The model does not follow the instructions

Hello, I would like to know if, in order for the model to follow the instructions, I need to change any parameters or create a different prompt. At the moment, any instruction I add is not respected. Here is an example of a prompt. Or do I need to add a custom system prompt?

Temperature: 0.2

Prompt:

```
Please provide a detailed description of the video, focusing on:

- The viewing angle (front, side, back, etc.)
- The position and orientation of the subject/object
- Notable details or actions visible
- Background or environment elements
- Other relevant visual details.

Additional Instructions to Follow:
- Start your final description with the word 'l0f1'
```

The issue names no source files, tests, or entry points. Start by reproducing the provided video-description prompt with the stated temperature, then determine whether the model or its prompting interface supports a custom system prompt; done would require a confirmed explanation or documented fix for enforcing the requested opening word.

Written by the indexing model from the issue text.