"campSlug":5,"stage":6,"classroomId":7,"hasClassroom":4},false,"",0,null,{"total":9,"list":10,"pageNum":44,"pageSize":45,"size":9,"startRow":44,"endRow":9,"pages":44,"prePage":6,"nextPage":6,"isFirstPage":46,"isLastPage":46,"hasPreviousPage":4,"hasNextPage":4,"navigatePages":47,"navigatepageNums":48,"navigateFirstPage":44,"navigateLastPage":44},3,[11,25,35],{"slug":12,"title":13,"logo":7,"liveUrl":14,"prize":15,"sponsor":16,"description":17,"organizer":18,"startTime":19,"endTime":20,"registrationDeadline":21,"status":22,"dashboardVisibility":23,"createdAt":24},"spring2026","2026春季启元人工智能大赛","https:\u002F\u002Fwww.bilibili.com\u002Fblackboard\u002Flive\u002Flive-activity-player.html?cid=1822703816&quality=0","200,000","麒麟软件有限公司","“2026春季启元人工智能大赛”是由启元实验室主办的面向全国人工智能领域开发者、科研人员及高校学生的权威赛事。大赛旨在促进与培养具备工程实践能力和创新能力的优秀人工智能领域开发者、推广国产软硬件技术与生态、助力国家人工智能产业高质量发展。\n各参赛选手需在规定的比赛周期内，依据赛事规则选定赛题并完成相应挑战，优胜者将共同角逐总额20万元的赛事奖金池，并且还将有机会获得实习和知名企业优质岗位内推资格。","启元实验室","2026-04-19 16:00:00","2026-07-13 16:00:00","2026-06-16 16:00:00","ongoing","private","2026-04-16 08:00:00",{"slug":26,"title":27,"logo":7,"liveUrl":14,"prize":28,"sponsor":7,"description":29,"organizer":18,"startTime":30,"endTime":31,"registrationDeadline":32,"status":33,"dashboardVisibility":23,"createdAt":34},"autumn2025","2025秋季启元人工智能大赛","150,000","“2025秋季启元人工智能大赛”是由启元实验室主办的面向全国人工智能领域开发者、科研人员及高校学生的权威赛事。大赛旨在促进与培养具备工程实践能力和创新能力的优秀人工智能领域开发者、推广国产软硬件技术与生态、助力国家人工智能产业高质量发展。","2025-11-08 08:00:00","2026-01-13 08:00:00","2025-12-30 08:00:00","ended","2025-11-03 08:00:00",{"slug":36,"title":37,"logo":7,"liveUrl":14,"prize":38,"sponsor":7,"description":39,"organizer":18,"startTime":40,"endTime":41,"registrationDeadline":42,"status":33,"dashboardVisibility":23,"createdAt":43},"summer2025","启元人工智能大赛(系统方向)","80,000","启元人工智能大赛由启元实验室主办，其中系统赛道基于启元实验室九源软件栈的相关开源项目，通过拉动开源社区力量，一方面促进国产智能计算生态的发展，另一方面也为人工智能系统方向的学生及从业者提供一个实践平台。","2025-07-08 08:00:00","2025-08-25 08:00:00","2025-07-22 08:00:00","2025-06-22 08:00:00",1,5,true,8,[44],{"total":45,"list":50,"pageNum":44,"pageSize":45,"size":45,"startRow":44,"endRow":45,"pages":44,"prePage":6,"nextPage":6,"isFirstPage":46,"isLastPage":46,"hasPreviousPage":4,"hasNextPage":4,"navigatePages":47,"navigatepageNums":103,"navigateFirstPage":44,"navigateLastPage":44},[51,62,73,84,94],{"id":52,"slug":53,"title":54,"logo":55,"publicLiveEmbedUrl":14,"description":56,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":16,"startTime":58,"endTime":59,"status":22,"dashboardVisibility":60,"createdAt":58,"updatedAt":61},"019eeeff-6b00-7d55-a629-efafded3e232","summer2026","2026年夏季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fuploads\u002F2026-06\u002F019eef7f-7816-7828-9309-79c2c61dd568.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=Loo9mNcAiqrEJeoxKJGChNjsIP0%3D","由启元实验室发起的InfiniTensor大模型与人工智能系统训练营，面向开源社区、高校及产业界伙伴开放。