[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-mli--paper-reading":3,"tool-mli--paper-reading":61},[4,18,26,36,44,53],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},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,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},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",143909,2,"2026-04-07T11:33:18",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},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。",107888,"2026-04-06T11:32:50",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":32,"last_commit_at":50,"category_tags":51,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":10,"last_commit_at":59,"category_tags":60,"status":17},4487,"LLMs-from-scratch","rasbt\u002FLLMs-from-scratch","LLMs-from-scratch 是一个基于 PyTorch 的开源教育项目，旨在引导用户从零开始一步步构建一个类似 ChatGPT 的大型语言模型（LLM）。它不仅是同名技术著作的官方代码库，更提供了一套完整的实践方案，涵盖模型开发、预训练及微调的全过程。\n\n该项目主要解决了大模型领域“黑盒化”的学习痛点。许多开发者虽能调用现成模型，却难以深入理解其内部架构与训练机制。通过亲手编写每一行核心代码，用户能够透彻掌握 Transformer 架构、注意力机制等关键原理，从而真正理解大模型是如何“思考”的。此外，项目还包含了加载大型预训练权重进行微调的代码，帮助用户将理论知识延伸至实际应用。\n\nLLMs-from-scratch 特别适合希望深入底层原理的 AI 开发者、研究人员以及计算机专业的学生。对于不满足于仅使用 API，而是渴望探究模型构建细节的技术人员而言，这是极佳的学习资源。其独特的技术亮点在于“循序渐进”的教学设计：将复杂的系统工程拆解为清晰的步骤，配合详细的图表与示例，让构建一个虽小但功能完备的大模型变得触手可及。无论你是想夯实理论基础，还是为未来研发更大规模的模型做准备",90106,"2026-04-06T11:19:32",[35,15,13,14],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":66,"readme_en":67,"readme_zh":68,"quickstart_zh":69,"use_case_zh":70,"hero_image_url":71,"owner_login":72,"owner_name":73,"owner_avatar_url":74,"owner_bio":75,"owner_company":76,"owner_location":77,"owner_email":75,"owner_twitter":75,"owner_website":75,"owner_url":78,"languages":75,"stars":79,"forks":80,"last_commit_at":81,"license":82,"difficulty_score":83,"env_os":84,"env_gpu":85,"env_ram":85,"env_deps":86,"category_tags":89,"github_topics":90,"view_count":32,"oss_zip_url":75,"oss_zip_packed_at":75,"status":17,"created_at":94,"updated_at":95,"faqs":96,"releases":97},5253,"mli\u002Fpaper-reading","paper-reading","深度学习经典、新论文逐段精读","paper-reading 是一个专注于深度学习领域的论文精读项目，旨在通过视频形式逐段拆解经典与前沿的学术文章。面对大模型时代海量且晦涩的技术论文，许多学习者和从业者往往难以快速抓住核心逻辑与实现细节，paper-reading 正是为了解决这一痛点而生。它将复杂的理论推导、模型架构设计及训练基础设施等内容，转化为通俗易懂的视频讲解，帮助观众高效吸收知识。\n\n该项目特别适合人工智能研究人员、算法工程师、高校学生以及对深度学习有浓厚兴趣的开发者使用。无论是想深入理解 Llama 3.1 的训练全过程、探究 Sora 作为世界模拟器的原理，还是回顾 GPT-4 的技术突破，都能在这里找到系统化的解读内容。其独特亮点在于“逐段精读”的深度模式，不仅覆盖模型结构、预训练数据等关键技术点，还包含科研思路分享，部分热门视频已在 B 站和 YouTube 获得广泛关注。对于希望在有限时间内掌握高质量技术内容的用户而言，paper-reading 提供了一条清晰、可靠的学习路径。","# 深度学习论文精读\n\n## 录制完成的论文\n\n| 日期 | 标题 | 封面 | 时长 | 视频（播放数） |\n| --: | -- | -- | --: | -- |\n| 1\u002F10\u002F25 | [OpenAI Sora](https:\u002F\u002Fopenai.com\u002Findex\u002Fvideo-generation-models-as-world-simulators\u002F) 上\u003Cbr \u002F>(包含Movie Gen和HunyuanVideo) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_f047496f2903.jpg\" width=\"200px\"\u002F> | 1:04:18 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1VdcxesEAt)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1VdcxesEAt\u002F?share_source=copy_web&vd_source=5d037e935914fc22e2e978cdccf5cdfe)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F5MGq7dSOghY?style=social)](https:\u002F\u002Fyoutu.be\u002F5MGq7dSOghY?si=lY-OsadDsTeKf-ub)  |\n| 9\u002F04\u002F24 | Llama 3.1论文精读 · 5. 模型训练过程 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_fc3deb80ee56.jpg\" width=\"200px\"\u002F> | 10:41| [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1c8HbeaEXi)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1c8HbeaEXi)\u003Cbr \u002F>  |\n| 8\u002F28\u002F24 | Llama 3.1论文精读 · 4. 训练infra | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_b0ceaf0a2286.webp\" width=\"200px\"\u002F> | 25:04| [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1b4421f7fa)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1b4421f7fa)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F6XidEHVjS1A?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=6XidEHVjS1A)  |\n| 8\u002F13\u002F24 | Llama 3.1论文精读 · 3. 模型 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_8432575b5da9.webp\" width=\"200px\"\u002F> | 26:14| [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Q4421Z7Tj)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Q4421Z7Tj)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FG6gF-5g1Gg4?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=G6gF-5g1Gg4)  |\n| 8\u002F05\u002F24 | [Llama 3.1论文精读 · 2. 预训练数据](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2407.21783) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_26c789c1391d.jpg\" width=\"200px\"\u002F> | 23:37| [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1u142187S5)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1u142187S5)[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FwXFr3zIE8FM?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=wXFr3zIE8FM)|\n| 7\u002F31\u002F24 | Llama 3.1论文精读 · 1. 导言 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_a85efbc1858a.jpg\" width=\"200px\"\u002F> | 18:53| [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1WM4m1y7Uh)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1WM4m1y7Uh)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F-PztagF3wQE?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=-PztagF3wQE)  |\n| 3\u002F30\u002F23 | [GPT-4](https:\u002F\u002Fopenai.com\u002Fresearch\u002Fgpt-4) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_6595c171965e.jpg\" width=\"200px\"\u002F> | 1:20:38 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1vM4y1U7b5)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1vM4y1U7b5)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FK0SZ9mdygTw?style=social)](https:\u002F\u002Fyoutu.be\u002FK0SZ9mdygTw)  |\n| 3\u002F23\u002F23 | 大模型时代下做科研的四个思路 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_024dacda1a7d.jpg\" width=\"200px\"\u002F> | 1:06:29 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1oX4y1d7X6)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1oX4y1d7X6)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fsh79Z8i15PI?style=social)](https:\u002F\u002Fyoutu.be\u002Fsh79Z8i15PI) |\n| 3\u002F10\u002F23 | [Anthropic LLM](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2204.05862.pdf) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_480c62e2b3cb.jpg\" width=\"200px\"\u002F> | 1:01:51 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1XY411B7nM)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1XY411B7nM)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FiqX0pgNDon0?style=social)](https:\u002F\u002Fyoutu.be\u002FiqX0pgNDon0) |\n| 1\u002F20\u002F23 | [Helm](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2211.09110.pdf) 全面语言模型评测 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_9923f5a739af.jpg\" width=\"200px\"\u002F> | 1:23:37 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1z24y1B7uX)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1z24y1B7uX)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FWgFEw9U3BXA?style=social)](https:\u002F\u002Fyoutu.be\u002FWgFEw9U3BXA) |\n| 1\u002F11\u002F23 | 多模态论文串讲·下 |  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_a432406afc4b.jpg\" width=\"200px\"\u002F> | 1:03:29 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1fA411Z772)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1fA411Z772) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FS1le41J76lQ?style=social)](https:\u002F\u002Fyoutu.be\u002FS1le41J76lQ) |\n| 12\u002F29\u002F22 | [Instruct GPT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2203.02155.pdf) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_020031862d7c.jpg\" width=\"200px\"\u002F> | 1:07:10 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1hd4y187CR)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1hd4y187CR) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FzfIGAwD1jOQ?style=social)](https:\u002F\u002Fyoutu.be\u002FzfIGAwD1jOQ) |\n| 12\u002F19\u002F22 | [Neural Corpus Indexer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2206.02743.pdf) 文档检索 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_94fcccfd8f01.jpg\" width=\"200px\"\u002F> | 55:47 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Se411w7Sn)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Se411w7Sn) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FQRffZMSGJyU?style=social)](https:\u002F\u002Fyoutu.be\u002FQRffZMSGJyU) |\n| 12\u002F12\u002F22 | 多模态论文串讲·上 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_a23028e6932d.jpg\" width=\"200px\"\u002F> | 1:12:27 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Vd4y1v77v)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Vd4y1v77v) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F6pzBOQAXUB8?style=social)](https:\u002F\u002Fyoutu.be\u002F6pzBOQAXUB8)  |\n| 11\u002F14\u002F22 | [OpenAI Whisper](https:\u002F\u002Fcdn.openai.com\u002Fpapers\u002Fwhisper.pdf) 精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_8040040b6d46.jpg\" width=\"200px\"\u002F> | 1:12:16 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1VG4y1t74x)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1VG4y1t74x) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F3eXCJd32UnM?style=social)](https:\u002F\u002Fyoutu.be\u002F3eXCJd32UnM) |\n| 11\u002F07\u002F22 | 在讲 OpenAI Whisper 前先做了一个剪视频小工具 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_0e2f637177e3.jpg\" width=\"200px\"\u002F> | 23:39 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Pe4y1t7de)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Pe4y1t7de) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FPwVlvCPDnrI?style=social)](https:\u002F\u002Fyoutu.be\u002FPwVlvCPDnrI)  |\n| 10\u002F23\u002F22 | [Chain of Thought](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2201.11903.pdf) 论文、代码和资源 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_2891c6ab6cf9.jpg\" width=\"200px\"\u002F> | 33:21 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1t8411e7Ug)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1t8411e7Ug)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FH4J59iG3t5o?style=social)](https:\u002F\u002Fyoutu.be\u002FH4J59iG3t5o) |\n| 9\u002F17\u002F22 | CLIP 改进工作串讲（下） | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_3bd53e33b5fe.jpg\" width=\"200px\"\u002F> | 1:04:26 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1gg411U7n4)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1gg411U7n4)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FugJeBivv65s?style=social)](https:\u002F\u002Fyoutu.be\u002FugJeBivv65s) |\n| 9\u002F2\u002F22 | CLIP 改进工作串讲（上） | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_f45d79988447.jpg\" width=\"200px\"\u002F> | 1:14:43 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1FV4y1p7Lm)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1FV4y1p7Lm)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fx4CDhZz_Dvg?style=social)](https:\u002F\u002Fyoutu.be\u002Fx4CDhZz_Dvg) |\n| 7\u002F29\u002F22 | [ViLT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2102.03334.pdf) 论文精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_e15b97fc3e8d.jpg\" width=\"200px\"\u002F> | 1:03:26 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV14r4y1j74y)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV14r4y1j74y)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fug8YvZOjOCE?style=social)](https:\u002F\u002Fyoutu.be\u002Fug8YvZOjOCE) |\n| 7\u002F22\u002F22 | 理由、论据和担保【[研究的艺术](https:\u002F\u002Fpress.uchicago.edu\u002Fucp\u002Fbooks\u002Fbook\u002Fchicago\u002FC\u002Fbo23521678.html)·四】 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_7b41e03a2154.jpg\" width=\"200px\"\u002F> | 44:14 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1SB4y1a75c)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1SB4y1a75c) |\n| 7\u002F15\u002F22 | 如何讲好故事、故事里的论点【[研究的艺术](https:\u002F\u002Fpress.uchicago.edu\u002Fucp\u002Fbooks\u002Fbook\u002Fchicago\u002FC\u002Fbo23521678.html)·三】| \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_ece7b5076813.jpg\" width=\"200px\"\u002F> | 43:56 |[![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1WB4y1v7ST)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1WB4y1v7ST)|\n| 7\u002F8\u002F22 | [DALL·E 2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2204.06125.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_fdf447b70ea7.jpg\" width=\"200px\"\u002F> | 1:27:54 |[![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV17r4y1u77B)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV17r4y1u77B)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FhO57mntSMl0?style=social)](https:\u002F\u002Fyoutu.be\u002FhO57mntSMl0)|\n| 7\u002F1\u002F22 | 明白问题的重要性【[研究的艺术](https:\u002F\u002Fpress.uchicago.edu\u002Fucp\u002Fbooks\u002Fbook\u002Fchicago\u002FC\u002Fbo23521678.html)·二】| \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_08b16fc18f81.jpg\" width=\"200px\"\u002F> | 1:03:40 |[![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV11S4y1v7S2)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV11S4y1v7S2\u002F)|\n| 6\u002F24\u002F22 | 跟读者建立联系【[研究的艺术](https:\u002F\u002Fpress.uchicago.edu\u002Fucp\u002Fbooks\u002Fbook\u002Fchicago\u002FC\u002Fbo23521678.html)·一】 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_001d3d2b77bc.jpg\" width=\"200px\"\u002F> | 45:01 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1hY411T7vy)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1hY411T7vy\u002F) |\n| 6\u002F17\u002F22 | [Zero](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1910.02054.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_7f007a07c22a.jpg\" width=\"200px\"\u002F> | 52:21 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1tY411g7ZT)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1tY411g7ZT\u002F) |\n| 6\u002F10\u002F22 | [DETR](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2005.12872.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_847fd1574cf7.jpg\" width=\"200px\"\u002F> | 54:22 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1GB4y1X72R)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1GB4y1X72R\u002F) |\n| 6\u002F3\u002F22 | [Megatron LM](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1909.08053.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_735a66dffc84.jpg\" width=\"200px\"\u002F> | 56:07 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1nB4y1R7Yz)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1nB4y1R7Yz\u002F) |\n| 5\u002F27\u002F22 | [GPipe](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2019\u002Ffile\u002F093f65e080a295f8076b1c5722a46aa2-Paper.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_50ca82d49de5.jpg\" width=\"200px\"\u002F> | 58:47 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1v34y1E7zu)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1v34y1E7zu\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FeXjRpS_BTbs?style=social)](https:\u002F\u002Fyoutu.be\u002FeXjRpS_BTbs)  |\n| 5\u002F5\u002F22 | [Pathways](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2203.12533.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_dc63f64be22a.jpg\" width=\"200px\"\u002F> | 1:02:13 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1xB4y1m7Xi)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1xB4y1m7Xi\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F8hS1ZtgG0wU?style=social)](https:\u002F\u002Fyoutu.be\u002F8hS1ZtgG0wU) |\n| 4\u002F28\u002F22 | [视频理解论文串讲](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2012.06567.pdf)（下） | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_3dd7beab0554.jpg\" width=\"200px\"\u002F> | 1:08:32 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV11Y411P7ep)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV11Y411P7ep\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FJ2YC0-k57NM?style=social)](https:\u002F\u002Fyoutu.be\u002FJ2YC0-k57NM) |\n| 4\u002F21\u002F22 | [参数服务器（Parameter Server）](https:\u002F\u002Fwww.usenix.org\u002Fsystem\u002Ffiles\u002Fconference\u002Fosdi14\u002Fosdi14-paper-li_mu.