GPT-4-1106-preview has trouble response Chinese text in function calling

If the prompts contain the phrase “use Chinese for output,” the resulting responses will consist of 4096 “\n” (newline) characters. For example, \n\n\n\n\n\n\n\n.
The issue is specific to function calls; Chat Completions do not exhibit this problem and the same call works perfectly fine with gpt-3.5-turbo and gpt-4.

2 Likes

I’m using function call for the same model and it outputs the correct parameters and then \n until the 4096 tokens are used up. So it may not be a language specific issue.

Anyway, given the state of OpenAI at the moment, I say we take a step back and chill for a while.