WebMC
收藏魔搭社区2025-12-05 更新2025-09-27 收录
下载链接：
https://modelscope.cn/datasets/LLM360/WebMC
下载链接
链接失效反馈官方服务：
资源简介：
This is the benchmark used in Crystal, including 600 multiple-choice questions regarding website understanding, editing and generation.
## Introduction
Large language models have demonstrated their potential and garnered increased attention, particularly in the realm of software development engineering, where there's a growing demand for models that specialize in code. Yet, a crucial area remains underexplored: applications related to websites, such as WebAgent [1]. To improve the integration of HTML code structure with natural language content and to advance the evaluation of web-related skills, we have created a new dataset, called WebMC. This dataset comprises training data, including 50,000 instruction pairs with questions and answers for instruction tuning, as well as evaluation data featuring 600 multiple-choice questions.
The development of our WebMC dataset adheres to the methodology of the Code Alpaca [2] pipeline, which itself is derived from the original Alpaca [3] pipeline. Our approach involves crafting data generation prompts that establish clear guidelines for generating HTML data, specifying the instructions, input, and output formats. After these prompts, we put a variety of examples at random from a sample set, all of which have been manually chosen by humans. We then use GPT-3.5 to generate more data samples, adhering to the established rules and following the provided examples. This process is similar for both training and evaluation phases, albeit with variations in the prompts and the formats required. By modifying the **system prompt** and **pre-selected examples**, we can influence GPT to produce the desired output.
## Prompts for data generation:
<pre>
You are asked to come up with a set of 20 diverse website source code generation task instructions. These task instructions will be given to a GPT model and we will evaluate the GPT model for completing the instructions.
Here are the requirements: <font color="green"># requirements about HTML</font>
Try not to repeat the verb for each instruction to maximize diversity.
…
…
List of 20 tasks:
1. Instruction: …. Input: … Output: …. <font color="green"># sample from pre-selected examples</font>
2. Instruction: …. Input: … Output: …. <font color="green"># sample from pre-selected examples</font>
3. <font color="green"># generate by GPT</font>
4. …
5. …
</pre>
Below, we present examples for each training and evaluation task. In the **training** phase, there are two types of instruction pairs: the first type pertains to generation, while the second type focuses on editing.
## Generated training data example:
**Generation** example for training:
"instruction": "Create the homepage for a restaurant website that showcases a carousel of their top dishes and includes an interactive Google Maps view of their location.",
"input": "The restaurant's name is 'Le Bistro'. The restaurant is located at '40.748817,-73.985428'. The top dishes to display are 'Spaghetti Carbonara', 'French Onion Soup', 'Lobster Thermidor', and 'Filet Mignon'.",
"output": "<!DOCTYPE html>\n<html lang=\"en\">\n<head>\n <meta charset=\"UTF-8\">\n <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n <link rel=\"stylesheet\" href=\"https://stackpath.bootstrapcdn.com/bootstrap/4.5.2/css/bootstrap.min.css\" integrity=\"sha384-pzjw8f+ua7Kw1TIq0v8FqFjcJ6pajs/rfdfs3SO+kD4Ck5BdPtF+to8xM6B5z6W5\" crossorigin=\"anonymous\">\n <title>Le Bistro</title>\n <style>\n #map {\n height: 200px;\n }\n .carousel-item {\n height: 200px;\n background-size: cover;\n background-position: center;\n }\n </style>\n <script>\n function initMap() {\n var location = {lat: 40.748817, lng: -73.985428};\n var map = new google.maps.Map(document.getElementById('map'), {\n zoom: 16,\n center: location\n });\n