-
-
Notifications
You must be signed in to change notification settings - Fork 98
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Incorrect numbering in ordered lists #22
Comments
@ansysic Could you please share this chat and paste the share URL here for easier debugging? Looks like the completion markdown is parsed as a bunch of ordered lists resulting in a DOM structure like this <ol>
<li>Point</li>
</ol>
<ol>
<li>Point</li>
</ol> That would explain the large spacing and the numbering. Does it also happen if you rephrase your prompt as "Give me a list of ..." or add that to the context like here: https://slickgpt.vercel.app/shared/flat-poor-train |
Unfortunately I don't have sharing features enabled. I could however reproduce it. Source code of the screenshot above:
|
Thanks for the clarification. The problem is that there are some linebreaks in the completion that our Markdown parser (mis)interprets as different list entities. This is a bit fiddly to fix and might have side effects on other completions. Because we have a workaround (asking for a list explicitly), I'd also consider this a low priority issue. I'll leave the issue open in case somebody wants to give it a shot and for documentation purposes. |
@jondcoleman |
It seems it simply does not respect the numbers of the list items and always renders them as "1". For example, I have the reduced answer (shortened the text and removed code blocks content, but keep {
"role": "assistant",
"content": "Here are a few more ways to test:\n\n4. **Scan-based testing**: In this approach...\n\nExample:\n```...```\n5. **Boundary-scan testing**: In this approach...\n\nExample:\n```...```\n6. **Structural testing**: In this approach...\n\nExample:\n```...```\nThese are just a few more examples...."
} And this response is rendered as 1... 1... 1. list instead of 4... 5... 6. |
Numbered lists don't go higher than 1. Also spacing between is off.
Example:
![image](https://user-images.githubusercontent.com/5899228/228568889-2d2b20e3-434f-455e-bafe-d2f8b140a798.png)
The text was updated successfully, but these errors were encountered: