poor text handling

[pseudotensor]

Reminder

• [X] I have searched the Github Discussion and issues and have not found anything similar to this.

Environment

ubuntu 20
python 3.10
torch 2.1.2
cuda 12.1

Current Behavior

Expected Behavior

No repetition and description of image's text

Steps to Reproduce

prompt:

Describe the image in detail.

Tried many other prompts, same kinds of failures.

Anything Else?

No response

[Yimi81]

The current version of the model does not have enhancements for OCR capabilities at the moment.