var marker = new google.maps.Marker({\n position: location,\n map: map\n });\n }\n </script>\n</head>\n<body onload=\"initMap()\">\n <h1>Welcome to Le Bistro</h1>\n <div id=\"carouselExampleIndicators\" class=\"carousel slide\" data-ride=\"carousel\">\n <div class=\"carousel-inner\">\n <div class=\"carousel-item active\" style=\"background-image: url('img/carbonara.jpg');\">\n <div class=\"carousel-caption d-none d-md-block\">\n <h5>Spaghetti Carbonara</h5>\n </div>\n </div>\n <div class=\"carousel-item\" style=\"background-image: url('img/onion-soup.jpg');\">\n <div class=\"carousel-caption d-none d-md-block\">\n <h5>French Onion Soup</h5>\n </div>\n </div>\n <div class=\"carousel-item\" style=\"background-image: url('img/lobster.jpg');\">\n <div class=\"carousel-caption d-none d-md-block\">\n <h5>Lobster Thermidor</h5>\n </div>\n </div>\n <div class=\"carousel-item\" style=\"background-image: url('img/filet-mignon.jpg');\">\n <div class=\"carousel-caption d-none d-md-block\">\n <h5>Filet Mignon</h5>\n </div>\n </div>\n </div>\n </div>\n <div id=\"map\"></div>\n <script src=\"https://maps.googleapis.com/maps/api/js?key=YOUR_API_KEY&callback=initMap\"></script>\n</body>\n</html>",
**Editing** example for training:
"instruction": "Redesign an existing site of non-profit organization focused on marine preservation. Use shades of blue for the color scheme and include images of marine life. Also, change the main sections to \"About Us\", \"Our Mission\", \"Join Us\" and \"Support Us\". Here is the existing code.",
"input": "<!DOCTYPE html>\n<html lang=\"en\">\n<head>\n <meta charset=\"UTF-8\">\n <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n <title>Generic Non-Profit</title>\n <style>\n body {\n font-family: 'Times New Roman', serif;\n background-color: #f5f5f5;\n }\n header {\n background-color: #663300;\n padding: 20px;\n text-align: center;\n }\n #about {\n background-color: #fff;\n padding: 20px;\n }\n </style>\n</head>\n<body>\n <header>\n <h1>Generic Non-Profit</h1>\n </header>\n <section id=\"about\">\n <h2>About</h2>\n <p>Generic details here.</p>\n </section>\n</body>\n</html>",
"output": "<!DOCTYPE html>\n<html lang=\"en\">\n<head>\n <meta charset=\"UTF-8\">\n <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n <title>Marine Preservation Society</title>\n <style>\n body {\n font-family: 'Arial', sans-serif;\n background-color: #e6f7ff;\n }\n header {\n background-color: #0070b3;\n padding: 20px;\n text-align: center;\n color: #ffffff;\n }\n section {\n background-color: #ffffff;\n padding: 20px;\n margin-bottom: 20px;\n }\n </style>\n</head>\n<body>\n <header>\n <h1>Marine Preservation Society</h1>\n </header>\n <section id=\"about\">\n <h2>About Us</h2>\n <p>We are dedicated to preserving marine life and habitats.</p>\n <img src=\"about_us_image.jpg\" alt=\"Marine Life\">\n </section>\n <section id=\"mission\">\n <h2>Our Mission</h2>\n <p>Our mission is to protect and restore the world's oceans.</p>\n <img src=\"our_mission_image.jpg\" alt=\"Ocean view\">\n </section>\n <section id=\"join\">\n <h2>Join Us</h2>\n <p>Help us in our mission by becoming a member.</p>\n <img src=\"join_us_image.jpg\" alt=\"Beach Cleanup\">\n </section>\n <section id=\"support\">\n <h2>Support Us</h2>\n <p>Make a difference by donating or volunteering in our projects.</p>\n <img src=\"support_us_image.jpg\" alt=\"Support Us\">\n </section>\n</body>\n</html>",
## Generated evaluation data example:
During the **evaluation** phase, three specific tasks are formulated, each consisting of 200 questions: generation, editing, and understanding. In the generation and editing tasks, the objective is to select the most suitable source codes that fulfill the respective requirements. In the understanding task, questions are posed regarding both the HTML source code structure and content, and the optimal solution is identified.