活动以「普及AI全栈知识技术、培养工程实践与创新型人才、推广国产软硬件技术生态」为核心目标，助力国家人工智能产业高质量发展。","...","2026-06-22 08:00:00","2026-10-19 08:00:00","public","2026-10-12 08:00:00",{"id":63,"slug":64,"title":65,"logo":66,"publicLiveEmbedUrl":7,"description":67,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":68,"startTime":69,"endTime":70,"status":33,"dashboardVisibility":23,"createdAt":71,"updatedAt":72},"019af3f3-e06a-7245-aa8d-37dc99a69c73","winter2025","2025年冬季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fuploads\u002F2025-12\u002F019afc69-b7ca-7805-91fb-605429671c01.png?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=aQ7DEIPmeZ6G2GUnnO42fZDzkoY%3D","由启元实验室发起的InfiniTensor大模型与人工智能系统训练营，面向开源社区、高校及产业界伙伴开放。活动以「普及AI全栈知识技术、培养工程实践与创新型人才、推广国产软硬件技术生态」为核心目标，助力国家人工智能产业高质量发展。 \n","新华三技术有限公司","2025-12-25 08:00:00","2026-03-27 08:00:00","2025-12-06 08:00:00","2026-10-22 20:24:30",{"id":74,"slug":36,"title":75,"logo":76,"publicLiveEmbedUrl":7,"description":77,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":78,"startTime":79,"endTime":80,"status":81,"dashboardVisibility":23,"createdAt":82,"updatedAt":83},"01980342-208c-7803-b3f3-19e1d31960ba","2025年夏季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fuploads\u002F2025-07\u002F0198045e-28fe-78f0-95fe-23835ce58395.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=rRUH7HCePjJ0x1Z1esoWYs0tjmE%3D","InfiniTensor大模型与人工智能系统训练营由启元实验室及清华大学 AI 系统方向团队发起，面向开源社区、高校和产业界，普及人工智能各方面知识和技术，培养相关人才。训练营联合开源项目开发者、学术界大牛、工业界资深工程师参与授课与项目指导。课程采用「线上直播授课 + 在线答疑 + 项目实践辅导 」的方式进行，确保每位同学都能获得充分成长。训练营全程免费，通过视频、直播课、习题和项目实践，在人工智能飞速发展的背景下，为学员提供一窥深层技术的机会。","北京积算科技有限公司","2025-07-14 08:00:00","2025-09-30 08:00:00","upcoming","2025-07-13 08:00:00","2026-10-22 20:24:58",{"id":85,"slug":86,"title":87,"logo":88,"publicLiveEmbedUrl":7,"description":89,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":5,"startTime":90,"endTime":91,"status":81,"dashboardVisibility":23,"createdAt":92,"updatedAt":93},"01980cf9-a7d8-7796-ae86-0bbe05334add","winter2024","2024年冬季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Ftesting-cdn.infinitensor.com\u002Fuploads\u002F2025-07\u002F01981269-ce23-7624-a287-9f07dd0d85f6.jpg","InfiniTensor大模型与人工智能系统训练营由清华大学发起，面向开源社区、高校和产业界，普及人工智能各方面知识和技术，培养相关人才。训练营全程免费，通过视频、直播课、习题和项目实践，在人工智能飞速发展的背景下，为学员提供一窥深层技术的机会。报名学员请阅读“导学”课程页的训练营介绍。","2025-01-05 08:00:00","2025-03-22 08:00:00","2025-07-15 08:00:00","2026-10-22 20:24:39",{"id":95,"slug":96,"title":97,"logo":98,"publicLiveEmbedUrl":7,"description":99,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":5,"startTime":100,"endTime":101,"status":81,"dashboardVisibility":23,"createdAt":92,"updatedAt":102},"01980c20-363a-7dc3-8f5a-58a68e08dcc4","summer2024","2024年夏季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Ftesting-cdn.infinitensor.com\u002Fuploads\u002F2025-07\u002F01981269-fbfa-788a-afcd-17fa9be74ff9.jpg","InfiniTensor大模型与人工智能系统训练营由清华大学发起，面向开源社区和高校，普及人工智能各方面知识和技术，培养相关人才。训练营全程免费，通过课程、习题和项目实践，在人工智能飞速发展的背景下，为学员提供一窥深层技术的机会。","2024-07-20 08:00:00","2024-09-29 08:00:00","2026-10-22 20:24:49",[44],{"currentItem":105,"processedData":125},{"img":106,"keywords":107,"treeType":44,"description":108,"showStatus":44,"updateTime":109,"type":110,"title":111,"content":112,"nextArticle":113,"createTime":109,"sortOrder":117,"canonicalPath":118,"id":119,"contentType":120,"prevArticle":121},"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_fm.