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_08556a3c4ef8.jpg\" width=\"200px\"\u002F> | 1:37:40 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1YA4y197G8)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1YA4y197G8\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fxt-AwUrDxQk?style=social)](https:\u002F\u002Fyoutu.be\u002Fxt-AwUrDxQk) |\n| 4\u002F14\u002F22 | [视频理解论文串讲](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2012.06567.pdf)（上） | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_2d8297a7a75b.jpg\" width=\"200px\"\u002F> | 51:15 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1fL4y157yA)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1fL4y157yA\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FgK7AGO6okhc?style=social)](https:\u002F\u002Fyoutu.be\u002FgK7AGO6okhc) |\n| 3\u002F31\u002F22 | [I3D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1705.07750.pdf) 论文精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_1c7a03f63be0.jpg\" width=\"200px\"\u002F> | 52:31 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1tY4y1p7hq)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1tY4y1p7hq\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F9lIkKiAn6uE?style=social)](https:\u002F\u002Fyoutu.be\u002F9lIkKiAn6uE) |\n| 3\u002F24\u002F22 | 斯坦福 2022 年 [AI 指数报告](https:\u002F\u002Faiindex.stanford.edu\u002Fwp-content\u002Fuploads\u002F2022\u002F03\u002F2022-AI-Index-Report_Master.pdf) 精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_8db0fddec393.jpg\" width=\"200px\"\u002F> | 1:19:56 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1s44y1N7eu)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1s44y1N7eu\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FK8h_xjQ6ufY?style=social)](https:\u002F\u002Fyoutu.be\u002FK8h_xjQ6ufY) |\n| 3\u002F17\u002F22 | [AlphaCode](https:\u002F\u002Fstorage.googleapis.com\u002Fdeepmind-media\u002FAlphaCode\u002Fcompetition_level_code_generation_with_alphacode.pdf) 论文精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_22dc2b1b039c.jpg\" width=\"200px\"\u002F> | 44:00 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1ab4y1s7rc)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1ab4y1s7rc\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Ft8Gzkca9pW4?style=social)](https:\u002F\u002Fyoutu.be\u002Ft8Gzkca9pW4) |\n| 3\u002F10\u002F22 | [OpenAI Codex](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2107.03374.pdf) 论文精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_83c68e6d2ef4.jpg\" width=\"200px\"\u002F> | 47:58 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1iY41137Zi)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1iY41137Zi\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1490959755963666432)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1490959755963666432)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FoZriUGkQSNM?style=social)](https:\u002F\u002Fyoutu.be\u002FoZriUGkQSNM) |\n| 3\u002F3\u002F22 | [GPT](https:\u002F\u002Fs3-us-west-2.amazonaws.com\u002Fopenai-assets\u002Fresearch-covers\u002Flanguage-unsupervised\u002Flanguage_understanding_paper.pdf), [GPT-2](https:\u002F\u002Fd4mucfpksywv.cloudfront.net\u002Fbetter-language-models\u002Flanguage_models_are_unsupervised_multitask_learners.pdf), [GPT-3](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.14165) 精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_20ff972ace3b.jpg\" width=\"200px\"\u002F> | 1:29:58 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1AF411b7xQ)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1AF411b7xQ\u002F)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Ft70Bl3w7bxY?style=social)](https:\u002F\u002Fyoutu.be\u002Ft70Bl3w7bxY) |\n| 2\u002F24\u002F22 | [Two-Stream](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2014\u002Ffile\u002F00ec53c4682d36f5c4359f4ae7bd7ba1-Paper.pdf) 逐段精读 |  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_1a2750415c48.jpg\" width=\"200px\"\u002F> | 52:57 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1mq4y1x7RU)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1mq4y1x7RU\u002F)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FvuqwKP2iDe0?style=social)](https:\u002F\u002Fyoutu.be\u002FvuqwKP2iDe0) |\n| 2\u002F10\u002F22 | [CLIP](https:\u002F\u002Fopenai.com\u002Fblog\u002Fclip\u002F) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_1be3ca6c3a75.jpg\" width=\"200px\"\u002F> | 1:38:25 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1SL4y1s7LQ)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1SL4y1s7LQ\u002F)\u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1475706654562299904)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1475706654562299904) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FOZF1t_Hieq8?style=social)](https:\u002F\u002Fyoutu.be\u002FOZF1t_Hieq8) |\n| 2\u002F6\u002F22 | 你（被）吐槽过[论文不够 novel](https:\u002F\u002Fperceiving-systems.blog\u002Fen\u002Fpost\u002Fnovelty-in-science) 吗？| \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_35889677b888.jpg\" width=\"200px\"\u002F> | 14:11 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1ea41127Bq)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1ea41127Bq\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1475719090198876161)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1475719090198876161) |\n| 1\u002F23\u002F22 | [AlphaFold 2](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs41586-021-03819-2.pdf) 精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_7c1c47dd9e8b.jpg\" width=\"200px\"\u002F> |  1:15:28 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1oR4y1K7Xr)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1oR4y1K7Xr\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1469132410537717760)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1469132410537717760)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FOy3OCoGUr-w?style=social)](https:\u002F\u002Fyoutu.be\u002FOy3OCoGUr-w) |\n| 1\u002F18\u002F22 | 如何判断（你自己的）研究工作的价值 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_63e5502f2633.jpg\" width=\"200px\"\u002F> |  9:59 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1oL411c7Us)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1oL411c7Us\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1475716940051869696)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1475716940051869696) |\n| 1\u002F15\u002F22 | [Swin Transformer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.14030.pdf) 精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_81dd84403994.jpg\" width=\"200px\"\u002F> | 1:00:21 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV13L4y1475U)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV13L4y1475U\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1466282983652691968)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1466282983652691968)   \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FluP3-Fs0QCo?style=social)](https:\u002F\u002Fyoutu.be\u002FluP3-Fs0QCo) |\n| 1\u002F7\u002F22 | [指导数学直觉](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs41586-021-04086-x.pdf) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_d39d05b9e0d0.jpg\" width=\"200px\"\u002F> | 52:51 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1YZ4y1S72j)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1YZ4y1S72j\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1464060386375299072)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1464060386375299072)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FczFGjvhtss8?style=social)](https:\u002F\u002Fyoutu.be\u002FczFGjvhtss8) |\n| 1\u002F5\u002F22 | AlphaFold 2 预告 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_b2a0459aca0e.jpg\" width=\"200px\"\u002F> | 03:28 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Eu411U7Te)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Eu411U7Te\u002F) |\n| 12\u002F20\u002F21 | [对比学习](#contrastive_learning)论文综述 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_12633fdc9471.jpg\" width=\"200px\"\u002F> | 1:32:01 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV19S4y1M7hm)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV19S4y1M7hm\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1460828005077164032)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1460828005077164032)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F1pvxufGRuW4?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=1pvxufGRuW4) |\n| 12\u002F15\u002F21 | [MoCo](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1911.05722.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_25987fa17019.jpg\" width=\"200px\"\u002F> | 1:24:11 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1C3411s7t9)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1C3411s7t9\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1454723120678936576)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1454723120678936576)   \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F1pvxufGRuW4?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=1pvxufGRuW4) |\n| 12\u002F9\u002F21 | 如何找研究想法 1 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_80c06ea1f09b.jpg\" width=\"200px\"\u002F> | 5:34 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1qq4y1z7F2)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1qq4y1z7F2\u002F) |\n| 12\u002F8\u002F21 | [MAE](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2111.06377.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_c9c4c48e57ec.jpg\" width=\"200px\"\u002F> | 47:04 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1sq4y1q77t)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1sq4y1q77t\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1452458167968251904)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1452458167968251904)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FmYlX2dpdHHM?style=social)](https:\u002F\u002Fyoutu.be\u002FmYlX2dpdHHM) |\n| 11\u002F29\u002F21 | [ViT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2010.11929.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_e3541b53dea0.jpg\" width=\"200px\"\u002F> | 1:11:30 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV15P4y137jb)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV15P4y137jb\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1449195245754380288)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1449195245754380288)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FFRFt3x0bO94?style=social)](https:\u002F\u002Fyoutu.be\u002FFRFt3x0bO94) |\n| 11\u002F18\u002F21 | [BERT](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.04805) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_3a938b86f843.jpg\" width=\"200px\"\u002F> | 45:49  | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1PL411M7eQ)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1PL411M7eQ\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1445340200976785408)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1445340200976785408)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FULD3uIb2MHQ?style=social)](https:\u002F\u002Fyoutu.be\u002FULD3uIb2MHQ) |\n| 11\u002F9\u002F21 | [GAN](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2014\u002Ffile\u002F5ca3e9b122f61f8f06494c97b1afccf3-Paper.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_99eeaac972eb.jpg\" width=\"200px\"\u002F> | 46:16  | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1rb4y187vD)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1rb4y187vD\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1442091389241159681)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1442091389241159681)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fg_0HtlrLiDo?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=g_0HtlrLiDo) |\n| 11\u002F3\u002F21 | 零基础多图详解 [图神经网络](https:\u002F\u002Fdistill.pub\u002F2021\u002Fgnn-intro\u002F)（GNN\u002FGCN） | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_f2552664a7c0.jpg\" width=\"200px\"\u002F> | 1:06:19 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1iT4y1d7zP)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1iT4y1d7zP\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1439540657619087360)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1439540657619087360)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FsejA2PtCITw?style=social)](https:\u002F\u002Fyoutu.be\u002FsejA2PtCITw) |\n| 10\u002F27\u002F21 | [Transformer](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.03762) 逐段精读\u003Cbr> （视频中提到的文献 [^transformer]) |\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_e7ec6347400b.jpg\" width=\"200px\"\u002F> | 1:27:05 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1pu411o7BE)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1pu411o7BE\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1437034536677404672)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1437034536677404672)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FnzqlFIcCSWQ?style=social)](https:\u002F\u002Fyoutu.be\u002FnzqlFIcCSWQ) |\n| 10\u002F22\u002F21 | [ResNet](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.03385) 论文逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_22e4b07b303b.jpg\" width=\"200px\"\u002F> | 53:46 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1P3411y7nn)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1P3411y7nn\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1434795406001180672)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1434795406001180672)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FpWMnzCX4cwQ?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=pWMnzCX4cwQ) |\n| 10\u002F21\u002F21 | 撑起计算机视觉半边天的 [ResNet](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.03385) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_081a2b3fb354.jpg\" width=\"200px\"\u002F> | 11:50 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Fb4y1h73E)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Fb4y1h73E\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1434787226101751808)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1434787226101751808)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FNnSldWhSqvY?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=NnSldWhSqvY) |\n| 10\u002F15\u002F21 | [AlexNet](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2012\u002Ffile\u002Fc399862d3b9d6b76c8436e924a68c45b-Paper.pdf) 论文逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_554ce4eea08a.jpg\" width=\"200px\"\u002F> | 55:21 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1hq4y157t1)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1hq4y157t1\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1432354207483871232)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1432354207483871232)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FwYmlILPsLlY?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=wYmlILPsLlY) |\n| 10\u002F14\u002F21 | 9年后重读深度学习奠基作之一：[AlexNet](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2012\u002Ffile\u002Fc399862d3b9d6b76c8436e924a68c45b-Paper.pdf) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_42196dca7f17.jpg\" width=\"200px\"\u002F> | 19:59 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1ih411J7Kz)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1ih411J7Kz\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1432155856322920448)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1432155856322920448)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FvdYH0fE6thY?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=vdYH0fE6thY) |\n| 10\u002F06\u002F21 | 如何读论文 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_18ece7c81b03.jpg\" width=\"200px\"\u002F> | 06:39 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1H44y1t75x)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1H44y1t75x\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1428973951632969728)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1428973951632969728)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Ftxjl_Q4jCyQ?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=txjl_Q4jCyQ&list=PLFXJ6jwg0qW-7UM8iUTj3qKqdhbQULP5I&index=1) |\n\n[^transformer]: 1 [斯坦福100+作者的200+页综述](https:\u002F\u002Farxiv.org\u002Fabs\u002F2108.07258)，2 [对LayerNorm的新研究](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1911.07013.pdf)，3 [对Attention在Transformer里面作用的研究](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.03404)\n\n\n## 所有论文\n\n包括已经录制完成和之后将要介绍的论文。选取的原则是10年内深度学习里有影响力文章（必读文章），或者近期比较有意思的文章。当然这十年里重要的工作太多了，不可能一一过一遍。在选取的时候我会偏向一些之前 [直播课](https:\u002F\u002Fc.d2l.ai\u002Fzh-v2\u002F) 中没讲到过的。 欢迎大家在 [讨论区](https:\u002F\u002Fgithub.com\u002Fmli\u002Fpaper-reading\u002Fdiscussions) 里提供建（点）议（歌）。\n\n总论文数 67，录制完成数 32\n\n（这里引用采用的是 semanticscholar，是因为它提供 [API](https:\u002F\u002Fapi.semanticscholar.org\u002Fapi-docs\u002Fgraph#operation\u002Fget_graph_get_paper) 可以自动获取，不用手动更新。）\n\n### 计算机视觉 - CNN\n\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅      | 2012 | [AlexNet](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2012\u002Ffile\u002Fc399862d3b9d6b76c8436e924a68c45b-Paper.pdf) | 深度学习热潮的奠基作                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fabd1c342495432171beb7ca8fd9551ef13cbd0ff%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImageNet-classification-with-deep-convolutional-Krizhevsky-Sutskever\u002Fabd1c342495432171beb7ca8fd9551ef13cbd0ff) |\n| | 2014 | [VGG](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1409.1556.pdf) | 使用 3x3 卷积构造更深的网络                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Feb42cf88027de515750f230b23b1a057dc782108%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FVery-Deep-Convolutional-Networks-for-Large-Scale-Simonyan-Zisserman\u002Feb42cf88027de515750f230b23b1a057dc782108) |\n| | 2014 | [GoogleNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1409.4842.pdf) | 使用并行架构构造更深的网络                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fe15cf50aa89fee8535703b9f9512fca5bfc43327%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FGoing-deeper-with-convolutions-Szegedy-Liu\u002Fe15cf50aa89fee8535703b9f9512fca5bfc43327) |\n|  ✅  | 2015 |  [ResNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1512.03385.pdf) | 构建深层网络都要有的残差连接。               |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F2c03df8b48bf3fa39054345bafabfeff15bfd11d%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FDeep-Residual-Learning-for-Image-Recognition-He-Zhang\u002F2c03df8b48bf3fa39054345bafabfeff15bfd11d)  |\n|  | 2017 | [MobileNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1704.04861.pdf) | 适合终端设备的小CNN                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F3647d6d0f151dc05626449ee09cc7bce55be497e%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMobileNets%3A-Efficient-Convolutional-Neural-Networks-Howard-Zhu\u002F3647d6d0f151dc05626449ee09cc7bce55be497e)  |\n| | 2019 | [EfficientNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1905.11946.pdf) | 通过架构搜索得到的CNN                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F4f2eda8077dc7a69bb2b4e0a1a086cf054adb3f9%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FEfficientNet%3A-Rethinking-Model-Scaling-for-Neural-Tan-Le\u002F4f2eda8077dc7a69bb2b4e0a1a086cf054adb3f9)  |\n| | 2021 |  [Non-deep networks](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2110.07641.pdf) | 让不深的网络也能在ImageNet刷到SOTA                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F0d7f6086772079bc3e243b7b375a9ca1a517ba8b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FNon-deep-Networks-Goyal-Bochkovskiy\u002F0d7f6086772079bc3e243b7b375a9ca1a517ba8b) |\n\n### 计算机视觉 - Transformer\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅ | 2020 | [ViT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2010.11929.pdf) | Transformer杀入CV界                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F7b15fa1b8d413fbe14ef7a97f651f47f5aff3903%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAn-Image-is-Worth-16x16-Words%3A-Transformers-for-at-Dosovitskiy-Beyer\u002F7b15fa1b8d413fbe14ef7a97f651f47f5aff3903)  |\n| ✅ | 2021 | [Swin Transformer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.14030.pdf) | 多层次的Vision Transformer                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc8b25fab5608c3e033d34b4483ec47e68ba109b7%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FSwin-Transformer%3A-Hierarchical-Vision-Transformer-Liu-Lin\u002Fc8b25fab5608c3e033d34b4483ec47e68ba109b7) |\n| | 2021 | [MLP-Mixer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2105.01601.pdf) | 使用MLP替换self-attention            |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F2def61f556f9a5576ace08911496b7c7e4f970a4%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMLP-Mixer%3A-An-all-MLP-Architecture-for-Vision-Tolstikhin-Houlsby\u002F2def61f556f9a5576ace08911496b7c7e4f970a4)  |\n| ✅ | 2021 | [MAE](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2111.06377.pdf) | BERT的CV版             |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc1962a8cf364595ed2838a097e9aa7cd159d3118%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMasked-Autoencoders-Are-Scalable-Vision-Learners-He-Chen\u002Fc1962a8cf364595ed2838a097e9aa7cd159d3118)  |\n\n### 生成模型\n\n| 已录制 | 年份 | 名字                                              | 简介         | 引用 |\n| ------ | ---- | ------------------------------------------------- | ------------ | ------------------------------------------------------------ |\n|  ✅ | 2014 | [GAN](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2014\u002Ffile\u002F5ca3e9b122f61f8f06494c97b1afccf3-Paper.pdf) | 生成模型的开创工作                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F54e325aee6b2d476bbbb88615ac15e251c6e8214%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FGenerative-Adversarial-Nets-Goodfellow-Pouget-Abadie\u002F54e325aee6b2d476bbbb88615ac15e251c6e8214)  |\n|  | 2015 | [DCGAN](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1511.06434.pdf) | 使用CNN的GAN          |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F8388f1be26329fa45e5807e968a641ce170ea078%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnsupervised-Representation-Learning-with-Deep-Radford-Metz\u002F8388f1be26329fa45e5807e968a641ce170ea078)  |\n|  | 2016 | [pix2pix](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1611.07004.pdf) |           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F8acbe90d5b852dadea7810345451a99608ee54c7%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImage-to-Image-Translation-with-Conditional-Isola-Zhu\u002F8acbe90d5b852dadea7810345451a99608ee54c7)  |\n|  | 2016 | [SRGAN](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1609.04802.pdf) | 图片超分辨率          |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fdf0c54fe61f0ffb9f0e36a17c2038d9a1964cba3%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FPhoto-Realistic-Single-Image-Super-Resolution-Using-Ledig-Theis\u002Fdf0c54fe61f0ffb9f0e36a17c2038d9a1964cba3)  |\n|  | 2017 | [WGAN](https:\u002F\u002Farxiv.org\u002Fabs\u002F1701.07875) | 训练更加容易          |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F2f85b7376769473d2bed56f855f115e23d727094%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FWasserstein-GAN-Arjovsky-Chintala\u002F2f85b7376769473d2bed56f855f115e23d727094)  |\n|  | 2017 | [CycleGAN](https:\u002F\u002Farxiv.org\u002Fabs\u002F1703.10593) |           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc43d954cf8133e6254499f3d68e45218067e4941%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnpaired-Image-to-Image-Translation-Using-Networks-Zhu-Park\u002Fc43d954cf8133e6254499f3d68e45218067e4941)  |\n|  | 2018 | [StyleGAN](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.04948) |           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fceb2ebef0b41e31c1a21b28c2734123900c005e2%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FA-Style-Based-Generator-Architecture-for-Generative-Karras-Laine\u002Fceb2ebef0b41e31c1a21b28c2734123900c005e2)  |\n| | 2019 | [StyleGAN2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1912.04958.pdf) |        |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Ff3e3d1f86a534a3654d0ee263142e44f4e2c61e9%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAnalyzing-and-Improving-the-Image-Quality-of-Karras-Laine\u002Ff3e3d1f86a534a3654d0ee263142e44f4e2c61e9)  |\n| | 2020 | [DDPM](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2006.11239.pdf) | Diffusion Models   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F289db3be7bf77e06e75541ba93269de3d604ac72%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FDenoising-Diffusion-Probabilistic-Models-Ho-Jain\u002F289db3be7bf77e06e75541ba93269de3d604ac72)  |\n| | 2021 | [Improved DDPM](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2102.09672.pdf) | 改进的 DDPM   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fde18baa4964804cf471d85a5a090498242d2e79f%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImproved-Denoising-Diffusion-Probabilistic-Models-Nichol-Dhariwal\u002Fde18baa4964804cf471d85a5a090498242d2e79f)  |\n| | 2021 | [Guided Diffusion Models](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2105.05233.pdf) | 号称超越 GAN  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F64ea8f180d0682e6c18d1eb688afdb2027c02794%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FDiffusion-Models-Beat-GANs-on-Image-Synthesis-Dhariwal-Nichol\u002F64ea8f180d0682e6c18d1eb688afdb2027c02794)  |\n| | 2021 | [StyleGAN3](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2106.12423.pdf) |        |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc1ff08b59f00c44f34dfdde55cd53370733a2c19%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAlias-Free-Generative-Adversarial-Networks-Karras-Aittala\u002Fc1ff08b59f00c44f34dfdde55cd53370733a2c19)  |\n|  ✅  | 2022 | [DALL.E 2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2204.06125.pdf) | CLIP + Diffusion models，文本生成图像新高度     |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc57293882b2561e1ba03017902df9fc2f289dea2%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FHierarchical-Text-Conditional-Image-Generation-with-Ramesh-Dhariwal\u002Fc57293882b2561e1ba03017902df9fc2f289dea2)  |\n|  ✅  | 2024 | [Sora](https:\u002F\u002Fopenai.com\u002Findex\u002Fvideo-generation-models-as-world-simulators\u002F) | 开启视频生成热潮     |  |\n|  ✅  | 2024 | [Movie Gen](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2410.13720) | 精确的文本指导视频编辑、个性化视频生成     |  |\n|  ✅  | 2025 | [HunyuanVideo](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2412.03603) | 开源视频生成框架     |  |\n\n### 计算机视觉 - Object Detection\n\n| 已录制 | 年份 | 名字                                              | 简介         | 引用 |\n| ------ | ---- | ------------------------------------------------- | ------------ | ------------------------------------------------------------ |\n|        | 2014 | [R-CNN](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1311.2524v5.pdf)    | Two-stage             |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F2f4df08d9072fc2ac181b7fced6a245315ce05c8%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F2f4df08d9072fc2ac181b7fced6a245315ce05c8)  |\n|        | 2015 | [Fast R-CNN](http:\u002F\u002Farxiv.org\u002Fabs\u002F1504.08083v2)   |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F7ffdbc358b63378f07311e883dddacc9faeeaf4b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F7ffdbc358b63378f07311e883dddacc9faeeaf4b)  |\n|        | 2015 | [Faster R-CNN](http:\u002F\u002Farxiv.org\u002Fabs\u002F1506.01497v3) |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F424561d8585ff8ebce7d5d07de8dbf7aae5e7270%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F424561d8585ff8ebce7d5d07de8dbf7aae5e7270)  |\n|        | 2016 | [SSD](http:\u002F\u002Farxiv.org\u002Fabs\u002F1512.02325v5)          | Single stage          |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F4d7a9197433acbfb24ef0e9d0f33ed1699e4a5b0%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F4d7a9197433acbfb24ef0e9d0f33ed1699e4a5b0)  |\n|        | 2016 | [YOLO](http:\u002F\u002Farxiv.org\u002Fabs\u002F1506.02640v5)         |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Ff8e79ac0ea341056ef20f2616628b3e964764cfd%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002Ff8e79ac0ea341056ef20f2616628b3e964764cfd)  |\n|        | 2017 | [Mask R-CNN](http:\u002F\u002Farxiv.org\u002Fabs\u002F1703.06870v3)   |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fea99a5535388196d0d44be5b4d7dd02029a43bb2%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002Fea99a5535388196d0d44be5b4d7dd02029a43bb2)  |\n|        | 2017 | [YOLOv2](http:\u002F\u002Farxiv.org\u002Fabs\u002F1612.08242v1)       |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F7d39d69b23424446f0400ef603b2e3e22d0309d6%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F7d39d69b23424446f0400ef603b2e3e22d0309d6)  |\n|        | 2018 | [YOLOv3](http:\u002F\u002Farxiv.org\u002Fabs\u002F1804.02767v1)       |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fe4845fb1e624965d4f036d7fd32e8dcdd2408148%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002Fe4845fb1e624965d4f036d7fd32e8dcdd2408148)  |\n|        | 2019 | [CenterNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1904.07850.pdf) | Anchor free           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F6a2e2fd1b5bb11224daef98b3fb6d029f68a73f2%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FObjects-as-Points-Zhou-Wang\u002F6a2e2fd1b5bb11224daef98b3fb6d029f68a73f2)  |\n|   ✅     | 2020 | [DETR](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2005.12872.pdf)      | Transformer           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F962dc29fdc3fbdc5930a10aba114050b82fe5a3e%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FEnd-to-End-Object-Detection-with-Transformers-Carion-Massa\u002F962dc29fdc3fbdc5930a10aba114050b82fe5a3e)  |\n\n\u003Ca name=\"contrastive_learning\">\u003C\u002Fa>\n\n### 计算机视觉 - 对比学习\n\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅      | 2018 | [InstDisc](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1805.01978.pdf) | 提出实例判别和memory bank做对比学习                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F155b7782dbd713982a4133df3aee7adfd0b6b304%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnsupervised-Feature-Learning-via-Non-parametric-Wu-Xiong\u002F155b7782dbd713982a4133df3aee7adfd0b6b304)  |\n| ✅      | 2018 | [CPC](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1807.03748.pdf) | 对比预测编码，图像语音文本强化学习全都能做                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fb227f3e4c0dc96e5ac5426b85485a70f2175a205%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FRepresentation-Learning-with-Contrastive-Predictive-Oord-Li\u002Fb227f3e4c0dc96e5ac5426b85485a70f2175a205) |\n| ✅      | 2019 | [InvaSpread](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1904.03436.pdf) | 一个编码器的端到端对比学习                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fe4bde6fe33b6c2cf9d1647ac0b041f7d1ba29c5b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnsupervised-Embedding-Learning-via-Invariant-and-Ye-Zhang\u002Fe4bde6fe33b6c2cf9d1647ac0b041f7d1ba29c5b)  |\n| ✅  | 2019 |  [CMC](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1906.05849.pdf) | 多视角下的对比学习               |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F97f4d09175705be4677d675fa27e55defac44800%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FContrastive-Multiview-Coding-Tian-Krishnan\u002F97f4d09175705be4677d675fa27e55defac44800)  |\n| ✅ | 2019 | [MoCov1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1911.05722.pdf) | 无监督训练效果也很好                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fec46830a4b275fd01d4de82bffcabe6da086128f%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMomentum-Contrast-for-Unsupervised-Visual-Learning-He-Fan\u002Fec46830a4b275fd01d4de82bffcabe6da086128f) |\n|  ✅ | 2020 |  [SimCLRv1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2002.05709.pdf) |  简单的对比学习 (数据增强 + MLP head + 大batch训练久)                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F34733eaf66007516347a40ad5d9bbe1cc9dacb6b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FA-Simple-Framework-for-Contrastive-Learning-of-Chen-Kornblith\u002F34733eaf66007516347a40ad5d9bbe1cc9dacb6b)  |\n|  ✅ | 2020 | [MoCov2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2003.04297.pdf) | MoCov1 + improvements from SimCLRv1                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fa1b8a8df281bbaec148a897927a49ea47ea31515%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImproved-Baselines-with-Momentum-Contrastive-Chen-Fan\u002Fa1b8a8df281bbaec148a897927a49ea47ea31515)  |\n|  ✅ | 2020 |  [SimCLRv2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2006.10029.pdf) | 大的自监督预训练模型很适合做半监督学习                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F3e7f5f4382ac6f9c4fef6197dd21abf74456acd1%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FBig-Self-Supervised-Models-are-Strong-Learners-Chen-Kornblith\u002F3e7f5f4382ac6f9c4fef6197dd21abf74456acd1)  |\n| ✅  | 2020 |  [BYOL](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2006.07733.pdf) | 不需要负样本的对比学习                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F38f93092ece8eee9771e61c1edaf11b1293cae1b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FBootstrap-Your-Own-Latent%3A-A-New-Approach-to-Grill-Strub\u002F38f93092ece8eee9771e61c1edaf11b1293cae1b) |\n|  ✅ | 2020 |  [SWaV](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2006.09882.pdf) | 聚类对比学习                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F10161d83d29fc968c4612c9e9e2b61a2fc25842e%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnsupervised-Learning-of-Visual-Features-by-Cluster-Caron-Misra\u002F10161d83d29fc968c4612c9e9e2b61a2fc25842e) |\n|  ✅ | 2020 |  [SimSiam](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2011.10566.pdf) | 化繁为简的孪生表征学习                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F0e23d2f14e7e56e81538f4a63e11689d8ac1eb9d%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FExploring-Simple-Siamese-Representation-Learning-Chen-He\u002F0e23d2f14e7e56e81538f4a63e11689d8ac1eb9d)  |\n| ✅ | 2021 | [MoCov3](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2104.02057.pdf) | 如何更稳定的自监督训练ViT                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F739ceacfafb1c4eaa17509351b647c773270b3ae%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAn-Empirical-Study-of-Training-Self-Supervised-Chen-Xie\u002F739ceacfafb1c4eaa17509351b647c773270b3ae)  |\n|  ✅ | 2021 |  [DINO](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2104.14294.pdf) | transformer加自监督在视觉也很香                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fad4a0938c48e61b7827869e4ac3baffd0aefab35%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FEmerging-Properties-in-Self-Supervised-Vision-Caron-Touvron\u002Fad4a0938c48e61b7827869e4ac3baffd0aefab35)  |\n\n\n### 计算机视觉 - 视频理解\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅ | 2014 |  [DeepVideo](https:\u002F\u002Fcs.stanford.edu\u002Fpeople\u002Fkarpathy\u002Fdeepvideo\u002F) | 提出sports1M数据集，用深度学习做视频理解 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F6d4c9c923e9f145d1c01a2de2afc38ec23c44253%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLarge-Scale-Video-Classification-with-Convolutional-Karpathy-Toderici\u002F6d4c9c923e9f145d1c01a2de2afc38ec23c44253)  |\n| ✅ | 2014 |  [Two-stream](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1406.2199.pdf) | 引入光流做时序建模，神经网络首次超越手工特征 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F67dccc9a856b60bdc4d058d83657a089b8ad4486%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FTwo-Stream-Convolutional-Networks-for-Action-in-Simonyan-Zisserman\u002F67dccc9a856b60bdc4d058d83657a089b8ad4486)  |\n| ✅ | 2014 |  [C3D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1412.0767.pdf) |  比较深的3D-CNN做视频理解 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fd25c65d261ea0e6a458be4c50c40ffe5bc508f77%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLearning-Spatiotemporal-Features-with-3D-Networks-Tran-Bourdev\u002Fd25c65d261ea0e6a458be4c50c40ffe5bc508f77)  |\n| ✅ | 2015 |  [Beyond-short-snippets](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1503.08909.pdf) | 尝试使用LSTM  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F5418b2a482720e013d487a385c26fae0f017c6a6%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FBeyond-short-snippets%3A-Deep-networks-for-video-Ng-Hausknecht\u002F5418b2a482720e013d487a385c26fae0f017c6a6)  |\n| ✅ | 2016 |  [Convolutional fusion](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1604.06573.pdf) | 做early fusion来加强时空间建模    |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F9d9aced120e530484609164c836da64548693484%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FConvolutional-Two-Stream-Network-Fusion-for-Video-Feichtenhofer-Pinz\u002F9d9aced120e530484609164c836da64548693484)  |\n| ✅ | 2016 |  [TSN](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1608.00859.pdf) | 超级有效的视频分段建模，bag of tricks in video |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fea3d7de6c0880e14455b9acb28f1bc1234321456%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FTemporal-Segment-Networks%3A-Towards-Good-Practices-Wang-Xiong\u002Fea3d7de6c0880e14455b9acb28f1bc1234321456)  |\n| ✅ | 2017 |  [I3D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1705.07750.pdf) | 提出Kinetics数据集，膨胀2D网络到3D，开启3D-CNN时代  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fb61a3f8b80bbd44f24544dc915f52fd30bbdf485%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FQuo-Vadis%2C-Action-Recognition-A-New-Model-and-the-Carreira-Zisserman\u002Fb61a3f8b80bbd44f24544dc915f52fd30bbdf485)  |\n| ✅ | 2017 |  [R2+1D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1711.11248.pdf) | 拆分3D卷积核，使3D网络容易优化  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F89c3050522a0bb9820c32dc7444e003ef0d3e2e4%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FA-Closer-Look-at-Spatiotemporal-Convolutions-for-Tran-Wang\u002F89c3050522a0bb9820c32dc7444e003ef0d3e2e4)  |\n| ✅ | 2017 |  [Non-local](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1711.07971.pdf) | 引入自注意力做视觉问题  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F8899094797e82c5c185a0893896320ef77f60e64%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FNon-local-Neural-Networks-Wang-Girshick\u002F8899094797e82c5c185a0893896320ef77f60e64)  |\n| ✅ | 2018 |  [SlowFast](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1812.03982.pdf) | 快慢两支提升效率   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F8b47b9c3c35b2b2a78bff7822605b3040f87d699%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FSlowFast-Networks-for-Video-Recognition-Feichtenhofer-Fan\u002F8b47b9c3c35b2b2a78bff7822605b3040f87d699)  |\n| ✅ | 2021 |  [TimeSformer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2102.05095.pdf) | 视频中第一个引入transformer，开启video transformer时代 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc143ea9e30b1f2d93a9c060253845423f9e60e1f%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FIs-Space-Time-Attention-All-You-Need-for-Video-Bertasius-Wang\u002Fc143ea9e30b1f2d93a9c060253845423f9e60e1f)  |\n\n\n### 多模态学习\n\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅ | 2021 |  [CLIP](https:\u002F\u002Fopenai.com\u002Fblog\u002Fclip\u002F) | 图片和文本之间的对比学习                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F6f870f7f02a8c59c3e23f407f3ef00dd1dcf8fc4%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLearning-Transferable-Visual-Models-From-Natural-Radford-Kim\u002F6f870f7f02a8c59c3e23f407f3ef00dd1dcf8fc4)  |\n| ✅ | 2021 |  [ViLT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2102.03334.pdf) | 第一个摆脱了目标检测的视觉文本模型      |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F0839722fb5369c0abaff8515bfc08299efc790a1%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FViLT%3A-Vision-and-Language-Transformer-Without-or-Kim-Son\u002F0839722fb5369c0abaff8515bfc08299efc790a1)  |\n| ✅ | 2021 |  [ViLD](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2104.13921.pdf) | CLIP蒸馏帮助开集目标检测      |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fcf9b8da26d9b92e75ba49616ed2a1033f59fce14%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FOpen-vocabulary-Object-Detection-via-Vision-and-Gu-Lin\u002Fcf9b8da26d9b92e75ba49616ed2a1033f59fce14)  |\n| ✅ | 2021 |  [GLIP](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2112.03857.pdf) | 联合目标检测和文本定位           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F5341b412383c43f4a693ad63ec4489e3ec7688c8%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FGrounded-Language-Image-Pre-training-Li-Zhang\u002F5341b412383c43f4a693ad63ec4489e3ec7688c8)  |\n| ✅ | 2021 |  [CLIP4Clip](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2104.08860.pdf) | 拿CLIP直接做视频文本retrieval       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F281ad83e06d731d5d686acf07cd701576f1188c4%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FCLIP4Clip%3A-An-Empirical-Study-of-CLIP-for-End-to-Luo-Ji\u002F281ad83e06d731d5d686acf07cd701576f1188c4)  |\n| ✅ | 2021 |  [ActionCLIP](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2109.08472.pdf) | 用多模态对比学习有监督的做视频动作分类   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fdc05240a06326b5b1664f7e8c95c330b08cd0349%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FActionCLIP%3A-A-New-Paradigm-for-Video-Action-Wang-Xing\u002Fdc05240a06326b5b1664f7e8c95c330b08cd0349)  |\n| ✅ | 2021 |  [PointCLIP](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2112.02413.pdf) | 3D变2D，巧妙利用CLIP做点云  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Ff3ce9ba3fcec362b70263a7ed63d9404975496a0%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FPointCLIP%3A-Point-Cloud-Understanding-by-CLIP-Zhang-Guo\u002Ff3ce9ba3fcec362b70263a7ed63d9404975496a0)  |\n| ✅ | 2022 |  [LSeg](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2201.03546.pdf) | 有监督的开集分割                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fcc9826c222ac1e81b4b374dd9e0df130f298b1e8%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLanguage-driven-Semantic-Segmentation-Li-Weinberger\u002Fcc9826c222ac1e81b4b374dd9e0df130f298b1e8)  |\n| ✅ | 2022 |  [GroupViT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2202.11094.pdf) | 只用图像文本对也能无监督做分割        |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F0b5f27a5766c5d1394a6282ad94fec21d620bd6b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FGroupViT%3A-Semantic-Segmentation-Emerges-from-Text-Xu-Mello\u002F0b5f27a5766c5d1394a6282ad94fec21d620bd6b)  |\n| ✅ | 2022 |  [CLIPasso](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2202.05822.pdf) | CLIP跨界生成简笔画   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F9dec819778bebae4a468c7813f7638534c826f52%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FCLIPasso%3A-Semantically-Aware-Object-Sketching-Vinker-Pajouheshgar\u002F9dec819778bebae4a468c7813f7638534c826f52)  |\n| ✅ | 2022 |  [DepthCLIP](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2207.01077.pdf) | 用文本跨界估计深度   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F9d0afe58801fe9e5537902e853d6e9e385340a92%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FCan-Language-Understand-Depth-Zhang-Zeng\u002F9d0afe58801fe9e5537902e853d6e9e385340a92)  |\n\n\n\n### 自然语言处理 - Transformer\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅ | 2017 | [Transformer](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.03762) | 继MLP、CNN、RNN后的第四大类架构                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F204e3073870fae3d05bcbc2f6a8e263d9b72e776%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAttention-is-All-you-Need-Vaswani-Shazeer\u002F204e3073870fae3d05bcbc2f6a8e263d9b72e776)  |\n| ✅ | 2018 | [GPT](https:\u002F\u002Fs3-us-west-2.amazonaws.com\u002Fopenai-assets\u002Fresearch-covers\u002Flanguage-unsupervised\u002Flanguage_understanding_paper.pdf) | 使用 Transformer 解码器来做预训练               |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fcd18800a0fe0b668a1cc19f2ec95b5003d0a5035%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImproving-Language-Understanding-by-Generative-Radford-Narasimhan\u002Fcd18800a0fe0b668a1cc19f2ec95b5003d0a5035)  |\n| ✅ | 2018 | [BERT](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.04805) | Transformer一统NLP的开始                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fdf2b0e26d0599ce3e70df8a9da02e51594e0e992%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FBERT%3A-Pre-training-of-Deep-Bidirectional-for-Devlin-Chang\u002Fdf2b0e26d0599ce3e70df8a9da02e51594e0e992)  |\n| ✅ | 2019 | [GPT-2](https:\u002F\u002Fd4mucfpksywv.cloudfront.net\u002Fbetter-language-models\u002Flanguage_models_are_unsupervised_multitask_learners.pdf)  |  更大的 GPT 模型，朝着zero-shot learning迈了一大步             |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F9405cc0d6169988371b2755e573cc28650d14dfe%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLanguage-Models-are-Unsupervised-Multitask-Learners-Radford-Wu\u002F9405cc0d6169988371b2755e573cc28650d14dfe)  |\n| ✅ | 2020 |  [GPT-3](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.14165) | 100倍更大的 GPT-2，few-shot learning效果显著                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F6b85b63579a916f705a8e10a49bd8d849d91b1fc%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLanguage-Models-are-Few-Shot-Learners-Brown-Mann\u002F6b85b63579a916f705a8e10a49bd8d849d91b1fc)  |\n| ✅ | 2024 |  [Llama 3.1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2407.21783) | 强大的Meta开源模型 - 动态扩展，多模态学习，零样本学习，高效计算                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F4176a4cecfaef26b2c503827493867e703f3411a%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F4176a4cecfaef26b2c503827493867e703f3411a)  |\n\n### 系统\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n|  ✅ |  2014 | [参数服务器](https:\u002F\u002Fwww.usenix.org\u002Fsystem\u002Ffiles\u002Fconference\u002Fosdi14\u002Fosdi14-paper-li_mu.pdf) | 支持千亿参数的传统机器学习模型       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F0de0c3240bda7972bd0a3c8369ebc4b4f2e4f9c2%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FScaling-Distributed-Machine-Learning-with-the-Li-Andersen\u002F0de0c3240bda7972bd0a3c8369ebc4b4f2e4f9c2)  |\n| ✅  | 2018 | [GPipe](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2019\u002Ffile\u002F093f65e080a295f8076b1c5722a46aa2-Paper.pdf) | 流水线（Pipeline）并行      |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc18663fea10c8a303d045fd2c1f33cacf9b73ca3%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FGPipe%3A-Efficient-Training-of-Giant-Neural-Networks-Huang-Cheng\u002Fc18663fea10c8a303d045fd2c1f33cacf9b73ca3)  |\n| ✅ | 2019 | [Megatron-LM](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1909.08053.pdf) | 张量（Tensor）并行      |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F8323c591e119eb09b28b29fd6c7bc76bd889df7a%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMegatron-LM%3A-Training-Multi-Billion-Parameter-Using-Shoeybi-Patwary\u002F8323c591e119eb09b28b29fd6c7bc76bd889df7a) |\n| ✅ | 2019 | [Zero](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1910.02054.pdf) | 参数分片      |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F00c957711b12468cb38424caccdf5291bb354033%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FZeRO%3A-Memory-optimizations-Toward-Training-Trillion-Rajbhandari-Rasley\u002F00c957711b12468cb38424caccdf5291bb354033)  |\n| ✅ |  2022 | [Pathways](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2203.12533.pdf) |  将Jax拓展到上千TPU核上       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F512e9aa873a1cd7eb61ce1f25ca7df6acb7e2352%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FPathways%3A-Asynchronous-Distributed-Dataflow-for-ML-Barham-Chowdhery\u002F512e9aa873a1cd7eb61ce1f25ca7df6acb7e2352)  |\n\n### 图神经网络\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n|  ✅ |  2021 | [图神经网络介绍](https:\u002F\u002Fdistill.pub\u002F2021\u002Fgnn-intro\u002F) | GNN的可视化介绍                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F2c0e0440882a42be752268d0b64243243d752a74%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FA-Gentle-Introduction-to-Graph-Neural-Networks-S%C3%A1nchez-Lengeling-Reif\u002F2c0e0440882a42be752268d0b64243243d752a74)  |\n\n### 优化算法\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| | 2014 | [Adam](https:\u002F\u002Farxiv.org\u002Fabs\u002F1412.6980) | 深度学习里最常用的优化算法之一                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fa6cb366736791bcccc5c8639de5a8f9636bf87e8%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAdam%3A-A-Method-for-Stochastic-Optimization-Kingma-Ba\u002Fa6cb366736791bcccc5c8639de5a8f9636bf87e8)  |\n| | 2016 |  [为什么超大的模型泛化性不错](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.03530)   |               |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F54ddb00fa691728944fd8becea90a373d21597cf%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnderstanding-deep-learning-requires-rethinking-Zhang-Bengio\u002F54ddb00fa691728944fd8becea90a373d21597cf)  |\n| | 2017 | [为什么Momentum有效](https:\u002F\u002Fdistill.pub\u002F2017\u002Fmomentum\u002F) | Distill的可视化介绍            |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F3e8ccf9d3d843c9855c5d76ab66d3e775384da72%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FWhy-Momentum-Really-Works-Goh\u002F3e8ccf9d3d843c9855c5d76ab66d3e775384da72)  |\n\n\n### 新领域应用\n\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| | 2016 | [AlphaGo](https:\u002F\u002Fstorage.googleapis.com\u002Fdeepmind-media\u002Falphago\u002FAlphaGoNaturePaper.pdf) | 强化学习出圈                 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F846aedd869a00c09b40f1f1f35673cb22bc87490%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMastering-the-game-of-Go-with-deep-neural-networks-Silver-Huang\u002F846aedd869a00c09b40f1f1f35673cb22bc87490)  |\n| | 2020 | [AlphaFold](https:\u002F\u002Fdiscovery.ucl.ac.uk\u002Fid\u002Feprint\u002F10089234\u002F1\u002F343019_3_art_0_py4t4l_convrt.pdf) | 赢得比赛的的蛋白质3D结构预测 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F3a083d843f891b3574494c385699c21766ce8b7a%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImproved-protein-structure-prediction-using-from-Senior-Evans\u002F3a083d843f891b3574494c385699c21766ce8b7a)  |\n| ✅ | 2021 | [AlphaFold 2](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs41586-021-03819-2.pdf) | 原子级别精度的蛋白质3D结构预测       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fdc32a984b651256a8ec282be52310e6bd33d9815%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FHighly-accurate-protein-structure-prediction-with-Jumper-Evans\u002Fdc32a984b651256a8ec282be52310e6bd33d9815)  |\n| ✅ | 2021 | [Codex](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2107.03374.pdf) | 使用注释生成代码       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Facbdbf49f9bc3f151b93d9ca9a06009f4f6eb269%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FEvaluating-Large-Language-Models-Trained-on-Code-Chen-Tworek\u002Facbdbf49f9bc3f151b93d9ca9a06009f4f6eb269)  |\n| ✅ | 2021 | [指导数学直觉](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs41586-021-04086-x.pdf) | 分析不同数学物体之前的联系来帮助发现新定理         |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Ff672b8fb430606fee0bb368f16603531ce1e90c4%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAdvancing-mathematics-by-guiding-human-intuition-AI-Davies-Velickovic\u002Ff672b8fb430606fee0bb368f16603531ce1e90c4)  |\n| ✅ | 2022 | [AlphaCode](https:\u002F\u002Fstorage.googleapis.com\u002Fdeepmind-media\u002FAlphaCode\u002Fcompetition_level_code_generation_with_alphacode.pdf) | 媲美一般程序员的编程解题水平       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F5cbe278b65a81602a864184bbca37de91448a5f5%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FCompetition-Level-Code-Generation-with-AlphaCode-Li-Choi\u002F5cbe278b65a81602a864184bbca37de91448a5f5)  |\n\n","# 深度学习论文精读\n\n## 录制完成的论文\n\n| 日期 | 标题 | 封面 | 时长 | 视频（播放数） |\n| --: | -- | -- | --: | -- |\n| 1\u002F10\u002F25 | [OpenAI Sora](https:\u002F\u002Fopenai.com\u002Findex\u002Fvideo-generation-models-as-world-simulators\u002F) 上\u003Cbr \u002F>(包含Movie Gen和HunyuanVideo) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_f047496f2903.jpg\" width=\"200px\"\u002F> | 1:04:18 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1VdcxesEAt)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1VdcxesEAt\u002F?share_source=copy_web&vd_source=5d037e935914fc22e2e978cdccf5cdfe)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F5MGq7dSOghY?style=social)](https:\u002F\u002Fyoutu.be\u002F5MGq7dSOghY?si=lY-OsadDsTeKf-ub)  |\n| 9\u002F04\u002F24 | Llama 3.1论文精读 · 5. 