**Generation** task for evaluation:
"instruction": "Generate an HTML code for a photography portfolio website. The main page should provide your name as a photographer, a brief biography, and a rotating image carousel showcasing selected works. There should also be navigation links to sections such as \"Collections\", \"Contact\", and \"About me\".",
"input": "A.<!DOCTYPE html>\n<html>\n<head>\n <title>Photography Portfolio</title>\n</head>\n<body>\n <header>\n <h1>John Doe Photography</h1>\n <p>Welcome! I'm John Doe, a passionate photographer.</p>\n <nav>\n <a href=\"#collections\">Collections</a>\n <a href=\"#about\">About</a>\n <a href=\"#contact\">Contact</a>\n </nav>\n <!-- Image Carousel -->\n </header>\n</body>\n</html> \n\nB.<!DOCTYPE html>\n<html>\n<head>\n <title>Photography Portfolio</title>\n</head>\n<body>\n <h1>John Doe Photography</h1>\n <nav>\n <a href=\"#about\">About</a>\n <a href=\"#collections\">Collections</a>\n <a href=\"#contact\">Contact</a>\n </nav>\n <div id=\"about\">\n <p>Hi! My name is John Doe. Welcome to my photography page. </p>\n <!-- Image Carousel -->\n </div>\n</body>\n</html> \n\nC. <!DOCTYPE html>\n<html>\n<head>\n <title>Photography Portfolio</title>\n</head>\n<body>\n <header>\n <h1>John Doe Photography</h1>\n <nav>\n <a href=\"#collections\">Collections</a>\n <a href=\"#about\">About</a>\n <a href=\"#contact\">Contact</a>\n </nav>\n <p>Welcome! I'm John Doe, a passionate photographer.</p>\n <!-- Image Carousel -->\n </header>\n</body>\n</html> \n\nD.<!DOCTYPE html>\n<html>\n<head>\n <title>Photography Portfolio</title>\n</head>\n<body>\n <div>\n <h1>John Doe Photography</h1>\n <p>Hi, my name's John Doe. Welcome to my photography page. </p>\n <nav>\n <a href=\"#about\">About</a>\n <a href=\"#collections\">Collections</a>\n <a href=\"#contact\">Contact</a>\n </nav>\n <!-- Image Carousel -->\n </div>\n</body>\n</html>",
"output": "C",
**Editing** task for evaluation:
"instruction": "The business portfolio website needs a facelift. I want to add more professional feels to the website by changing the current font-family from \"Arial\" to \"Helvetica\" and also the font-color from \"#000000\" (black) to \"#353535\" (very dark gray) in the presented source code.\n<!DOCTYPE html>\n<html lang=\"en\">\n<head>\n <meta charset=\"UTF-8\">\n <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n <style>\n body{\n font-family: Arial;\n color: #000000;\n }\n </style>\n <title>Business Portfolio</title>\n</head>\n<body>\n <header>\n <h1>Welcome to our Business Portfolio Site!</h1>\n </header>\n</body>\n</html>",
"input": "A. <!DOCTYPE html>\n<html lang=\"en\">\n<head>\n <meta charset=\"UTF-8\">\n <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n <style>\n body{\n font-family: Arial;\n color: #353535;\n }\n </style>\n <title>Business Portfolio</title>\n</head>\n<body>\n <header>\n <h1>Welcome to our Business Portfolio Site!