GPT-4 Abrubtly stops any output in Hebrew

I’m trying to build a resume generation workflow out of candidate information. However GPT stops halfway into any Hebrew output. Here is an example of me trying to produce a CV for Nikola Tesla based on his Wikipedia introduction section.

Not changing the Hebrew input (still Nikola Teslas Wikipedia introduction in Hebrew), GPT prints it’s response to proper completion upon me asking it to print a response in English at the prompt. (I’m a new user so unfortunately I cannot show an additional screenshot)

Why does it stop midway? And why is the Hebrew CV so poor if the English CV is good? My only guess is that maybe GPT4 resorts to producing a CV for Nikola Tesla based on prior knowledge.

Your max tokens is set to 256. Try increasing this.

1 Like