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=2rbV3V3172LzXJiKwdBT298ihN0%3D","InfiniTensor,冬季训练营,RAG,检索增强,语言模型","围绕检索增强语言模型展开，讲解 RAG 思路、知识接入流程与问答效果提升策略。","2026-10-22 20:38:11",2,"InfiniTensor 冬季训练营前沿模型第三课《检索增强的语言模型》","![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw01.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=C%2B%2FEK%2B0HVRsp1SLfNa3CZgmfw58%3D)\r\n\r\n\r\n\r\n2026 年 1 月 21 日，2025 冬季 InfiniTensor 大模型与人工智能系统训练营前沿模型系列第三课《检索增强的语言模型》开讲。\r\n\r\n本节课由九格大模型团队讲述 RAG 技术相关基础知识、技术发展历程以及最新研究进展。\r\n\r\n### 语言模型基础\r\n#### 1. 语言模型分类：\r\n- 自回归语言模型：GPT、千问等\r\n- 掩码语言模型：BERT\r\n#### 2. 评价指标：困惑度（Perplexity）\r\n#### 3. 典型结构：\r\n- Encoder-Decoder 架构\r\n- Encoder-only\r\n- Decoder-only 的区别\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw02.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=FlsLsOULNAH2Ybtw00orLcDBISw%3D)\r\n\r\n### RAG 技术概述\r\n#### 1. RAG 提出背景：\r\n- 预训练模型的局限性：知识密集型场景、时效性信息、幻觉问题\r\n#### 2. RAG 基本原理：\r\n- 外部信息以平文本形式存在，通过检索增强模型生成能力\r\n- 基础范式：用户 Query → 检索知识库 → 检索信息与 Query 结合 → 模型生成回答\r\n\r\n#### 3. 检索范式\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw03.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=bCQvDJC%2FxySRmfHP%2B7KSiJSjhE8%3D)\r\n\r\n\r\n### RAG 早期经典工作\r\n#### 1. REALM：\r\n- 掩码模型，检索相关片段拼接后预测\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw04.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=XBy0qIg%2BklFxb%2BYBdp7Y8N9XoyU%3D)\r\n\r\n- 检索模型训练：\r\n  - 单塔与双塔模型原理\r\n  - 对比学习训练 Embedding 模型\r\n- 检索后处理：多次推理与概率平均\r\n- 后续工作：DPR、RAG、Atlas 等面向开放域问答任务\r\n#### 2. Retrieval-in-context LM\r\n- **顺序预测模型**\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw05.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=brzbZcVQ%2B1nJy5aSQZna3Y2hNCA%3D)\r\n\r\n- **检索对象选择:**\r\n  - 查询语句长度与噪声问题\r\n  - 实验分析：查询语句长度对检索效果的影响\r\n- **检索频率优化：**\r\n  - 提高检索频率对模型性能的影响\r\n  - 计算量与检索成本的平衡\r\n- **召回文本数量影响：**\r\n  - 召回数量与模型性能的关系\r\n  - 噪声引入与性能下降问题\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw06.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=ql4fidZxCC%2BpspDWL4ucmXdwz70%3D)\r\n\r\n#### 3. RETRO\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw07.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=ZY3K2NUTnZAOgMiOgYA9F%2BPFW9A%3D)\r\n\r\n#### 4. kNN-LM\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw08.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=OvG098ZYFnRyfqvfCPBjjUTtQMA%3D)\r\n\r\n#### 5. 对比\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw09.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=NSZ6DOlmJwtybtrC6VesaQ6gsZM%3D)\r\n\r\n\r\n\r\n### RAG 技术发展\r\n#### 1. 发展阶段：\r\n- 2022年：建立 RAG 基本范式\r\n- 2023年：关注检索时机与方式优化\r\n- 近期：结合思维链（CoT）、多模态知识注入\r\n#### 2. 核心共识：\r\n- 基于 Embedding 模型的 Dense Retrieve 成为标配\r\n- 输入范式：输入层直接拼接文本\r\n\r\n### RAG 最新研究进展\r\n#### 1. 检索过程优化：\r\n- **多轮检索**（难以避免噪声问题）与**自适应检索**（信息割裂，缺少交互整合）\r\n- 引入**大模型笔记**概念：记录检索信息、规划检索过程、减少噪声\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw10.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=SWX633EFOI3HkLPjpLoeKmJfpCs%3D)\r\n\r\n#### 2. 