[{"data":1,"prerenderedAt":81},["ShallowReactive",2],{"tool-cmmlu":3},{"tool":4,"relatedTools":43},{"id":5,"name":6,"slug":7,"tagline":8,"description":9,"websiteUrl":10,"logoText":11,"pricing":12,"editorScore":13,"clickCount":14,"featured":15,"latest":15,"updatedAt":16,"category":17,"tags":26,"overview":28,"highlights":29,"useCases":38},"b9ac668d63adbb3ea59a8f255cf9e7ae","CMMLU","cmmlu","一个综合性的大模型中文评估基准","CMMLU是综合性的中文评估基准，专门用在评估语言模型在中文语境下的知识和推理能力，涵盖从基础学科到高级专业水平的67个主题。包括需要计算和推理的自然科学，需要知识的人文科学和社会科学，及需要生活常识的中国驾驶规则等。","https:\u002F\u002Fgithub.com\u002Fhaonan-li\u002FCMMLU\u002F","CMM","官网可查看",98,4,false,"2026-04-15T04:26:34+08:00",{"id":18,"name":19,"slug":20,"description":21,"iconLabel":22,"accentColor":23,"toolCount":24,"sortOrder":25},"d6d303a9f41677e5f1a5f9b12c520225","智能体与自动化","agents","适合搭建业务工作流、插件编排和多步骤自动化，用来把 AI 接进真实流程。","AUTO","#6C7A2D",216,7,[27],"AI模型评测","CMMLU是综合性的中文评估基准，专门用在评估语言模型在中文语境下的知识和推理能力，涵盖从基础学科到高级专业水平的67个主题。包括需要计算和推理的自然科学，需要知识的人文科学和社会科学，及需要生活常识的中国驾驶规则等。CMMLU中的许多任务具有中国特定的答案，可能在其他地区或语言中并不普遍适用。CMMLU提供丰富的测试数据和排行榜，支持多种评估方式，如five-shot和zero-shot测试，是衡量中文语言模型性能的重要工具。 排行榜 ：展示不同语言模型在five-shot和zero-shot测试下的表现，帮助比较模型性能。",[30,31,32,33,34,35,36,37],"排行榜","数据集","预处理代码","评估工具","获取数据集","从GitHub下载","通过Hugging Face获取","准备测试环境",[39,40,41,42],"从基础学科到高级专业水平的67个主题","需要计算和推理的自然科学","丰富的测试数据和排行榜","开发和测试数据",[44,57,68],{"id":45,"name":46,"slug":47,"tagline":48,"description":49,"websiteUrl":50,"logoText":51,"pricing":12,"editorScore":13,"clickCount":52,"featured":53,"latest":53,"updatedAt":16,"category":54,"tags":55},"a638bfd4cc081d0294d7d12975a727dd","GPT-4","gpt-4","OpenAI旗下最新的GPT-4模型","GPT-4（Generative Pre-trained Transformer）是 OpenAI 开发的自然语言处理模型 GPT 家族中的第四个版本，该模型依靠强大的神经网络来理解和生成类似人类的语言。","https:\u002F\u002Fcn.bing.com\u002Fsearch?q=GPT-4&ensearch=1","GPT",2356,true,{"id":18,"name":19,"slug":20,"description":21,"iconLabel":22,"accentColor":23,"toolCount":24,"sortOrder":25},[56],"AI训练模型",{"id":58,"name":59,"slug":60,"tagline":61,"description":62,"websiteUrl":63,"logoText":64,"pricing":12,"editorScore":13,"clickCount":65,"featured":53,"latest":53,"updatedAt":16,"category":66,"tags":67},"a23e89d9e4203bd1c683c41e837f370e","LMArena","lmarena","AI模型评估平台","LMArena是加州大学伯克利分校推出的创新AI模型评估平台，基于让用户对不同AI模型的回答进行匿名投票，衡量模型的表现。用户输入问题后，平台提供两个模型的回答，用户根据偏好选择更优答案，投票结果直接塑造公共排行榜。","https:\u002F\u002Flmarena.ai\u002F?utm_source=ai-bot.cn","LMA",681,{"id":18,"name":19,"slug":20,"description":21,"iconLabel":22,"accentColor":23,"toolCount":24,"sortOrder":25},[27],{"id":69,"name":70,"slug":71,"tagline":72,"description":73,"websiteUrl":74,"logoText":75,"pricing":76,"editorScore":13,"clickCount":77,"featured":53,"latest":53,"updatedAt":16,"category":78,"tags":79},"32bb7d109995125694dc4a175a840117","Coze","coze","海量AI智能体免费用，一键复制同款","Coze是字节跳动推出的AI办公助手，能通过AI技术简化工作流程，提升工作效率。通过Agent Skills技能系统和Agent Plan长期计划功能，实现从单一工具到智能工作伙伴的转变。","https:\u002F\u002Fwww.coze.cn\u002Foverview?utm_medium=daohang&utm_source=aibot&utm_term=49156769m","COZ","免费",316,{"id":18,"name":19,"slug":20,"description":21,"iconLabel":22,"accentColor":23,"toolCount":24,"sortOrder":25},[80],"AI开发平台",1776262182475]