</h1>\n </header>\n</body>\n</html>\n\nB. <!DOCTYPE html>\n<html lang=\"en\">\n<head>\n <meta charset=\"UTF-8\">\n <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n <style>\n body{\n font-family: Helvetica;\n color: #000000;\n }\n </style>\n <title>Business Portfolio</title>\n</head>\n<body>\n <header>\n <h1>Welcome to our Business Portfolio Site!</h1>\n </header>\n</body>\n</html>\n\nC. <!DOCTYPE html>\n<html lang=\"en\">\n<head>\n <meta charset=\"UTF-8\">\n <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n <style>\n body{\n font-family: Helvetica;\n color: #353535;\n }\n </style>\n <title>Business Portfolio</title>\n</head>\n<body>\n <header>\n <h1>Welcome to our Business Portfolio Site!</h1>\n </header>\n</body>\n</html>\n\nD. <!DOCTYPE html>\n<html lang=\"en\">\n<head>\n <meta charset=\"UTF-8\">\n <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n <style>\n body{\n font-family: Arial;\n color: #000000;\n }\n </style>\n <title>Business Portfolio</title>\n</head>\n<body>\n <header>\n <h1>Welcome to our Business Portfolio Site!</h1>\n </header>\n</body>\n</html>",
"output": "C",
**Understanding** task for evaluation:
"instruction": "Source codes: <!DOCTYPE html> <html lang=\"en\"> <head> <meta charset=\"UTF-8\"> <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\"> <title>Best-Deals.com</title> </head> <body> <header> <h1>Best Deals in Town!</h1> </header> <nav> <a href=\"#electronics\">Electronics</a> <a href=\"#appliances\">Home Appliances</a> <a href=\"#groceries\">Groceries</a> </nav> </body> </html> Which links does the navbar section on this e-commerce site contain?",
"input": "A. Electronics, Home Appliances and Apparel. \n B. Electronics, Groceries and Home Appliances. \n C. Electronics, Books and Home Furniture. \n D. Electronics, Video Games, and Groceries.",
"output": "B",
## Reference
1. Gur, Izzeddin, et al. "A real-world webagent with planning, long context understanding, and program synthesis." arXiv preprint arXiv:2307.12856 (2023).
2. https://github.com/sahil280114/codealpaca
3. Taori, Rohan, et al. "Alpaca: A strong, replicable instruction-following model." Stanford Center for Research on Foundation Models. https://crfm. stanford. edu/2023/03/13/alpaca. html 3.6 (2023): 7.
---
license: apache-2.0
---
本基准集用于Crystal项目，包含600道涉及网站理解、编辑与生成的选择题。
## 引言
大语言模型（Large Language Model, LLM）已展现出巨大潜力并获得了越来越多的关注，尤其在软件开发工程领域，对专精代码的模型需求日益增长。然而，一个关键领域仍未得到充分探索：与网站相关的应用，例如WebAgent[1]。为推动超文本标记语言（HTML）代码结构与自然语言内容的融合，并完善网页相关技能的评估体系，我们构建了一款全新的数据集，命名为WebMC。该数据集包含训练数据与评估数据两部分：训练数据包含50000条问答指令对，用于指令微调；评估数据则包含600道选择题。
我们的WebMC数据集开发遵循Code Alpaca[2]流程的方法论，而该流程本身源自原始Alpaca[3]流程。我们的实现步骤为：首先构建数据生成提示词，为HTML数据生成制定清晰的规范，明确指令、输入与输出格式；随后从人工精选的样本集中随机选取若干示例；再借助GPT-3.5按照既定规则与示例生成更多数据样本。训练与评估阶段的流程类似，但提示词与所需格式存在差异。通过修改**系统提示词（system prompt）**与**预选示例**，我们可以引导GPT生成符合预期的输出。
## 数据生成提示词
<pre>
You are asked to come up with a set of 20 diverse website source code generation task instructions. These task instructions will be given to a GPT model and we will evaluate the GPT model for completing the instructions.
Here are the requirements: <font color="green"># requirements about HTML</font>
Try not to repeat the verb for each instruction to maximize diversity.
