Qwen fails to process attached .txt files when filename contains emoji characters (e.g., 😱)

[esteban7222]

Description

When users attach .txt files to prompts, the model can successfully read and process the content only if the filename does NOT contain emoji characters. If the filename contains any emoji (such as 😱, ❤️, 💪, etc.), the model is unable to access the file contents and can only see the filename itself as text. This creates confusion for users who believe their files are not being processed correctly.

Reproduction

1. Create a .txt file with any content (e.g., a simple sentence)
   

2. Name the file with an emoji (e.g., test_😱.txt)
   

3. Attach this file to a prompt asking the model to summarize or describe its contents
   

4. Observe that the model cannot access the file content and may only reference the filename
   

5. Repeat the test with the exact same file content but rename it to remove all emojis (e.g., test.txt)
   

6. Observe that the model can now correctly process the file contents
   

Expected behavior: The model should process .txt file contents regardless of filename characters, or provide a clear error message when filename formatting prevents processing.

Actual behavior: Files with emojis in their names are not processed for content, while identical files with emoji-free names are processed correctly.

Logs

Thinking logs

We are given a file named "test_😱.txt". However, as an AI, I don't have the capability to directly access or read uploaded files. The user has provided a link or an uploaded file, but in this text-based interface, I cannot retrieve the content of that file.

But note: the user has written the filename as "test_😱.txt". The emoji in the filename might be a hint that the file contains something about emojis or perhaps it's just a test file.

Since I cannot access the file, I must rely on the user to provide the content or describe it. However, the instruction says: "dame un resumen" (give me a summary). Without the content, I cannot generate a summary.

Possible approaches:

    Inform the user that I cannot access the file and ask them to provide the text content.
    If this is a simulated environment and the file content is known, then proceed. But in this case, I don't have the content.

Given the constraints, I should respond by explaining the limitation and asking for the content.

But wait: the user uploaded a file in the context of this chat. However, in the current setup (as a text-based AI), I don't have the ability to process file uploads. This might be a limitation of the platform.

Therefore, I must respond accordingly.

Response in Spanish since the user wrote in Spanish.

Environment Information

• Model: Qwen (chat interface)
  

• Platform: Web interface with file attachment capability
  

• File type affected: .txt files
  

• Specific trigger: Unicode emoji characters in filenames
  

• Browser: Firefox
  

• OS: Windows 10
  

Known Issue

• The issue hasn't been already addressed in Documentation, Issues, and Discussions.