GPT-4o with canvas | 盛溪的博客

type

status

date

slug

summary

Better collaboration with ChatGPT与 ChatGPT 更好的协作

People use ChatGPT every day for help with writing and code. Although the chat interface is easy to use and works well for many tasks, it’s limited when you want to work on projects that require editing and revisions. Canvas offers a new interface for this kind of work.人们每天都使用 ChatGPT 来寻求写作和编码方面的帮助。尽管聊天界面易于使用并且适用于许多任务，但当您想要处理需要编辑和修订的项目时，它会受到限制。 Canvas 为此类工作提供了新的界面。

With canvas, ChatGPT can better understand the context of what you’re trying to accomplish. You can highlight specific sections to indicate exactly what you want ChatGPT to focus on. Like a copy editor or code reviewer, it can give inline feedback and suggestions with the entire project in mind.借助画布，ChatGPT 可以更好地理解您想要完成的任务的上下文。您可以突出显示特定部分以准确指示您希望 ChatGPT 关注的内容。就像文本编辑器或代码审阅者一样，它可以针对整个项目提供内联反馈和建议。

You control the project in canvas. You can directly edit text or code. There’s a menu of shortcuts for you to ask ChatGPT to adjust writing length, debug your code, and quickly perform other useful actions. You can also restore previous versions of your work by using the back button in canvas.您可以在画布中控制项目。您可以直接编辑文本或代码。有一个快捷方式菜单，您可以要求 ChatGPT 调整书写长度、调试代码以及快速执行其他有用的操作。您还可以使用画布中的后退按钮恢复以前版本的作品。

Canvas opens automatically when ChatGPT detects a scenario in which it could be helpful. You can also include “use canvas” in your prompt to open canvas and use it to work on an existing project.当 ChatGPT 检测到可能有帮助的场景时，Canvas 会自动打开。您还可以在提示中包含“使用画布”以打开画布并使用它来处理现有项目。

Writing shortcuts include:书写快捷方式包括：

Suggest edits: ChatGPT offers inline suggestions and feedback.建议编辑： ChatGPT 提供内联建议和反馈。

Adjust the length: Edits the document length to be shorter or longer.调整长度：将文档长度编辑得更短或更长。

Change reading level: Adjusts the reading level, from Kindergarten to Graduate School.更改阅读级别：调整阅读级别，从幼儿园到研究生院。

Add final polish: Checks for grammar, clarity, and consistency.最后润色：检查语法、清晰度和一致性。

Add emojis: Adds relevant emojis for emphasis and color.添加表情符号：添加相关表情符号以强调重点和颜色。

Coding in canvas 在画布上编码

Coding is an iterative process, and it can be hard to follow all the revisions to your code in chat. Canvas makes it easier to track and understand ChatGPT’s changes, and we plan to continue improving transparency into these kinds of edits.编码是一个迭代过程，很难在聊天中跟踪代码的所有修订。 Canvas 使跟踪和理解 ChatGPT 的更改变得更加容易，我们计划继续提高此类编辑的透明度。Coding shortcuts include:编码快捷键包括：

Review code: ChatGPT provides inline suggestions to improve your code.审查代码： ChatGPT 提供内联建议来改进您的代码。

Add logs: Inserts print statements to help you debug and understand your code.添加日志：插入打印语句以帮助您调试和理解代码。

Add comments: Adds comments to the code to make it easier to understand.添加注释：为代码添加注释，使其更易于理解。

Fix bugs: Detects and rewrites problematic code to resolve errors.修复错误：检测并重写有问题的代码以解决错误。

Port to a language: Translates your code into JavaScript, TypeScript, Python, Java, C++, or PHP.移植到一种语言：将您的代码转换为 JavaScript、TypeScript、Python、Java、C++ 或 PHP。

Training the model to become a collaborator训练模型成为合作者

We trained GPT-4o to collaborate as a creative partner. The model knows when to open a canvas, make targeted edits, and fully rewrite. It also understands broader context to provide precise feedback and suggestions.我们训练 GPT-4o 作为创意合作伙伴进行协作。该模型知道何时打开画布、进行有针对性的编辑以及完全重写。它还了解更广泛的背景，以提供准确的反馈和建议。

To support this, our research team developed the following core behaviors:为了支持这一点，我们的研究团队制定了以下核心行为：

Triggering the canvas for writing and coding 触发画布进行书写和编码

Generating diverse content types生成多样化的内容类型

Making targeted edits 进行有针对性的编辑

Rewriting documents 重写文档

Providing inline critique提供内联批评

We measured progress with over 20 automated internal evaluations. We used novel synthetic data generation techniques, such as distilling outputs from OpenAI o1-preview, to post-train the model for its core behaviors. This approach allowed us to rapidly address writing quality and new user interactions, all without relying on human-generated data.我们通过 20 多项自动内部评估来衡量进展情况。我们使用新颖的合成数据生成技术（例如从 OpenAI o1-preview 中提取输出）来对模型的核心行为进行后训练。这种方法使我们能够快速解决写作质量和新的用户交互问题，而无需依赖人工生成的数据。

A key challenge was defining when to trigger a canvas. We taught the model to open a canvas for prompts like “Write a blog post about the history of coffee beans” while avoiding over-triggering for general Q&A tasks like “Help me cook a new recipe for dinner.” For writing tasks, we prioritized improving “correct triggers” (at the expense of “correct non-triggers”), reaching 83% compared to a baseline zero-shot GPT-4o with prompted instructions.一个关键的挑战是定义何时触发画布。我们教模型打开画布以显示“写一篇关于咖啡豆历史的博客文章”等提示，同时避免过度触发“帮我做一份新的晚餐食谱”等一般问答任务。对于编写任务，我们优先改进“正确的触发器”（以牺牲“正确的非触发器”为代价），与带有提示指令的基准零样本 GPT-4o 相比，达到了 83%。

It’s worth noting that the quality of such baselines is highly sensitive to the specific prompt used. With different prompts, the baseline may still perform poorly but in a different manner—for instance, by being evenly inaccurate across coding and writing tasks, resulting in a different distribution of errors and alternative forms of suboptimal performance. For coding, we intentionally biased the model against triggering to avoid disrupting our power users. We'll continue refining this based on user feedback.值得注意的是，此类基线的质量对所使用的特定提示高度敏感。在不同的提示下，基线可能仍然表现不佳，但方式不同，例如，在编码和写作任务中均匀不准确，导致错误分布不同以及次优性能的替代形式。对于编码，我们故意使模型偏向于触发，以避免干扰我们的高级用户。我们将根据用户反馈继续完善这一点。

For writing and coding tasks, we improved correctly triggering the canvas decision boundary, reaching 83% and 94% respectively compared to a baseline zero-shot GPT-4o with prompted instructions.对于写作和编码任务，我们改进了画布决策边界的正确触发，与带有提示指令的基线零样本 GPT-4o 相比，分别达到了 83% 和 94%。

A second challenge involved tuning the model's editing behavior once the canvas was triggered—specifically deciding when to make a targeted edit versus rewriting the entire content. We trained the model to perform targeted edits when users explicitly select text through the interface, otherwise favoring rewrites. This behavior continues to evolve as we refine the model.第二个挑战涉及在触发画布后调整模型的编辑行为，具体决定何时进行有针对性的编辑而不是重写整个内容。我们训练模型在用户通过界面明确选择文本时执行有针对性的编辑，否则倾向于重写。随着我们完善模型，这种行为不断发展。