知识有效利用：\r\n- 指令跟随能力在知识密集场景下的局限性\r\n- 强化训练方法：DPO、GRPO 等\r\n- RAG-DDR：基于 DPO 的生成链路强化调优\r\n  - 后向对齐策略：模块化训练与采样调节\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw11.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=GvhzbnqWgxOvSgYy5PbP3pT5gh4%3D)\r\n\r\n### UltraRAG 开源项目介绍\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw12.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=1ltbqRf1zIIrBFrrgLdFFyWCtng%3D)\r\n\r\n\r\n\r\n### 总结\r\n本节课讲解了 RAG 技术原理、技术面临的挑战与解决方案，RAG 发展历程与最新研究进展。\r\n\r\n\r\n\r\n![2025冬季训练营课程表](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw6.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=F9UYYN6GHyw0aDIvfyvnXUb6gKw%3D)\r\n\r\n### 💡 学习贴士\r\n更多细节及完整课程内容，请观看直播或查看课程回放。\r\n\r\n链接：https:\u002F\u002Fwww.infinitensor.com\u002Fcamp\u002Fwinter2025\r\n\r\n### 📚 训练营报名持续进行中\r\n扫码报名与加入群聊，或点击 **「阅读原文」** 直达官网报名\r\n\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw7.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=4qCORU7P3l9%2FAWZfM%2BMlpmzjXSU%3D)",{"publishTime":7,"createTime":109,"canonicalPath":114,"id":115,"title":116},"\u002Fstudy-material\u002Fstudy-28","study-28","InfiniTensor 冬季训练营推理引擎第三课《从大模型推理到 AI 对话》",80,"\u002Fstudy-material\u002Fstudy-29","study-29","markdown",{"publishTime":7,"createTime":109,"canonicalPath":122,"id":123,"title":124},"\u002Fstudy-material\u002Fstudy-30","study-30","InfiniTensor 冬季训练营 CUDA 编程第四课《CUDA 程序迁移至天数智芯 GPU》",{"htmlContent":126,"headings":127},"\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw01.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=C%2B%2FEK%2B0HVRsp1SLfNa3CZgmfw58%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cp>2026 年 1 月 21 日，2025 冬季 InfiniTensor 大模型与人工智能系统训练营前沿模型系列第三课《检索增强的语言模型》开讲。\u003C\u002Fp>\n\u003Cp>本节课由九格大模型团队讲述 RAG 技术相关基础知识、技术发展历程以及最新研究进展。\u003C\u002Fp>\n\u003Ch3 id=\"语言模型基础\">\u003Ca class=\"anchor\" href=\"#语言模型基础\" aria-label=\"跳转到标题：语言模型基础\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>语言模型基础\u003C\u002Fh3>\u003Ch4 id=\"1-语言模型分类\">\u003Ca class=\"anchor\" href=\"#1-语言模型分类\" aria-label=\"跳转到标题：1. 语言模型分类：\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>1. 语言模型分类：\u003C\u002Fh4>\u003Cul>\n\u003Cli>自回归语言模型：GPT、千问等\u003C\u002Fli>\n\u003Cli>掩码语言模型：BERT\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch4 id=\"2-评价指标-困惑度-perplexity\">\u003Ca class=\"anchor\" href=\"#2-评价指标-困惑度-perplexity\" aria-label=\"跳转到标题：2. 评价指标：困惑度（Perplexity）\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>2. 评价指标：困惑度（Perplexity）\u003C\u002Fh4>\u003Ch4 id=\"3-典型结构\">\u003Ca class=\"anchor\" href=\"#3-典型结构\" aria-label=\"跳转到标题：3. 典型结构：\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>3. 典型结构：\u003C\u002Fh4>\u003Cul>\n\u003Cli>Encoder-Decoder 架构\u003C\u002Fli>\n\u003Cli>Encoder-only\u003C\u002Fli>\n\u003Cli>Decoder-only 的区别\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw02.