One of my users is getting this strange behavior when using the AI Helper. They are selecting a single word and selecting the “Translate to Polish” option. The user’s interface is Polish, and the error logs don’t show anything.
What is the word? What is the LLM?
For example “vigilance” and “door jimmies.” The model being used is Gemini 2.0 Flash. As far as I can tell for now it seems to have been happening randomly
2 Likes
Is this still the case?
We are working on moving to structured outputs at the moment, it should resolve issues like this.
1 Like
I totally forgot about this to be honest. I have since moved to GPT-4.1-mini but I just tested it again for good measure and this issue is no longer occurring.
1 Like