…
…
List of 20 tasks:
1. Instruction: …. Input: … Output: …. <font color="green"># sample from pre-selected examples</font>
2. Instruction: …. Input: … Output: …. <font color="green"># sample from pre-selected examples</font>
3. <font color="green"># generate by GPT</font>
4. …
5. …
</pre>
下文将展示训练与评估任务的各类示例。在**训练阶段**，包含两种类型的指令对：第一类针对代码生成任务，第二类则聚焦代码编辑任务。
## 训练生成数据示例
**代码生成**训练示例：
"指令": "为名为'Le Bistro'的餐厅创建主页，要求展示招牌菜品轮播图，并包含其位置的交互式谷歌地图视图。",
"输入": "餐厅名称为'Le Bistro'，地址为'40.748817,-73.985428'。",
"输出": "<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<link rel="stylesheet" href="https://stackpath.bootstrapcdn.com/bootstrap/4.5.2/css/bootstrap.min.css" integrity="sha384-pzjw8f+ua7Kw1TIq0v8FqFjcJ6pajs/rfdfs3SO+kD4Ck5BdPtF+to8xM6B5z6W5" crossorigin="anonymous">
<title>Le Bistro</title>
<style>
#map {
height: 200px;
}
.carousel-item {
height: 200px;
background-size: cover;
background-position: center;
}
</style>
<script>
function initMap() {
var location = {lat: 40.748817, lng: -73.985428};
var map = new google.maps.Map(document.getElementById('map'), {
zoom: 16,
center: location
});
var marker = new google.maps.Marker({
position: location,
map: map
});
}
</script>
</head>
<body onload="initMap()">
<h1>Welcome to Le Bistro</h1>
<div id="carouselExampleIndicators" class="carousel slide" data-ride="carousel">
<div class="carousel-inner">
<div class="carousel-item active" style="background-image: url('img/carbonara.jpg');">
<h5>Spaghetti Carbonara</h5>
</div>
</div>
</div>
<div id="map"></div>
<script src="https://maps.googleapis.com/maps/api/js?key=YOUR_API_KEY&callback=initMap"></script>
</body>
</html>"
**代码编辑**训练示例：
"指令": "重新设计一个专注于海洋保护的非营利组织现有网站。要求使用蓝色系作为配色方案，并添加海洋生物相关图片；同时将主要板块修改为'关于我们'、'我们的使命'、'加入我们'与'支持我们'。以下为现有代码。",
"输入": "<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Generic Non-Profit</title>
<style>
body {
font-family: 'Times New Roman', serif;
background-color: #f5f5f5;
}
header {
background-color: #663300;
padding: 20px;
text-align: center;
}
#about {
background-color: #fff;
padding: 20px;
}
</style>
</head>
<body>
<header>
<h1>Generic Non-Profit</h1>
</header>
<section id="about">
<h2>About</h2>
<p>Generic details here.</p>
</section>
</body>
</html>",
"输出": "<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Marine Preservation Society</title>
<style>
body {
font-family: Arial, sans-serif;
background-color: #e6f7ff;
}
header {
background-color: #0070b3;
padding: 20px;
text-align: center;
color: #ffffff;
}
section {
background-color: #ffffff;
padding: 20px;
margin-bottom: 20px;
}
</style>
</head>
<body>
<header>
<h1>Marine Preservation Society</h1>
</header>
<section id="about">
<h2>About Us</h2>
<p>We are dedicated to preserving marine ecosystems and their biodiversity.</p>
</section>
<section id="mission">
<h2>Our Mission</h2>
<p>Our mission is to protect and restore the world's oceans and marine habitats.</p>
</section>
<section id="join">
<h2>Join Us</h2>
<p>Become a member to support our marine conservation efforts.</p>
</section>
<section id="support">
<h2>Support Us</h2>
<p>Your donations will help us continue our important work.</p>
</section>
</body>
</html>"
## 评估生成数据示例
在**评估阶段**，我们设计了三类具体任务，每类包含200道题目：代码生成任务、代码编辑任务与代码理解任务。在生成与编辑任务中，目标为选择最符合对应需求的源代码；在理解任务中，题目围绕HTML源代码的结构与内容展开，需选出最优解。
**代码生成**评估任务：
"指令": "为摄影作品集网站生成HTML代码。主页需展示摄影师姓名、个人简介，以及精选作品轮播图；同时需包含导航链接，指向'作品集'、'关于我'与'联系'板块。",
"输入": "A.<!DOCTYPE html>
<html>
<head>
<title>Photography Portfolio</title>
</head>