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=FlsLsOULNAH2Ybtw00orLcDBISw%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch3 id=\"rag-技术概述\">\u003Ca class=\"anchor\" href=\"#rag-技术概述\" aria-label=\"跳转到标题：RAG 技术概述\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>RAG 技术概述\u003C\u002Fh3>\u003Ch4 id=\"1-rag-提出背景\">\u003Ca class=\"anchor\" href=\"#1-rag-提出背景\" aria-label=\"跳转到标题：1. RAG 提出背景：\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>1. RAG 提出背景：\u003C\u002Fh4>\u003Cul>\n\u003Cli>预训练模型的局限性：知识密集型场景、时效性信息、幻觉问题\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch4 id=\"2-rag-基本原理\">\u003Ca class=\"anchor\" href=\"#2-rag-基本原理\" aria-label=\"跳转到标题：2. RAG 基本原理：\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>2. RAG 基本原理：\u003C\u002Fh4>\u003Cul>\n\u003Cli>外部信息以平文本形式存在，通过检索增强模型生成能力\u003C\u002Fli>\n\u003Cli>基础范式：用户 Query → 检索知识库 → 检索信息与 Query 结合 → 模型生成回答\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch4 id=\"3-检索范式\">\u003Ca class=\"anchor\" href=\"#3-检索范式\" aria-label=\"跳转到标题：3. 检索范式\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>3. 检索范式\u003C\u002Fh4>\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw03.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=bCQvDJC%2FxySRmfHP%2B7KSiJSjhE8%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch3 id=\"rag-早期经典工作\">\u003Ca class=\"anchor\" href=\"#rag-早期经典工作\" aria-label=\"跳转到标题：RAG 早期经典工作\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>RAG 早期经典工作\u003C\u002Fh3>\u003Ch4 id=\"1-realm\">\u003Ca class=\"anchor\" href=\"#1-realm\" aria-label=\"跳转到标题：1. REALM：\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>1. REALM：\u003C\u002Fh4>\u003Cul>\n\u003Cli>掩码模型，检索相关片段拼接后预测\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw04.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=XBy0qIg%2BklFxb%2BYBdp7Y8N9XoyU%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cul>\n\u003Cli>检索模型训练：\u003Cul>\n\u003Cli>单塔与双塔模型原理\u003C\u002Fli>\n\u003Cli>对比学习训练 Embedding 模型\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003Cli>检索后处理：多次推理与概率平均\u003C\u002Fli>\n\u003Cli>后续工作：DPR、RAG、Atlas 等面向开放域问答任务\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch4 id=\"2-retrieval-in-context-lm\">\u003Ca class=\"anchor\" href=\"#2-retrieval-in-context-lm\" aria-label=\"跳转到标题：2. Retrieval-in-context LM\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>2. Retrieval-in-context LM\u003C\u002Fh4>\u003Cul>\n\u003Cli>\u003Cstrong>顺序预测模型\u003C\u002Fstrong>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw05.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=brzbZcVQ%2B1nJy5aSQZna3Y2hNCA%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>检索对象选择:\u003C\u002Fstrong>\u003Cul>\n\u003Cli>查询语句长度与噪声问题\u003C\u002Fli>\n\u003Cli>实验分析：查询语句长度对检索效果的影响\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003Cli>\u003Cstrong>检索频率优化：\u003C\u002Fstrong>\u003Cul>\n\u003Cli>提高检索频率对模型性能的影响\u003C\u002Fli>\n\u003Cli>计算量与检索成本的平衡\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003Cli>\u003Cstrong>召回文本数量影响：\u003C\u002Fstrong>\u003Cul>\n\u003Cli>召回数量与模型性能的关系\u003C\u002Fli>\n\u003Cli>噪声引入与性能下降问题\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw06.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=ql4fidZxCC%2BpspDWL4ucmXdwz70%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch4 id=\"3-retro\">\u003Ca class=\"anchor\" href=\"#3-retro\" aria-label=\"跳转到标题：3. RETRO\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>3. RETRO\u003C\u002Fh4>\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw07.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=ZY3K2NUTnZAOgMiOgYA9F%2BPFW9A%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch4 id=\"4-knn-lm\">\u003Ca class=\"anchor\" href=\"#4-knn-lm\" aria-label=\"跳转到标题：4. kNN-LM\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>4. kNN-LM\u003C\u002Fh4>\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw08.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=OvG098ZYFnRyfqvfCPBjjUTtQMA%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch4 id=\"5-对比\">\u003Ca class=\"anchor\" href=\"#5-对比\" aria-label=\"跳转到标题：5. 对比\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>5. 对比\u003C\u002Fh4>\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw09.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=NSZ6DOlmJwtybtrC6VesaQ6gsZM%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch3 id=\"rag-技术发展\">\u003Ca class=\"anchor\" href=\"#rag-技术发展\" aria-label=\"跳转到标题：RAG 技术发展\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>RAG 技术发展\u003C\u002Fh3>\u003Ch4 id=\"1-发展阶段\">\u003Ca class=\"anchor\" href=\"#1-发展阶段\" aria-label=\"跳转到标题：1. 发展阶段：\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>1. 发展阶段：\u003C\u002Fh4>\u003Cul>\n\u003Cli>2022年：建立 RAG 基本范式\u003C\u002Fli>\n\u003Cli>2023年：关注检索时机与方式优化\u003C\u002Fli>\n\u003Cli>近期：结合思维链（CoT）、多模态知识注入\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch4 id=\"2-核心共识\">\u003Ca class=\"anchor\" href=\"#2-核心共识\" aria-label=\"跳转到标题：2. 核心共识：\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>2. 核心共识：\u003C\u002Fh4>\u003Cul>\n\u003Cli>基于 Embedding 模型的 Dense Retrieve 成为标配\u003C\u002Fli>\n\u003Cli>输入范式：输入层直接拼接文本\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch3 id=\"rag-最新研究进展\">\u003Ca class=\"anchor\" href=\"#rag-最新研究进展\" aria-label=\"跳转到标题：RAG 最新研究进展\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>RAG 最新研究进展\u003C\u002Fh3>\u003Ch4 id=\"1-检索过程优化\">\u003Ca class=\"anchor\" href=\"#1-检索过程优化\" aria-label=\"跳转到标题：1. 检索过程优化：\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>1. 检索过程优化：\u003C\u002Fh4>\u003Cul>\n\u003Cli>\u003Cstrong>多轮检索\u003C\u002Fstrong>（难以避免噪声问题）与\u003Cstrong>自适应检索\u003C\u002Fstrong>（信息割裂，缺少交互整合）\u003C\u002Fli>\n\u003Cli>引入\u003Cstrong>大模型笔记\u003C\u002Fstrong>概念：记录检索信息、规划检索过程、减少噪声\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw10.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=SWX633EFOI3HkLPjpLoeKmJfpCs%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch4 id=\"2-知识有效利用\">\u003Ca class=\"anchor\" href=\"#2-知识有效利用\" aria-label=\"跳转到标题：2. 知识有效利用：\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>2. 知识有效利用：\u003C\u002Fh4>\u003Cul>\n\u003Cli>指令跟随能力在知识密集场景下的局限性\u003C\u002Fli>\n\u003Cli>强化训练方法：DPO、GRPO 等\u003C\u002Fli>\n\u003Cli>RAG-DDR：基于 DPO 的生成链路强化调优\u003Cul>\n\u003Cli>后向对齐策略：模块化训练与采样调节\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw11.