[Badcase]: Poor quality of outputs in Russian language

[ilya-corneli]

Model Series

Qwen2.5

What are the models used?

Qwen2.5 7B

What is the scenario where the problem happened?

Qwen2.5 7B

Is this badcase known and can it be solved using avaiable techniques?

• I have followed the GitHub README.
• I have checked the Qwen documentation and cannot find a solution there.
• I have checked the documentation of the related framework and cannot find useful information.
• I have searched the issues and there is not a similar one.

Information about environment

The model does not work well in Anything LLM, despite the fact that the manual asks to give answers in Russian, the model gives mixed answers in Russian and Chinese, no clarification does not help to correct the situation.

Description

Steps to reproduce

This happens to Qwen2.5-xB-Instruct-xxx and xxx.
The badcase can be reproduced with the following steps:

1. ...
2. ...

The following example input & output can be used:

system: ...
user: ...
...

Expected results

The results are expected to be ...

Attempts to fix

I have tried several ways to fix this, including:

1. adjusting the sampling parameters, but ...
2. prompt engineering, but ...

Anything else helpful for investigation

I find that this problem also happens to ...