<body>
<header>
<h1>John Doe Photography</h1>
<p>Welcome! I'm John Doe, a passionate photographer.</p>
<nav>
<a href="#collections">Collections</a>
<a href="#about">About</a>
<a href="#contact">Contact</a>
</nav>
<!-- Image Carousel -->
</header>
</body>
</html>
B.<!DOCTYPE html>
<html>
<head>
<title>Photography Portfolio</title>
</head>
<body>
<h1>John Doe Photography</h1>
<nav>
<a href="#about">About</a>
<a href="#collections">Collections</a>
<a href="#contact">Contact</a>
</nav>
<div id="about">
<p>Hi, I'm John Doe.</p>
</div>
<!-- Image Carousel -->
</div>
</body>
</html>
C. <!DOCTYPE html>
<html>
<head>
<title>Photography Portfolio</title>
</head>
<body>
<header>
<h1>John Doe Photography</h1>
<nav>
<a href="#collections">Collections</a>
<a href="#about">About</a>
<a href="#contact">Contact</a>
</nav>
<p>Welcome! I'm John Doe, a passionate photographer.</p>
<!-- Image Carousel -->
</header>
</body>
</html>
D.<!DOCTYPE html>
<html>
<head>
<title>Photography Portfolio</title>
</head>
<body>
<div>
<h1>John Doe Photography</h1>
<p>Hi, I'm John Doe.</p>
<nav>
<a href="#about">About</a>
<a href="#collections">Collections</a>
<a href="#contact">Contact</a>
</nav>
<!-- Image Carousel -->
</div>
</body>
</html>",
"输出": "C"
**代码编辑**评估任务：
"指令": "商业作品集网站需要改版。要求将当前字体族从'Arial'修改为'Helvetica'，并将字体颜色从'#000000'修改为'#353535'，以提升网站的专业感。以下为现有源代码。",
"输入": "<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<style>
body{
font-family: Arial;
color: #000000;
}
</style>
<title>Business Portfolio</title>
</head>
<body>
<header>
<h1>Welcome to our Business Portfolio Site!</h1>
</header>
</body>
</html>",
"A. <!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<style>
body{
font-family: Arial;
color: #353535;
}
</style>
<title>Business Portfolio</title>
</head>
<body>
<header>
<h1>Welcome to our Business Portfolio Site!</h1>
</header>
</body>
</html>
B. <!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<style>
body{
font-family: Helvetica;
color: #000000;
}
</style>
<title>Business Portfolio</title>
</head>
<body>
<header>
<h1>Welcome to our Business Portfolio Site!</h1>
</header>
</body>
</html>
C. <!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<style>
body{
font-family: Helvetica;
color: #353535;
}
</style>
<title>Business Portfolio</title>
</head>
<body>
<header>
<h1>Welcome to our Business Portfolio Site!</h1>
</header>
</body>
</html>
D. <!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<style>
body{
font-family: Arial;
color: #000000;
}
</style>
<title>Business Portfolio</title>
</head>
<body>
<header>
<h1>Welcome to our Business Portfolio Site!</h1>
</header>
</body>
</html>",
"输出": "C"
**代码理解**评估任务：
"指令": "现有源代码：<!DOCTYPE html> <html lang="en"> <head> <meta charset="UTF-8"> <meta name="viewport" content="width=device-width, initial-scale=1.0"> <title>Best-Deals.com</title> </head> <body> <header> <h1>Best Deals in Town!</h1> </header> <nav> <a href="#electronics">Electronics</a> <a href="#home-appliances">Home Appliances</a> <a href="#groceries">Groceries</a> </nav> </body> </html> 该电商网站的导航栏包含哪些链接？",
"输入": "A. 电子产品、家居用品与服装。 B. 电子产品、家居用品与杂货。 C. 家居用品、食品与杂货。 D. 食品、杂货与电子产品。",
"输出": "B"
## 参考文献
1. Gur, Izzeddin, 等. 具备规划、长上下文理解与程序合成能力的真实Web智能体. arXiv预印本arXiv:2307.12856 (2023).
2. https://github.com/sahil280114/codealpaca
3. Taori, Rohan, 等. Alpaca：一款强大且可复现的指令跟随模型. 斯坦福基础模型研究中心. https://crfm.stanford.edu/2023/03/13/alpaca.html 3.6 (2023): 7.
---
许可证：Apache-2.0
---
提供机构：
maas
创建时间：
2025-04-07