模型训练过程 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_fc3deb80ee56.jpg\" width=\"200px\"\u002F> | 10:41| [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1c8HbeaEXi)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1c8HbeaEXi)\u003Cbr \u002F>  |\n| 8\u002F28\u002F24 | Llama 3.1论文精读 · 4. 训练infra | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_b0ceaf0a2286.webp\" width=\"200px\"\u002F> | 25:04| [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1b4421f7fa)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1b4421f7fa)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F6XidEHVjS1A?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=6XidEHVjS1A)  |\n| 8\u002F13\u002F24 | Llama 3.1论文精读 · 3. 模型 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_8432575b5da9.webp\" width=\"200px\"\u002F> | 26:14| [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Q4421Z7Tj)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Q4421Z7Tj)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FG6gF-5g1Gg4?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=G6gF-5g1Gg4)  |\n| 8\u002F05\u002F24 | [Llama 3.1论文精读 · 2. 预训练数据](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2407.21783) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_26c789c1391d.jpg\" width=\"200px\"\u002F> | 23:37| [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1u142187S5)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1u142187S5)[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FwXFr3zIE8FM?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=wXFr3zIE8FM)|\n| 7\u002F31\u002F24 | Llama 3.1论文精读 · 1. 导言 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_a85efbc1858a.jpg\" width=\"200px\"\u002F> | 18:53| [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1WM4m1y7Uh)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1WM4m1y7Uh)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F-PztagF3wQE?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=-PztagF3wQE)  |\n| 3\u002F30\u002F23 | [GPT-4](https:\u002F\u002Fopenai.com\u002Fresearch\u002Fgpt-4) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_6595c171965e.jpg\" width=\"200px\"\u002F> | 1:20:38 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1vM4y1U7b5)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1vM4y1U7b5)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FK0SZ9mdygTw?style=social)](https:\u002F\u002Fyoutu.be\u002FK0SZ9mdygTw)  |\n| 3\u002F23\u002F23 | 大模型时代下做科研的四个思路 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_024dacda1a7d.jpg\" width=\"200px\"\u002F> | 1:06:29 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1oX4y1d7X6)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1oX4y1d7X6)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fsh79Z8i15PI?style=social)](https:\u002F\u002Fyoutu.be\u002Fsh79Z8i15PI) |\n| 3\u002F10\u002F23 | [Anthropic LLM](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2204.05862.pdf) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_480c62e2b3cb.jpg\" width=\"200px\"\u002F> | 1:01:51 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1XY411B7nM)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1XY411B7nM)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FiqX0pgNDon0?style=social)](https:\u002F\u002Fyoutu.be\u002FiqX0pgNDon0) |\n| 1\u002F20\u002F23 | [Helm](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2211.09110.pdf) 全面语言模型评测 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_9923f5a739af.jpg\" width=\"200px\"\u002F> | 1:23:37 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1z24y1B7uX)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1z24y1B7uX)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FWgFEw9U3BXA?style=social)](https:\u002F\u002Fyoutu.be\u002FWgFEw9U3BXA) |\n| 1\u002F11\u002F23 | 多模态论文串讲·下 |  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_a432406afc4b.jpg\" width=\"200px\"\u002F> | 1:03:29 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1fA411Z772)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1fA411Z772) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FS1le41J76lQ?style=social)](https:\u002F\u002Fyoutu.be\u002FS1le41J76lQ) |\n| 12\u002F29\u002F22 | [Instruct GPT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2203.02155.pdf) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_020031862d7c.jpg\" width=\"200px\"\u002F> | 1:07:10 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1hd4y187CR)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1hd4y187CR) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FzfIGAwD1jOQ?style=social)](https:\u002F\u002Fyoutu.be\u002FzfIGAwD1jOQ) |\n| 12\u002F19\u002F22 | [Neural Corpus Indexer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2206.02743.pdf) 文档检索 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_94fcccfd8f01.jpg\" width=\"200px\"\u002F> | 55:47 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Se411w7Sn)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Se411w7Sn) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FQRffZMSGJyU?style=social)](https:\u002F\u002Fyoutu.be\u002FQRffZMSGJyU) |\n| 12\u002F12\u002F22 | 多模态论文串讲·上 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_a23028e6932d.jpg\" width=\"200px\"\u002F> | 1:12:27 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Vd4y1v77v)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Vd4y1v77v) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F6pzBOQAXUB8?style=social)](https:\u002F\u002Fyoutu.be\u002F6pzBOQAXUB8)  |\n| 11\u002F14\u002F22 | [OpenAI Whisper](https:\u002F\u002Fcdn.openai.com\u002Fpapers\u002Fwhisper.pdf) 精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_8040040b6d46.jpg\" width=\"200px\"\u002F> | 1:12:16 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1VG4y1t74x)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1VG4y1t74x) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F3eXCJd32UnM?style=social)](https:\u002F\u002Fyoutu.be\u002F3eXCJd32UnM) |\n| 11\u002F07\u002F22 | 在讲 OpenAI Whisper 前先做了一个剪视频小工具 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_0e2f637177e3.jpg\" width=\"200px\"\u002F> | 23:39 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Pe4y1t7de)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Pe4y1t7de) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FPwVlvCPDnrI?style=social)](https:\u002F\u002Fyoutu.be\u002FPwVlvCPDnrI)  |\n| 10\u002F23\u002F22 | [Chain of Thought](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2201.11903.pdf) 论文、代码和资源 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_2891c6ab6cf9.jpg\" width=\"200px\"\u002F> | 33:21 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1t8411e7Ug)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1t8411e7Ug)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FH4J59iG3t5o?style=social)](https:\u002F\u002Fyoutu.be\u002FH4J59iG3t5o) |\n| 9\u002F17\u002F22 | CLIP 改进工作串讲（下） | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_3bd53e33b5fe.jpg\" width=\"200px\"\u002F> | 1:04:26 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1gg411U7n4)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1gg411U7n4)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FugJeBivv65s?style=social)](https:\u002F\u002Fyoutu.be\u002FugJeBivv65s) |\n| 9\u002F2\u002F22 | CLIP 改进工作串讲（上） | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_f45d79988447.jpg\" width=\"200px\"\u002F> | 1:14:43 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1FV4y1p7Lm)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1FV4y1p7Lm)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fx4CDhZz_Dvg?style=social)](https:\u002F\u002Fyoutu.be\u002Fx4CDhZz_Dvg) |\n| 7\u002F29\u002F22 | [ViLT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2102.03334.pdf) 论文精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_e15b97fc3e8d.jpg\" width=\"200px\"\u002F> | 1:03:26 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV14r4y1j74y)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV14r4y1j74y)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fug8YvZOjOCE?style=social)](https:\u002F\u002Fyoutu.be\u002Fug8YvZOjOCE) |\n| 7\u002F22\u002F22 | 理由、论据和担保【[研究的艺术](https:\u002F\u002Fpress.uchicago.edu\u002Fucp\u002Fbooks\u002Fbook\u002Fchicago\u002FC\u002Fbo23521678.html)·四】 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_7b41e03a2154.jpg\" width=\"200px\"\u002F> | 44:14 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1SB4y1a75c)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1SB4y1a75c) |\n| 7\u002F15\u002F22 | 如何讲好故事、故事里的论点【[研究的艺术](https:\u002F\u002Fpress.uchicago.edu\u002Fucp\u002Fbooks\u002Fbook\u002Fchicago\u002FC\u002Fbo23521678.html)·三】| \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_ece7b5076813.jpg\" width=\"200px\"\u002F> | 43:56 |[![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1WB4y1v7ST)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1WB4y1v7ST)|\n| 7\u002F8\u002F22 | [DALL·E 2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2204.06125.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_fdf447b70ea7.jpg\" width=\"200px\"\u002F> | 1:27:54 |[![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV17r4y1u77B)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV17r4y1u77B)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FhO57mntSMl0?style=social)](https:\u002F\u002Fyoutu.be\u002FhO57mntSMl0)|\n| 7\u002F1\u002F22 | 明白问题的重要性【[研究的艺术](https:\u002F\u002Fpress.uchicago.edu\u002Fucp\u002Fbooks\u002Fbook\u002Fchicago\u002FC\u002Fbo23521678.html)·二】| \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_08b16fc18f81.jpg\" width=\"200px\"\u002F> | 1:03:40 |[![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV11S4y1v7S2)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV11S4y1v7S2\u002F)|\n| 6\u002F24\u002F22 | 跟读者建立联系【[研究的艺术](https:\u002F\u002Fpress.uchicago.edu\u002Fucp\u002Fbooks\u002Fbook\u002Fchicago\u002FC\u002Fbo23521678.html)·一】 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_001d3d2b77bc.jpg\" width=\"200px\"\u002F> | 45:01 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1hY411T7vy)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1hY411T7vy\u002F) |\n| 6\u002F17\u002F22 | [Zero](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1910.02054.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_7f007a07c22a.jpg\" width=\"200px\"\u002F> | 52:21 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1tY411g7ZT)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1tY411g7ZT\u002F) |\n| 6\u002F10\u002F22 | [DETR](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2005.12872.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_847fd1574cf7.jpg\" width=\"200px\"\u002F> | 54:22 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1GB4y1X72R)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1GB4y1X72R\u002F) |\n| 6\u002F3\u002F22 | [Megatron LM](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1909.08053.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_735a66dffc84.jpg\" width=\"200px\"\u002F> | 56:07 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1nB4y1R7Yz)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1nB4y1R7Yz\u002F) |\n| 5\u002F27\u002F22 | [GPipe](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2019\u002Ffile\u002F093f65e080a295f8076b1c5722a46aa2-Paper.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_50ca82d49de5.jpg\" width=\"200px\"\u002F> | 58:47 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1v34y1E7zu)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1v34y1E7zu\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FeXjRpS_BTbs?style=social)](https:\u002F\u002Fyoutu.be\u002FeXjRpS_BTbs)  |\n| 5\u002F5\u002F22 | [Pathways](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2203.12533.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_dc63f64be22a.jpg\" width=\"200px\"\u002F> | 1:02:13 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1xB4y1m7Xi)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1xB4y1m7Xi\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F8hS1ZtgG0wU?style=social)](https:\u002F\u002Fyoutu.be\u002F8hS1ZtgG0wU) |\n| 4\u002F28\u002F22 | [视频理解论文串讲](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2012.06567.pdf)（下） | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_3dd7beab0554.jpg\" width=\"200px\"\u002F> | 1:08:32 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV11Y411P7ep)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV11Y411P7ep\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FJ2YC0-k57NM?style=social)](https:\u002F\u002Fyoutu.be\u002FJ2YC0-k57NM) |\n| 4\u002F21\u002F22 | [参数服务器（Parameter Server）](https:\u002F\u002Fwww.usenix.org\u002Fsystem\u002Ffiles\u002Fconference\u002Fosdi14\u002Fosdi14-paper-li_mu.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_08556a3c4ef8.jpg\" width=\"200px\"\u002F> | 1:37:40 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1YA4y197G8)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1YA4y197G8\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fxt-AwUrDxQk?style=social)](https:\u002F\u002Fyoutu.be\u002Fxt-AwUrDxQk) |\n| 4\u002F14\u002F22 | [视频理解论文串讲](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2012.06567.pdf)（上） | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_2d8297a7a75b.jpg\" width=\"200px\"\u002F> | 51:15 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1fL4y157yA)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1fL4y157yA\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FgK7AGO6okhc?style=social)](https:\u002F\u002Fyoutu.be\u002FgK7AGO6okhc) |\n| 3\u002F31\u002F22 | [I3D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1705.07750.pdf) 论文精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_1c7a03f63be0.jpg\" width=\"200px\"\u002F> | 52:31 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1tY4y1p7hq)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1tY4y1p7hq\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F9lIkKiAn6uE?style=social)](https:\u002F\u002Fyoutu.be\u002F9lIkKiAn6uE) |\n| 3\u002F24\u002F22 | 斯坦福 2022 年 [AI 指数报告](https:\u002F\u002Faiindex.stanford.edu\u002Fwp-content\u002Fuploads\u002F2022\u002F03\u002F2022-AI-Index-Report_Master.pdf) 精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_8db0fddec393.jpg\" width=\"200px\"\u002F> | 1:19:56 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1s44y1N7eu)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1s44y1N7eu\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FK8h_xjQ6ufY?style=social)](https:\u002F\u002Fyoutu.be\u002FK8h_xjQ6ufY) |\n| 3\u002F17\u002F22 | [AlphaCode](https:\u002F\u002Fstorage.googleapis.com\u002Fdeepmind-media\u002FAlphaCode\u002Fcompetition_level_code_generation_with_alphacode.pdf) 论文精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_22dc2b1b039c.jpg\" width=\"200px\"\u002F> | 44:00 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1ab4y1s7rc)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1ab4y1s7rc\u002F) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Ft8Gzkca9pW4?style=social)](https:\u002F\u002Fyoutu.be\u002Ft8Gzkca9pW4) |\n| 3\u002F10\u002F22 | [OpenAI Codex](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2107.03374.pdf) 论文精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_83c68e6d2ef4.jpg\" width=\"200px\"\u002F> | 47:58 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1iY41137Zi)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1iY41137Zi\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1490959755963666432)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1490959755963666432)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FoZriUGkQSNM?style=social)](https:\u002F\u002Fyoutu.be\u002FoZriUGkQSNM) |\n| 3\u002F3\u002F22 | [GPT](https:\u002F\u002Fs3-us-west-2.amazonaws.com\u002Fopenai-assets\u002Fresearch-covers\u002Flanguage-unsupervised\u002Flanguage_understanding_paper.pdf), [GPT-2](https:\u002F\u002Fd4mucfpksywv.cloudfront.net\u002Fbetter-language-models\u002Flanguage_models_are_unsupervised_multitask_learners.pdf), [GPT-3](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.14165) 精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_20ff972ace3b.jpg\" width=\"200px\"\u002F> | 1:29:58 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1AF411b7xQ)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1AF411b7xQ\u002F)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Ft70Bl3w7bxY?style=social)](https:\u002F\u002Fyoutu.be\u002Ft70Bl3w7bxY) |\n| 2\u002F24\u002F22 | [Two-Stream](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2014\u002Ffile\u002F00ec53c4682d36f5c4359f4ae7bd7ba1-Paper.pdf) 逐段精读 |  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_1a2750415c48.jpg\" width=\"200px\"\u002F> | 52:57 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1mq4y1x7RU)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1mq4y1x7RU\u002F)\u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FvuqwKP2iDe0?style=social)](https:\u002F\u002Fyoutu.be\u002FvuqwKP2iDe0) |\n| 2\u002F10\u002F22 | [CLIP](https:\u002F\u002Fopenai.com\u002Fblog\u002Fclip\u002F) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_1be3ca6c3a75.jpg\" width=\"200px\"\u002F> | 1:38:25 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1SL4y1s7LQ)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1SL4y1s7LQ\u002F)\u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1475706654562299904)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1475706654562299904) \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FOZF1t_Hieq8?style=social)](https:\u002F\u002Fyoutu.be\u002FOZF1t_Hieq8) |\n| 2\u002F6\u002F22 | 你（被）吐槽过[论文不够 novel](https:\u002F\u002Fperceiving-systems.blog\u002Fen\u002Fpost\u002Fnovelty-in-science) 吗？