[BUG] Image recognition content is missing

[szhjia]

Describe the bug
During the process of image recognition, the leftmost part of the image is consistently missing. No matter how the prompt is modified, the system fails to recognize and provide the specific content from that section.

To Reproduce
Upload an image of middle school mathematical rules and theorems.

The image content is divided into three columns.

The prompt is as follows:

The image contains common middle school mathematical rules and theorems. Please process the image according to the following requirements:

There are a total of 75 rules and theorems in the image.

Organize them into a table.

Expand on the specific content.

Check the completeness of the content.

Ensure there are no omissions.

Expected behavior
The system should be able to recognize and include the content from the leftmost part of the image.

Screenshots

Additional context
I think it is a common issue, would you please fix ASAP, thanks.

[mowentian]

看上去和 ocr 的模型能力有关，我们研究一下

[vishalmishra369]

if you don't mind would you like me to correct this issue as i am exploring and trying todo open source contributions
if yes then please assign it to me, thanks.

[szhjia]

Hi there ,any update for this ?thanks

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. If you believe this issue is still relevant, please leave a comment to keep it open. Thank you for your contributions!

[github-actions]

false