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=GvhzbnqWgxOvSgYy5PbP3pT5gh4%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch3 id=\"ultrarag-开源项目介绍\">\u003Ca class=\"anchor\" href=\"#ultrarag-开源项目介绍\" aria-label=\"跳转到标题：UltraRAG 开源项目介绍\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>UltraRAG 开源项目介绍\u003C\u002Fh3>\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_29_zw12.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=1ltbqRf1zIIrBFrrgLdFFyWCtng%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch3 id=\"总结\">\u003Ca class=\"anchor\" href=\"#总结\" aria-label=\"跳转到标题：总结\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>总结\u003C\u002Fh3>\u003Cp>本节课讲解了 RAG 技术原理、技术面临的挑战与解决方案，RAG 发展历程与最新研究进展。\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw6.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=F9UYYN6GHyw0aDIvfyvnXUb6gKw%3D\" alt=\"2025冬季训练营课程表\">\u003C\u002Fp>\n\u003Ch3 id=\"学习贴士\">\u003Ca class=\"anchor\" href=\"#学习贴士\" aria-label=\"跳转到标题：💡 学习贴士\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>💡 学习贴士\u003C\u002Fh3>\u003Cp>更多细节及完整课程内容，请观看直播或查看课程回放。\u003C\u002Fp>\n\u003Cp>链接：\u003Ca href=\"https:\u002F\u002Fwww.infinitensor.com\u002Fcamp\u002Fwinter2025\">https:\u002F\u002Fwww.infinitensor.com\u002Fcamp\u002Fwinter2025\u003C\u002Fa>\u003C\u002Fp>\n\u003Ch3 id=\"训练营报名持续进行中\">\u003Ca class=\"anchor\" href=\"#训练营报名持续进行中\" aria-label=\"跳转到标题：📚 训练营报名持续进行中\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>📚 训练营报名持续进行中\u003C\u002Fh3>\u003Cp>扫码报名与加入群聊，或点击 \u003Cstrong>「阅读原文」\u003C\u002Fstrong> 直达官网报名\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw7.jpg?Expires=1792752563&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=4qCORU7P3l9%2FAWZfM%2BMlpmzjXSU%3D\" alt=\"\">\u003C\u002Fp>\n",[128,130,134,137,140,143,146,149,152,155,158,161,164,167,170,173,176,179,182,185,188,191,193,196],{"id":129,"text":129,"level":9},"语言模型基础",{"id":131,"text":132,"level":133},"1-语言模型分类","1. 语言模型分类：",4,{"id":135,"text":136,"level":133},"2-评价指标-困惑度-perplexity","2. 评价指标：困惑度（Perplexity）",{"id":138,"text":139,"level":133},"3-典型结构","3. 典型结构：",{"id":141,"text":142,"level":9},"rag-技术概述","RAG 技术概述",{"id":144,"text":145,"level":133},"1-rag-提出背景","1. RAG 提出背景：",{"id":147,"text":148,"level":133},"2-rag-基本原理","2. RAG 基本原理：",{"id":150,"text":151,"level":133},"3-检索范式","3. 检索范式",{"id":153,"text":154,"level":9},"rag-早期经典工作","RAG 早期经典工作",{"id":156,"text":157,"level":133},"1-realm","1. REALM：",{"id":159,"text":160,"level":133},"2-retrieval-in-context-lm","2. Retrieval-in-context LM",{"id":162,"text":163,"level":133},"3-retro","3. RETRO",{"id":165,"text":166,"level":133},"4-knn-lm","4. kNN-LM",{"id":168,"text":169,"level":133},"5-对比","5. 对比",{"id":171,"text":172,"level":9},"rag-技术发展","RAG 技术发展",{"id":174,"text":175,"level":133},"1-发展阶段","1. 发展阶段：",{"id":177,"text":178,"level":133},"2-核心共识","2. 核心共识：",{"id":180,"text":181,"level":9},"rag-最新研究进展","RAG 最新研究进展",{"id":183,"text":184,"level":133},"1-检索过程优化","1. 检索过程优化：",{"id":186,"text":187,"level":133},"2-知识有效利用","2. 知识有效利用：",{"id":189,"text":190,"level":9},"ultrarag-开源项目介绍","UltraRAG 开源项目介绍",{"id":192,"text":192,"level":9},"总结",{"id":194,"text":195,"level":9},"学习贴士","💡 学习贴士",{"id":197,"text":198,"level":9},"训练营报名持续进行中","📚 训练营报名持续进行中"]