GPT4-o Generating lots of newlines characters instead of giving the actual results

API Bugs
1

Hi.

I’ve recently switched to gpt-4o in our services and we are having odd issues. Previously I’ve used gpt4-turbo-2024-04-09 and gpt-4-preview-1105 and all other versions. But This problem is very new.

This is actually the model result:


"\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n \n\n\n

I see this problem across all my prompts not just one. First I thought It was langchain’s fault. Then, I could reproduce this problem on colab with Pure OpenAI.

I wonder if anyone else has seen this problem and what are the possible solutions?

1 Like
2

I found that this only happens when using response_format={"type": "json_object" }. This never happens when response_format={"type":"text"}.

So, I can disable json_object for now. But this should be considered.

I wasted hours of work to solve this problem.

3

Hello

It happened to me too. I share this regarding the JSON mode, I don’t know if it makes a difference for you to also indicate in the prompt that you need a JSON format :

“Important Notes:
Explicit JSON Instructions: When using JSON mode, explicitly instruct the model to output JSON in your prompts. Without this, the output may consist of endless whitespace or appear ‘stuck’.”

Thank you and have a nice day

4

We have written lots of lines to tell the model that we need json.

I can not emphasize more. We first tell the model that the output needs to be json. Then we emphasize two more times that we need only json. Nothing else.

It indeed returns json most of the times but inside the json items, it is all \n.

1 Like
5

Hey everyone,

I experienced this issue multiple times as well, but I think I finally figured out the root cause:

It turns out that having a blank line at the beginning of the prompt was causing the problem. Once I removed that initial blank line, everything started working perfectly, and I haven’t encountered excessive newlines since.

I hope this helps others facing the same challenge!