| \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_35889677b888.jpg\" width=\"200px\"\u002F> | 14:11 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1ea41127Bq)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1ea41127Bq\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1475719090198876161)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1475719090198876161) |\n| 1\u002F23\u002F22 | [AlphaFold 2](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs41586-021-03819-2.pdf) 精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_7c1c47dd9e8b.jpg\" width=\"200px\"\u002F> |  1:15:28 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1oR4y1K7Xr)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1oR4y1K7Xr\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1469132410537717760)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1469132410537717760)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FOy3OCoGUr-w?style=social)](https:\u002F\u002Fyoutu.be\u002FOy3OCoGUr-w) |\n| 1\u002F18\u002F22 | 如何判断（你自己的）研究工作的价值 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_63e5502f2633.jpg\" width=\"200px\"\u002F> |  9:59 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1oL411c7Us)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1oL411c7Us\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1475716940051869696)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1475716940051869696) |\n| 1\u002F15\u002F22 | [Swin Transformer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.14030.pdf) 精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_81dd84403994.jpg\" width=\"200px\"\u002F> | 1:00:21 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV13L4y1475U)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV13L4y1475U\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1466282983652691968)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1466282983652691968)   \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FluP3-Fs0QCo?style=social)](https:\u002F\u002Fyoutu.be\u002FluP3-Fs0QCo) |\n| 1\u002F7\u002F22 | [指导数学直觉](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs41586-021-04086-x.pdf) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_d39d05b9e0d0.jpg\" width=\"200px\"\u002F> | 52:51 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1YZ4y1S72j)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1YZ4y1S72j\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1464060386375299072)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1464060386375299072)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FczFGjvhtss8?style=social)](https:\u002F\u002Fyoutu.be\u002FczFGjvhtss8) |\n| 1\u002F5\u002F22 | AlphaFold 2 预告 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_b2a0459aca0e.jpg\" width=\"200px\"\u002F> | 03:28 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Eu411U7Te)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Eu411U7Te\u002F) |\n| 12\u002F20\u002F21 | [对比学习](#contrastive_learning)论文综述 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_12633fdc9471.jpg\" width=\"200px\"\u002F> | 1:32:01 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV19S4y1M7hm)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV19S4y1M7hm\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1460828005077164032)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1460828005077164032)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F1pvxufGRuW4?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=1pvxufGRuW4) |\n| 12\u002F15\u002F21 | [MoCo](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1911.05722.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_25987fa17019.jpg\" width=\"200px\"\u002F> | 1:24:11 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1C3411s7t9)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1C3411s7t9\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1454723120678936576)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1454723120678936576)   \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002F1pvxufGRuW4?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=1pvxufGRuW4) |\n| 12\u002F9\u002F21 | 如何找研究想法 1 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_80c06ea1f09b.jpg\" width=\"200px\"\u002F> | 5:34 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1qq4y1z7F2)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1qq4y1z7F2\u002F) |\n| 12\u002F8\u002F21 | [MAE](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2111.06377.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_c9c4c48e57ec.jpg\" width=\"200px\"\u002F> | 47:04 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1sq4y1q77t)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1sq4y1q77t\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1452458167968251904)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1452458167968251904)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FmYlX2dpdHHM?style=social)](https:\u002F\u002Fyoutu.be\u002FmYlX2dpdHHM) |\n| 11\u002F29\u002F21 | [ViT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2010.11929.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_e3541b53dea0.jpg\" width=\"200px\"\u002F> | 1:11:30 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV15P4y137jb)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV15P4y137jb\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1449195245754380288)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1449195245754380288)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FFRFt3x0bO94?style=social)](https:\u002F\u002Fyoutu.be\u002FFRFt3x0bO94) |\n| 11\u002F18\u002F21 | [BERT](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.04805) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_3a938b86f843.jpg\" width=\"200px\"\u002F> | 45:49  | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1PL411M7eQ)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1PL411M7eQ\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1445340200976785408)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1445340200976785408)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FULD3uIb2MHQ?style=social)](https:\u002F\u002Fyoutu.be\u002FULD3uIb2MHQ) |\n| 11\u002F9\u002F21 | [GAN](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2014\u002Ffile\u002F5ca3e9b122f61f8f06494c97b1afccf3-Paper.pdf) 逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_99eeaac972eb.jpg\" width=\"200px\"\u002F> | 46:16  | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1rb4y187vD)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1rb4y187vD\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1442091389241159681)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1442091389241159681)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fg_0HtlrLiDo?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=g_0HtlrLiDo) |\n| 11\u002F3\u002F21 | 零基础多图详解 [图神经网络](https:\u002F\u002Fdistill.pub\u002F2021\u002Fgnn-intro\u002F)（GNN\u002FGCN） | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_f2552664a7c0.jpg\" width=\"200px\"\u002F> | 1:06:19 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1iT4y1d7zP)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1iT4y1d7zP\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1439540657619087360)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1439540657619087360)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FsejA2PtCITw?style=social)](https:\u002F\u002Fyoutu.be\u002FsejA2PtCITw) |\n| 10\u002F27\u002F21 | [Transformer](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.03762) 逐段精读\u003Cbr> （视频中提到的文献 [^transformer]) |\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_e7ec6347400b.jpg\" width=\"200px\"\u002F> | 1:27:05 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1pu411o7BE)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1pu411o7BE\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1437034536677404672)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1437034536677404672)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FnzqlFIcCSWQ?style=social)](https:\u002F\u002Fyoutu.be\u002FnzqlFIcCSWQ) |\n| 10\u002F22\u002F21 | [ResNet](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.03385) 论文逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_22e4b07b303b.jpg\" width=\"200px\"\u002F> | 53:46 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1P3411y7nn)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1P3411y7nn\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1434795406001180672)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1434795406001180672)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FpWMnzCX4cwQ?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=pWMnzCX4cwQ) |\n| 10\u002F21\u002F21 | 撑起计算机视觉半边天的 [ResNet](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.03385) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_081a2b3fb354.jpg\" width=\"200px\"\u002F> | 11:50 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1Fb4y1h73E)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1Fb4y1h73E\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1434787226101751808)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1434787226101751808)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FNnSldWhSqvY?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=NnSldWhSqvY) |\n| 10\u002F15\u002F21 | [AlexNet](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2012\u002Ffile\u002Fc399862d3b9d6b76c8436e924a68c45b-Paper.pdf) 论文逐段精读 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_554ce4eea08a.jpg\" width=\"200px\"\u002F> | 55:21 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1hq4y157t1)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1hq4y157t1\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1432354207483871232)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1432354207483871232)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FwYmlILPsLlY?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=wYmlILPsLlY) |\n| 10\u002F14\u002F21 | 9年后重读深度学习奠基作之一：[AlexNet](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2012\u002Ffile\u002Fc399862d3b9d6b76c8436e924a68c45b-Paper.pdf) | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_42196dca7f17.jpg\" width=\"200px\"\u002F> | 19:59 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1ih411J7Kz)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1ih411J7Kz\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1432155856322920448)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1432155856322920448)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FvdYH0fE6thY?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=vdYH0fE6thY) |\n| 10\u002F06\u002F21 | 如何读论文 | \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_readme_18ece7c81b03.jpg\" width=\"200px\"\u002F> | 06:39 | [![bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=bilibili&query=data.stat.view&url=https%3A%2F%2Fapi.bilibili.com%2Fx%2Fweb-interface%2Fview%3Fbvid%3DBV1H44y1t75x)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1H44y1t75x\u002F) \u003Cbr \u002F>[![zhihu](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=views&style=social&logo=zhihu&query=video.play_count&url=https:\u002F\u002Fwww.zhihu.com\u002Fapi\u002Fv4\u002Fzvideos\u002F1428973951632969728)](https:\u002F\u002Fwww.zhihu.com\u002Fzvideo\u002F1428973951632969728)  \u003Cbr \u002F>[![](https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Ftxjl_Q4jCyQ?style=social)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=txjl_Q4jCyQ&list=PLFXJ6jwg0qW-7UM8iUTj3qKqdhbQULP5I&index=1) |\n\n[^transformer]: 1 [斯坦福100+作者的200+页综述](https:\u002F\u002Farxiv.org\u002Fabs\u002F2108.07258)，2 [对LayerNorm的新研究](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1911.07013.pdf)，3 [对Attention在Transformer里面作用的研究](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.03404)\n\n\n\n\n## 所有论文\n\n包括已经录制完成和之后将要介绍的论文。选取的原则是10年内深度学习里有影响力文章（必读文章），或者近期比较有意思的文章。当然这十年里重要的工作太多了，不可能一一过一遍。在选取的时候我会偏向一些之前 [直播课](https:\u002F\u002Fc.d2l.ai\u002Fzh-v2\u002F) 中没讲到过的。 欢迎大家在 [讨论区](https:\u002F\u002Fgithub.com\u002Fmli\u002Fpaper-reading\u002Fdiscussions) 里提供建（点）议（歌）。\n\n总论文数 67，录制完成数 32\n\n（这里引用采用的是 semanticscholar，是因为它提供 [API](https:\u002F\u002Fapi.semanticscholar.org\u002Fapi-docs\u002Fgraph#operation\u002Fget_graph_get_paper) 可以自动获取，不用手动更新。）\n\n### 计算机视觉 - CNN\n\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅      | 2012 | [AlexNet](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2012\u002Ffile\u002Fc399862d3b9d6b76c8436e924a68c45b-Paper.pdf) | 深度学习热潮的奠基作                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fabd1c342495432171beb7ca8fd9551ef13cbd0ff%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImageNet-classification-with-deep-convolutional-Krizhevsky-Sutskever\u002Fabd1c342495432171beb7ca8fd9551ef13cbd0ff) |\n| | 2014 | [VGG](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1409.1556.pdf) | 使用 3x3 卷积构造更深的网络                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Feb42cf88027de515750f230b23b1a057dc782108%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FVery-Deep-Convolutional-Networks-for-Large-Scale-Simonyan-Zisserman\u002Feb42cf88027de515750f230b23b1a057dc782108) |\n| | 2014 | [GoogleNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1409.4842.pdf) | 使用并行架构构造更深的网络                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fe15cf50aa89fee8535703b9f9512fca5bfc43327%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FGoing-deeper-with-convolutions-Szegedy-Liu\u002Fe15cf50aa89fee8535703b9f9512fca5bfc43327) |\n|  ✅  | 2015 |  [ResNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1512.03385.pdf) | 构建深层网络都要有的残差连接。               |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F2c03df8b48bf3fa39054345bafabfeff15bfd11d%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FDeep-Residual-Learning-for-Image-Recognition-He-Zhang\u002F2c03df8b48bf3fa39054345bafabfeff15bfd11d)  |\n|  | 2017 | [MobileNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1704.04861.pdf) | 适合终端设备的小CNN                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F3647d6d0f151dc05626449ee09cc7bce55be497e%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMobileNets%3A-Efficient-Convolutional-Neural-Networks-Howard-Zhu\u002F3647d6d0f151dc05626449ee09cc7bce55be497e)  |\n| | 2019 | [EfficientNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1905.11946.pdf) | 通过架构搜索得到的CNN                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F4f2eda8077dc7a69bb2b4e0a1a086cf054adb3f9%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FEfficientNet%3A-Rethinking-Model-Scaling-for-Neural-Tan-Le\u002F4f2eda8077dc7a69bb2b4e0a1a086cf054adb3f9)  |\n| | 2021 |  [Non-deep networks](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2110.07641.pdf) | 让不深的网络也能在ImageNet刷到SOTA                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F0d7f6086772079bc3e243b7b375a9ca1a517ba8b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FNon-deep-Networks-Goyal-Bochkovskiy\u002F0d7f6086772079bc3e243b7b375a9ca1a517ba8b) |\n\n### 计算机视觉 - Transformer\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅ | 2020 | [ViT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2010.11929.pdf) | Transformer杀入CV界                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F7b15fa1b8d413fbe14ef7a97f651f47f5aff3903%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAn-Image-is-Worth-16x16-Words%3A-Transformers-for-at-Dosovitskiy-Beyer\u002F7b15fa1b8d413fbe14ef7a97f651f47f5aff3903)  |\n| ✅ | 2021 | [Swin Transformer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.14030.pdf) | 多层次的Vision Transformer                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc8b25fab5608c3e033d34b4483ec47e68ba109b7%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FSwin-Transformer%3A-Hierarchical-Vision-Transformer-Liu-Lin\u002Fc8b25fab5608c3e033d34b4483ec47e68ba109b7) |\n| | 2021 | [MLP-Mixer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2105.01601.pdf) | 使用MLP替换self-attention            |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F2def61f556f9a5576ace08911496b7c7e4f970a4%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMLP-Mixer%3A-An-all-MLP-Architecture-for-Vision-Tolstikhin-Houlsby\u002F2def61f556f9a5576ace08911496b7c7e4f970a4)  |\n| ✅ | 2021 | [MAE](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2111.06377.pdf) | BERT的CV版             |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc1962a8cf364595ed2838a097e9aa7cd159d3118%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMasked-Autoencoders-Are-Scalable-Vision-Learners-He-Chen\u002Fc1962a8cf364595ed2838a097e9aa7cd159d3118)  |\n\n### 生成模型\n\n| 已录制 | 年份 | 名字                                              | 简介         | 引用 |\n| ------ | ---- | ------------------------------------------------- | ------------ | ------------------------------------------------------------ |\n|  ✅ | 2014 | [GAN](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2014\u002Ffile\u002F5ca3e9b122f61f8f06494c97b1afccf3-Paper.pdf) | 生成模型的开创工作                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F54e325aee6b2d476bbbb88615ac15e251c6e8214%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FGenerative-Adversarial-Nets-Goodfellow-Pouget-Abadie\u002F54e325aee6b2d476bbbb88615ac15e251c6e8214)  |\n|  | 2015 | [DCGAN](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1511.06434.pdf) | 使用CNN的GAN          |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F8388f1be26329fa45e5807e968a641ce170ea078%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnsupervised-Representation-Learning-with-Deep-Radford-Metz\u002F8388f1be26329fa45e5807e968a641ce170ea078)  |\n|  | 2016 | [pix2pix](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1611.07004.pdf) |           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F8acbe90d5b852dadea7810345451a99608ee54c7%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImage-to-Image-Translation-with-Conditional-Isola-Zhu\u002F8acbe90d5b852dadea7810345451a99608ee54c7)  |\n|  | 2016 | [SRGAN](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1609.04802.pdf) | 图片超分辨率          |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fdf0c54fe61f0ffb9f0e36a17c2038d9a1964cba3%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FPhoto-Realistic-Single-Image-Super-Resolution-Using-Ledig-Theis\u002Fdf0c54fe61f0ffb9f0e36a17c2038d9a1964cba3)  |\n|  | 2017 | [WGAN](https:\u002F\u002Farxiv.org\u002Fabs\u002F1701.07875) | 训练更加容易          |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F2f85b7376769473d2bed56f855f115e23d727094%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FWasserstein-GAN-Arjovsky-Chintala\u002F2f85b7376769473d2bed56f855f115e23d727094)  |\n|  | 2017 | [CycleGAN](https:\u002F\u002Farxiv.org\u002Fabs\u002F1703.10593) |           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc43d954cf8133e6254499f3d68e45218067e4941%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnpaired-Image-to-Image-Translation-Using-Networks-Zhu-Park\u002Fc43d954cf8133e6254499f3d68e45218067e4941)  |\n|  | 2018 | [StyleGAN](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.04948) |           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fceb2ebef0b41e31c1a21b28c2734123900c005e2%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FA-Style-Based-Generator-Architecture-for-Generative-Karras-Laine\u002Fceb2ebef0b41e31c1a21b28c2734123900c005e2)  |\n| | 2019 | [StyleGAN2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1912.04958.pdf) |        |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Ff3e3d1f86a534a3654d0ee263142e44f4e2c61e9%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAnalyzing-and-Improving-the-Image-Quality-of-Karras-Laine\u002Ff3e3d1f86a534a3654d0ee263142e44f4e2c61e9)  |\n| | 2020 | [DDPM](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2006.11239.pdf) | Diffusion Models   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F289db3be7bf77e06e75541ba93269de3d604ac72%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FDenoising-Diffusion-Probabilistic-Models-Ho-Jain\u002F289db3be7bf77e06e75541ba93269de3d604ac72)  |\n| | 2021 | [Improved DDPM](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2102.09672.pdf) | 改进的 DDPM   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fde18baa4964804cf471d85a5a090498242d2e79f%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImproved-Denoising-Diffusion-Probabilistic-Models-Nichol-Dhariwal\u002Fde18baa4964804cf471d85a5a090498242d2e79f)  |\n| | 2021 | [Guided Diffusion Models](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2105.05233.pdf) | 号称超越 GAN  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F64ea8f180d0682e6c18d1eb688afdb2027c02794%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FDiffusion-Models-Beat-GANs-on-Image-Synthesis-Dhariwal-Nichol\u002F64ea8f180d0682e6c18d1eb688afdb2027c02794)  |\n| | 2021 | [StyleGAN3](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2106.12423.pdf) |        |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc1ff08b59f00c44f34dfdde55cd53370733a2c19%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAlias-Free-Generative-Adversarial-Networks-Karras-Aittala\u002Fc1ff08b59f00c44f34dfdde55cd53370733a2c19)  |\n|  ✅  | 2022 | [DALL.E 2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2204.06125.pdf) | CLIP + Diffusion models，文本生成图像新高度     |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc57293882b2561e1ba03017902df9fc2f289dea2%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FHierarchical-Text-Conditional-Image-Generation-with-Ramesh-Dhariwal\u002Fc57293882b2561e1ba03017902df9fc2f289dea2)  |\n|  ✅  | 2024 | [Sora](https:\u002F\u002Fopenai.com\u002Findex\u002Fvideo-generation-models-as-world-simulators\u002F) | 开启视频生成热潮     |  |\n|  ✅  | 2024 | [Movie Gen](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2410.13720) | 精确的文本指导视频编辑、个性化视频生成     |  |\n|  ✅  | 2025 | [HunyuanVideo](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2412.03603) | 开源视频生成框架     |  |\n\n### 计算机视觉 - 目标检测\n\n| 已录制 | 年份 | 名字                                              | 简介         | 引用 |\n| ------ | ---- | ------------------------------------------------- | ------------ | ------------------------------------------------------------ |\n|        | 2014 | [R-CNN](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1311.2524v5.pdf)    | 两阶段             |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F2f4df08d9072fc2ac181b7fced6a245315ce05c8%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F2f4df08d9072fc2ac181b7fced6a245315ce05c8)  |\n|        | 2015 | [Fast R-CNN](http:\u002F\u002Farxiv.org\u002Fabs\u002F1504.08083v2)   |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F7ffdbc358b63378f07311e883dddacc9faeeaf4b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F7ffdbc358b63378f07311e883dddacc9faeeaf4b)  |\n|        | 2015 | [Faster R-CNN](http:\u002F\u002Farxiv.org\u002Fabs\u002F1506.01497v3) |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F424561d8585ff8ebce7d5d07de8dbf7aae5e7270%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F424561d8585ff8ebce7d5d07de8dbf7aae5e7270)  |\n|        | 2016 | [SSD](http:\u002F\u002Farxiv.org\u002Fabs\u002F1512.02325v5)          | 单阶段          |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F4d7a9197433acbfb24ef0e9d0f33ed1699e4a5b0%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F4d7a9197433acbfb24ef0e9d0f33ed1699e4a5b0)  |\n|        | 2016 | [YOLO](http:\u002F\u002Farxiv.org\u002Fabs\u002F1506.02640v5)         |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Ff8e79ac0ea341056ef20f2616628b3e964764cfd%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002Ff8e79ac0ea341056ef20f2616628b3e964764cfd)  |\n|        | 2017 | [Mask R-CNN](http:\u002F\u002Farxiv.org\u002Fabs\u002F1703.06870v3)   |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fea99a5535388196d0d44be5b4d7dd02029a43bb2%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002Fea99a5535388196d0d44be5b4d7dd02029a43bb2)  |\n|        | 2017 | [YOLOv2](http:\u002F\u002Farxiv.org\u002Fabs\u002F1612.08242v1)       |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F7d39d69b23424446f0400ef603b2e3e22d0309d6%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F7d39d69b23424446f0400ef603b2e3e22d0309d6)  |\n|        | 2018 | [YOLOv3](http:\u002F\u002Farxiv.org\u002Fabs\u002F1804.02767v1)       |                       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fe4845fb1e624965d4f036d7fd32e8dcdd2408148%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002Fe4845fb1e624965d4f036d7fd32e8dcdd2408148)  |\n|        | 2019 | [CenterNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1904.07850.pdf) | 无锚点           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F6a2e2fd1b5bb11224daef98b3fb6d029f68a73f2%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FObjects-as-Points-Zhou-Wang\u002F6a2e2fd1b5bb11224daef98b3fb6d029f68a73f2)  |\n|   ✅     | 2020 | [DETR](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2005.12872.pdf)      | 变压器           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F962dc29fdc3fbdc5930a10aba114050b82fe5a3e%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FEnd-to-End-Object-Detection-with-Transformers-Carion-Massa\u002F962dc29fdc3fbdc5930a10aba114050b82fe5a3e)  |\n\n\u003Ca name=\"contrastive_learning\">\u003C\u002Fa>\n\n### 计算机视觉 - 对比学习\n\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅      | 2018 | [InstDisc](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1805.01978.pdf) | 提出实例判别和memory bank做对比学习                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F155b7782dbd713982a4133df3aee7adfd0b6b304%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnsupervised-Feature-Learning-via-Non-parametric-Wu-Xiong\u002F155b7782dbd713982a4133df3aee7adfd0b6b304)  |\n| ✅      | 2018 | [CPC](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1807.03748.pdf) | 对比预测编码，图像语音文本强化学习全都能做                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fb227f3e4c0dc96e5ac5426b85485a70f2175a205%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FRepresentation-Learning-with-Contrastive-Predictive-Oord-Li\u002Fb227f3e4c0dc96e5ac5426b85485a70f2175a205) |\n| ✅      | 2019 | [InvaSpread](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1904.03436.pdf) | 一个编码器的端到端对比学习                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fe4bde6fe33b6c2cf9d1647ac0b041f7d1ba29c5b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnsupervised-Embedding-Learning-via-Invariant-and-Ye-Zhang\u002Fe4bde6fe33b6c2cf9d1647ac0b041f7d1ba29c5b)  |\n| ✅  | 2019 |  [CMC](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1906.05849.pdf) | 多视角下的对比学习               |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F97f4d09175705be4677d675fa27e55defac44800%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FContrastive-Multiview-Coding-Tian-Krishnan\u002F97f4d09175705be4677d675fa27e55defac44800)  |\n| ✅ | 2019 | [MoCov1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1911.05722.pdf) | 无监督训练效果也很好                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fec46830a4b275fd01d4de82bffcabe6da086128f%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMomentum-Contrast-for-Unsupervised-Visual-Learning-He-Fan\u002Fec46830a4b275fd01d4de82bffcabe6da086128f) |\n|  ✅ | 2020 |  [SimCLRv1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2002.05709.pdf) |  简单的对比学习 (数据增强 + MLP head + 大batch训练久)                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F34733eaf66007516347a40ad5d9bbe1cc9dacb6b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FA-Simple-Framework-for-Contrastive-Learning-of-Chen-Kornblith\u002F34733eaf66007516347a40ad5d9bbe1cc9dacb6b)  |\n|  ✅ | 2020 | [MoCov2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2003.04297.pdf) | MoCov1 + improvements from SimCLRv1                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fa1b8a8df281bbaec148a897927a49ea47ea31515%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImproved-Baselines-with-Momentum-Contrastive-Chen-Fan\u002Fa1b8a8df281bbaec148a897927a49ea47ea31515)  |\n|  ✅ | 2020 |  [SimCLRv2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2006.10029.pdf) | 大的自监督预训练模型很适合做半监督学习                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F3e7f5f4382ac6f9c4fef6197dd21abf74456acd1%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FBig-Self-Supervised-Models-are-Strong-Learners-Chen-Kornblith\u002F3e7f5f4382ac6f9c4fef6197dd21abf74456acd1)  |\n| ✅  | 2020 |  [BYOL](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2006.07733.pdf) | 不需要负样本的对比学习                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F38f93092ece8eee9771e61c1edaf11b1293cae1b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FBootstrap-Your-Own-Latent%3A-A-New-Approach-to-Grill-Strub\u002F38f93092ece8eee9771e61c1edaf11b1293cae1b) |\n|  ✅ | 2020 |  [SWaV](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2006.09882.pdf) | 聚类对比学习                   | [![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F10161d83d29fc968c4612c9e9e2b61a2fc25842e%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnsupervised-Learning-of-Visual-Features-by-Cluster-Caron-Misra\u002F10161d83d29fc968c4612c9e9e2b61a2fc25842e) |\n|  ✅ | 2020 |  [SimSiam](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2011.10566.pdf) | 化繁为简的孪生表征学习                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F0e23d2f14e7e56e81538f4a63e11689d8ac1eb9d%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FExploring-Simple-Siamese-Representation-Learning-Chen-He\u002F0e23d2f14e7e56e81538f4a63e11689d8ac1eb9d)  |\n| ✅ | 2021 | [MoCov3](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2104.02057.pdf) | 如何更稳定的自监督训练ViT                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F739ceacfafb1c4eaa17509351b647c773270b3ae%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAn-Empirical-Study-of-Training-Self-Supervised-Chen-Xie\u002F739ceacfafb1c4eaa17509351b647c773270b3ae)  |\n|  ✅ | 2021 |  [DINO](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2104.14294.pdf) | transformer加自监督在视觉也很香                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fad4a0938c48e61b7827869e4ac3baffd0aefab35%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FEmerging-Properties-in-Self-Supervised-Vision-Caron-Touvron\u002Fad4a0938c48e61b7827869e4ac3baffd0aefab35)  |\n\n### 计算机视觉 - 视频理解\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅ | 2014 |  [DeepVideo](https:\u002F\u002Fcs.stanford.edu\u002Fpeople\u002Fkarpathy\u002Fdeepvideo\u002F) | 提出sports1M数据集，用深度学习做视频理解 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F6d4c9c923e9f145d1c01a2de2afc38ec23c44253%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLarge-Scale-Video-Classification-with-Convolutional-Karpathy-Toderici\u002F6d4c9c923e9f145d1c01a2de2afc38ec23c44253)  |\n| ✅ | 2014 |  [Two-stream](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1406.2199.pdf) | 引入光流做时序建模，神经网络首次超越手工特征 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F67dccc9a856b60bdc4d058d83657a089b8ad4486%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FTwo-Stream-Convolutional-Networks-for-Action-in-Simonyan-Zisserman\u002F67dccc9a856b60bdc4d058d83657a089b8ad4486)  |\n| ✅ | 2014 |  [C3D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1412.0767.pdf) |  比较深的3D-CNN做视频理解 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fd25c65d261ea0e6a458be4c50c40ffe5bc508f77%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLearning-Spatiotemporal-Features-with-3D-Networks-Tran-Bourdev\u002Fd25c65d261ea0e6a458be4c50c40ffe5bc508f77)  |\n| ✅ | 2015 |  [Beyond-short-snippets](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1503.08909.pdf) | 尝试使用LSTM  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F5418b2a482720e013d487a385c26fae0f017c6a6%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FBeyond-short-snippets%3A-Deep-networks-for-video-Ng-Hausknecht\u002F5418b2a482720e013d487a385c26fae0f017c6a6)  |\n| ✅ | 2016 |  [Convolutional fusion](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1604.06573.pdf) | 做early fusion来加强时空间建模    |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F9d9aced120e530484609164c836da64548693484%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FConvolutional-Two-Stream-Network-Fusion-for-Video-Feichtenhofer-Pinz\u002F9d9aced120e530484609164c836da64548693484)  |\n| ✅ | 2016 |  [TSN](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1608.00859.pdf) | 超级有效的视频分段建模，bag of tricks in video |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fea3d7de6c0880e14455b9acb28f1bc1234321456%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FTemporal-Segment-Networks%3A-Towards-Good-Practices-Wang-Xiong\u002Fea3d7de6c0880e14455b9acb28f1bc1234321456)  |\n| ✅ | 2017 |  [I3D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1705.07750.pdf) | 提出Kinetics数据集，膨胀2D网络到3D，开启3D-CNN时代  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fb61a3f8b80bbd44f24544dc915f52fd30bbdf485%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FQuo-Vadis%2C-Action-Recognition-A-New-Model-and-the-Carreira-Zisserman\u002Fb61a3f8b80bbd44f24544dc915f52fd30bbdf485)  |\n| ✅ | 2017 |  [R2+1D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1711.11248.pdf) | 拆分3D卷积核，使3D网络容易优化  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F89c3050522a0bb9820c32dc7444e003ef0d3e2e4%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FA-Closer-Look-at-Spatiotemporal-Convolutions-for-Tran-Wang\u002F89c3050522a0bb9820c32dc7444e003ef0d3e2e4)  |\n| ✅ | 2017 |  [Non-local](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1711.07971.pdf) | 引入自注意力做视觉问题  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F8899094797e82c5c185a0893896320ef77f60e64%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FNon-local-Neural-Networks-Wang-Girshick\u002F8899094797e82c5c185a0893896320ef77f60e64)  |\n| ✅ | 2018 |  [SlowFast](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1812.03982.pdf) | 快慢两支提升效率   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F8b47b9c3c35b2b2a78bff7822605b3040f87d699%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FSlowFast-Networks-for-Video-Recognition-Feichtenhofer-Fan\u002F8b47b9c3c35b2b2a78bff7822605b3040f87d699)  |\n| ✅ | 2021 |  [TimeSformer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2102.05095.pdf) | 视频中第一个引入transformer，开启video transformer时代 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc143ea9e30b1f2d93a9c060253845423f9e60e1f%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FIs-Space-Time-Attention-All-You-Need-for-Video-Bertasius-Wang\u002Fc143ea9e30b1f2d93a9c060253845423f9e60e1f)  |\n\n### 多模态学习\n\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅ | 2021 |  [CLIP](https:\u002F\u002Fopenai.com\u002Fblog\u002Fclip\u002F) | 图片和文本之间的对比学习                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F6f870f7f02a8c59c3e23f407f3ef00dd1dcf8fc4%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLearning-Transferable-Visual-Models-From-Natural-Radford-Kim\u002F6f870f7f02a8c59c3e23f407f3ef00dd1dcf8fc4)  |\n| ✅ | 2021 |  [ViLT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2102.03334.pdf) | 第一个摆脱了目标检测的视觉文本模型      |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F0839722fb5369c0abaff8515bfc08299efc790a1%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FViLT%3A-Vision-and-Language-Transformer-Without-or-Kim-Son\u002F0839722fb5369c0abaff8515bfc08299efc790a1)  |\n| ✅ | 2021 |  [ViLD](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2104.13921.pdf) | CLIP蒸馏帮助开集目标检测      |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fcf9b8da26d9b92e75ba49616ed2a1033f59fce14%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FOpen-vocabulary-Object-Detection-via-Vision-and-Gu-Lin\u002Fcf9b8da26d9b92e75ba49616ed2a1033f59fce14)  |\n| ✅ | 2021 |  [GLIP](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2112.03857.pdf) | 联合目标检测和文本定位           |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F5341b412383c43f4a693ad63ec4489e3ec7688c8%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FGrounded-Language-Image-Pre-training-Li-Zhang\u002F5341b412383c43f4a693ad63ec4489e3ec7688c8)  |\n| ✅ | 2021 |  [CLIP4Clip](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2104.08860.pdf) | 拿CLIP直接做视频文本retrieval       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F281ad83e06d731d5d686acf07cd701576f1188c4%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FCLIP4Clip%3A-An-Empirical-Study-of-CLIP-for-End-to-Luo-Ji\u002F281ad83e06d731d5d686acf07cd701576f1188c4)  |\n| ✅ | 2021 |  [ActionCLIP](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2109.08472.pdf) | 用多模态对比学习有监督的做视频动作分类   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fdc05240a06326b5b1664f7e8c95c330b08cd0349%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FActionCLIP%3A-A-New-Paradigm-for-Video-Action-Wang-Xing\u002Fdc05240a06326b5b1664f7e8c95c330b08cd0349)  |\n| ✅ | 2021 |  [PointCLIP](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2112.02413.pdf) | 3D变2D，巧妙利用CLIP做点云  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Ff3ce9ba3fcec362b70263a7ed63d9404975496a0%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FPointCLIP%3A-Point-Cloud-Understanding-by-CLIP-Zhang-Guo\u002Ff3ce9ba3fcec362b70263a7ed63d9404975496a0)  |\n| ✅ | 2022 |  [LSeg](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2201.03546.pdf) | 有监督的开集分割                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fcc9826c222ac1e81b4b374dd9e0df130f298b1e8%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLanguage-driven-Semantic-Segmentation-Li-Weinberger\u002Fcc9826c222ac1e81b4b374dd9e0df130f298b1e8)  |\n| ✅ | 2022 |  [GroupViT](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2202.11094.pdf) | 只用图像文本对也能无监督做分割        |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F0b5f27a5766c5d1394a6282ad94fec21d620bd6b%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FGroupViT%3A-Semantic-Segmentation-Emerges-from-Text-Xu-Mello\u002F0b5f27a5766c5d1394a6282ad94fec21d620bd6b)  |\n| ✅ | 2022 |  [CLIPasso](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2202.05822.pdf) | CLIP跨界生成简笔画   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F9dec819778bebae4a468c7813f7638534c826f52%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FCLIPasso%3A-Semantically-Aware-Object-Sketching-Vinker-Pajouheshgar\u002F9dec819778bebae4a468c7813f7638534c826f52)  |\n| ✅ | 2022 |  [DepthCLIP](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2207.01077.pdf) | 用文本跨界估计深度   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F9d0afe58801fe9e5537902e853d6e9e385340a92%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FCan-Language-Understand-Depth-Zhang-Zeng\u002F9d0afe58801fe9e5537902e853d6e9e385340a92)  |\n\n### 自然语言处理 - Transformer\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| ✅ | 2017 | [Transformer](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.03762) | 继MLP、CNN、RNN后的第四大类架构                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F204e3073870fae3d05bcbc2f6a8e263d9b72e776%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAttention-is-All-you-Need-Vaswani-Shazeer\u002F204e3073870fae3d05bcbc2f6a8e263d9b72e776)  |\n| ✅ | 2018 | [GPT](https:\u002F\u002Fs3-us-west-2.amazonaws.com\u002Fopenai-assets\u002Fresearch-covers\u002Flanguage-unsupervised\u002Flanguage_understanding_paper.pdf) | 使用 Transformer 解码器来做预训练               |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fcd18800a0fe0b668a1cc19f2ec95b5003d0a5035%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImproving-Language-Understanding-by-Generative-Radford-Narasimhan\u002Fcd18800a0fe0b668a1cc19f2ec95b5003d0a5035)  |\n| ✅ | 2018 | [BERT](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.04805) | Transformer一统NLP的开始                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fdf2b0e26d0599ce3e70df8a9da02e51594e0e992%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FBERT%3A-Pre-training-of-Deep-Bidirectional-for-Devlin-Chang\u002Fdf2b0e26d0599ce3e70df8a9da02e51594e0e992)  |\n| ✅ | 2019 | [GPT-2](https:\u002F\u002Fd4mucfpksywv.cloudfront.net\u002Fbetter-language-models\u002Flanguage_models_are_unsupervised_multitask_learners.pdf)  |  更大的 GPT 模型，朝着zero-shot learning迈了一大步             |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F9405cc0d6169988371b2755e573cc28650d14dfe%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLanguage-Models-are-Unsupervised-Multitask-Learners-Radford-Wu\u002F9405cc0d6169988371b2755e573cc28650d14dfe)  |\n| ✅ | 2020 |  [GPT-3](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.14165) | 100倍更大的 GPT-2，few-shot learning效果显著                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F6b85b63579a916f705a8e10a49bd8d849d91b1fc%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FLanguage-Models-are-Few-Shot-Learners-Brown-Mann\u002F6b85b63579a916f705a8e10a49bd8d849d91b1fc)  |\n| ✅ | 2024 |  [Llama 3.1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2407.21783) | 强大的Meta开源模型 - 动态扩展，多模态学习，零样本学习，高效计算                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F4176a4cecfaef26b2c503827493867e703f3411a%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002F4176a4cecfaef26b2c503827493867e703f3411a)  |\n\n### 系统\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n|  ✅ |  2014 | [参数服务器](https:\u002F\u002Fwww.usenix.org\u002Fsystem\u002Ffiles\u002Fconference\u002Fosdi14\u002Fosdi14-paper-li_mu.pdf) | 支持千亿参数的传统机器学习模型       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F0de0c3240bda7972bd0a3c8369ebc4b4f2e4f9c2%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FScaling-Distributed-Machine-Learning-with-the-Li-Andersen\u002F0de0c3240bda7972bd0a3c8369ebc4b4f2e4f9c2)  |\n| ✅  | 2018 | [GPipe](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2019\u002Ffile\u002F093f65e080a295f8076b1c5722a46aa2-Paper.pdf) | 流水线（Pipeline）并行      |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fc18663fea10c8a303d045fd2c1f33cacf9b73ca3%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FGPipe%3A-Efficient-Training-of-Giant-Neural-Networks-Huang-Cheng\u002Fc18663fea10c8a303d045fd2c1f33cacf9b73ca3)  |\n| ✅ | 2019 | [Megatron-LM](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1909.08053.pdf) | 张量（Tensor）并行      |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F8323c591e119eb09b28b29fd6c7bc76bd889df7a%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMegatron-LM%3A-Training-Multi-Billion-Parameter-Using-Shoeybi-Patwary\u002F8323c591e119eb09b28b29fd6c7bc76bd889df7a) |\n| ✅ | 2019 | [Zero](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1910.02054.pdf) | 参数分片      |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F00c957711b12468cb38424caccdf5291bb354033%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FZeRO%3A-Memory-optimizations-Toward-Training-Trillion-Rajbhandari-Rasley\u002F00c957711b12468cb38424caccdf5291bb354033)  |\n| ✅ |  2022 | [Pathways](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2203.12533.pdf) |  将Jax拓展到上千TPU核上       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F512e9aa873a1cd7eb61ce1f25ca7df6acb7e2352%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FPathways%3A-Asynchronous-Distributed-Dataflow-for-ML-Barham-Chowdhery\u002F512e9aa873a1cd7eb61ce1f25ca7df6acb7e2352)  |\n\n### 图神经网络\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n|  ✅ |  2021 | [图神经网络介绍](https:\u002F\u002Fdistill.pub\u002F2021\u002Fgnn-intro\u002F) | GNN的可视化介绍                  |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F2c0e0440882a42be752268d0b64243243d752a74%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FA-Gentle-Introduction-to-Graph-Neural-Networks-S%C3%A1nchez-Lengeling-Reif\u002F2c0e0440882a42be752268d0b64243243d752a74)  |\n\n### 优化算法\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| | 2014 | [Adam](https:\u002F\u002Farxiv.org\u002Fabs\u002F1412.6980) | 深度学习里最常用的优化算法之一                   |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fa6cb366736791bcccc5c8639de5a8f9636bf87e8%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAdam%3A-A-Method-for-Stochastic-Optimization-Kingma-Ba\u002Fa6cb366736791bcccc5c8639de5a8f9636bf87e8)  |\n| | 2016 |  [为什么超大的模型泛化性不错](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.03530)   |               |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F54ddb00fa691728944fd8becea90a373d21597cf%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FUnderstanding-deep-learning-requires-rethinking-Zhang-Bengio\u002F54ddb00fa691728944fd8becea90a373d21597cf)  |\n| | 2017 | [为什么Momentum有效](https:\u002F\u002Fdistill.pub\u002F2017\u002Fmomentum\u002F) | Distill的可视化介绍            |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F3e8ccf9d3d843c9855c5d76ab66d3e775384da72%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FWhy-Momentum-Really-Works-Goh\u002F3e8ccf9d3d843c9855c5d76ab66d3e775384da72)  |\n\n\n### 新领域应用\n\n\n| 已录制 | 年份 | 名字                                                         | 简介                 | 引用 |\n| ------ | ---- | ------------------------------------------------------------ | -------------------- | ------------------------------------------------------------ |\n| | 2016 | [AlphaGo](https:\u002F\u002Fstorage.googleapis.com\u002Fdeepmind-media\u002Falphago\u002FAlphaGoNaturePaper.pdf) | 强化学习出圈                 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F846aedd869a00c09b40f1f1f35673cb22bc87490%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FMastering-the-game-of-Go-with-deep-neural-networks-Silver-Huang\u002F846aedd869a00c09b40f1f1f35673cb22bc87490)  |\n| | 2020 | [AlphaFold](https:\u002F\u002Fdiscovery.ucl.ac.uk\u002Fid\u002Feprint\u002F10089234\u002F1\u002F343019_3_art_0_py4t4l_convrt.pdf) | 赢得比赛的的蛋白质3D结构预测 |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F3a083d843f891b3574494c385699c21766ce8b7a%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FImproved-protein-structure-prediction-using-from-Senior-Evans\u002F3a083d843f891b3574494c385699c21766ce8b7a)  |\n| ✅ | 2021 | [AlphaFold 2](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs41586-021-03819-2.pdf) | 原子级别精度的蛋白质3D结构预测       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Fdc32a984b651256a8ec282be52310e6bd33d9815%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FHighly-accurate-protein-structure-prediction-with-Jumper-Evans\u002Fdc32a984b651256a8ec282be52310e6bd33d9815)  |\n| ✅ | 2021 | [Codex](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2107.03374.pdf) | 使用注释生成代码       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Facbdbf49f9bc3f151b93d9ca9a06009f4f6eb269%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FEvaluating-Large-Language-Models-Trained-on-Code-Chen-Tworek\u002Facbdbf49f9bc3f151b93d9ca9a06009f4f6eb269)  |\n| ✅ | 2021 | [指导数学直觉](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs41586-021-04086-x.pdf) | 分析不同数学物体之前的联系来帮助发现新定理         |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2Ff672b8fb430606fee0bb368f16603531ce1e90c4%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FAdvancing-mathematics-by-guiding-human-intuition-AI-Davies-Velickovic\u002Ff672b8fb430606fee0bb368f16603531ce1e90c4)  |\n| ✅ | 2022 | [AlphaCode](https:\u002F\u002Fstorage.googleapis.com\u002Fdeepmind-media\u002FAlphaCode\u002Fcompetition_level_code_generation_with_alphacode.pdf) | 媲美一般程序员的编程解题水平       |[![citation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdynamic\u002Fjson?label=citation&query=citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F5cbe278b65a81602a864184bbca37de91448a5f5%3Ffields%3DcitationCount)](https:\u002F\u002Fwww.semanticscholar.org\u002Fpaper\u002FCompetition-Level-Code-Generation-with-AlphaCode-Li-Choi\u002F5cbe278b65a81602a864184bbca37de91448a5f5)  |","# paper-reading 快速上手指南\n\n`paper-reading` 并非一个需要安装运行的软件工具，而是一个**深度学习论文精读视频资源库**。该项目由作者录制并整理了大量前沿 AI 论文（如 Sora, Llama 3.1, GPT-4, Whisper 等）的深度解读视频。\n\n开发者无需配置环境或运行代码，直接通过视频平台观看即可获取知识。\n\n## 1. 环境准备\n\n本项目无系统要求或前置依赖。您只需要：\n*   **网络设备**：能够访问互联网。\n*   **播放平台**：推荐使用 **哔哩哔哩 (Bilibili)** 以获得最佳国内访问速度，或使用 **YouTube**。\n*   **浏览器**：任意现代浏览器（Chrome, Edge, Firefox 等）。\n\n## 2. 访问方式\n\n无需执行安装命令。请直接访问以下平台查看最新录制的论文精读列表：\n\n*   **国内首选（哔哩哔哩）**：\n    访问作者 B 站主页或直接点击 README 中的视频链接。\n    *   优势：无需特殊网络环境，加载速度快，支持高清播放。\n    \n*   **国际备选（YouTube）**：\n    访问作者 YouTube 频道。\n    *   优势：适合海外开发者或习惯使用 Google 生态的用户。\n\n## 3. 基本使用\n\n### 浏览论文列表\n在项目的 `README.md` 中，作者维护了一份详细的**录制完成的论文**表格。表格包含以下关键信息：\n*   **日期**：视频发布时间。\n*   **标题**：论文名称及解读重点（例如：\"Llama 3.1 论文精读 · 模型训练过程”）。\n*   **时长**：视频长度，方便安排学习时间。\n*   **视频链接**：直接跳转到 B 站或 YouTube 播放页。\n\n### 学习示例\n假设您想学习 **OpenAI Sora** 的技术细节：\n\n1.  在 README 表格中找到第一行：`1\u002F10\u002F25 | OpenAI Sora 上 (包含 Movie Gen 和 HunyuanVideo)`。\n2.  点击对应的 **Bilibili** 徽章图标或链接。\n3.  观看时长为 `1:04:18` 的精读视频。\n4.  视频中通常包含对论文核心架构、训练数据、创新点及代码实现的详细拆解。\n\n### 推荐学习路径\n根据表格内容，建议按以下顺序进阶学习：\n1.  **基础入门**：从早期的 `CLIP`, `ViLT`, `DETR` 等多模态基础论文开始。\n2.  **大模型核心**：深入学习 `Llama 3.1` 系列（导言 -> 预训练数据 -> 模型架构 -> 训练基础设施 -> 训练过程）。\n3.  **前沿追踪**：观看最新的 `OpenAI Sora` 及 `GPT-4` 解读，了解行业最新动态。\n4.  **科研方法论**：参考《研究的艺术》系列视频，提升科研写作与逻辑思维能力的软技能。\n\n> **提示**：部分视频标题旁附有 arXiv 论文链接，建议先下载论文原文，配合视频进行对照阅读，效果更佳。","某 AI 算法工程师正急需复现 Sora 或 Llama 3.1 的核心机制，面对长达数十页、充满复杂公式与专业术语的英文论文，感到无从下手。\n\n### 没有 paper-reading 时\n- **语言与术语壁垒高**：非英语母语者需频繁查阅词典，对\"World Simulators\"或特定训练架构术语理解吃力，阅读速度极慢。\n- **核心逻辑难以捕捉**：容易陷入数学推导的细节迷宫，花费数小时仍无法理清模型训练的整体流程和数据构建思路。\n- **时间成本高昂**：精读一篇如 GPT-4 或 Llama 3.1 的重磅论文往往需要整整一个周末，严重挤占代码复现与实验调试的时间。\n- **缺乏直观认知**：仅靠文字描述难以想象视频生成模型的世界模拟过程或大规模集群的训练拓扑，理解停留在抽象层面。\n\n### 使用 paper-reading 后\n- **逐段精讲破障碍**：直接观看针对 OpenAI Sora 或 Llama 3.1 的逐段精读视频，专家用中文通俗拆解术语，瞬间扫清语言盲区。\n- **脉络梳理更清晰**：通过视频中对“预训练数据”、“训练 Infra\"及“模型架构”的分章节深度剖析，快速掌握技术演进的关键路径。\n- **学习效率倍增**：利用 20-60 分钟的高质量视频内容，即可替代数天的啃书过程，将节省下的时间投入到实际的代码验证中。\n- **可视化辅助理解**：结合封面图示与视频讲解，将抽象的“世界模拟器”概念转化为具象的认知，深刻理解数据如何驱动模型生成。\n\npaper-reading 通过将晦涩的学术文献转化为结构化的视频课程，让开发者能以最低的时间成本穿透技术迷雾，直达深度学习前沿的核心逻辑。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmli_paper-reading_2c8bc031.png","mli","Mu Li","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fmli_4b5cab9c.jpg",null,"Boson AI","Palo Alto, CA","https:\u002F\u002Fgithub.com\u002Fmli",32830,2781,"2026-04-07T15:35:47","Apache-2.0",1,"","未说明",{"notes":87,"python":85,"dependencies":88},"该项目并非可运行的 AI 软件工具，而是一个深度学习论文精读的视频教程列表（托管于 Bilibili 和 YouTube）。README 中仅包含视频标题、封面、时长及观看链接，不涉及任何代码运行环境、依赖库或硬件配置需求。用户只需具备能播放视频的设备即可观看内容。",[],[14],[91,92,93],"deep-learning","paper","reading-list","2026-03-27T02:49:30.150509","2026-04-08T05:10:41.093165",[],[]]