The AI is generating text like \\\\u5DF2\\\\u5BA1\\\\u6838
, this is actually supposed to be Chinese characters. I guess this might be solvable with better prompting, wondering there’s something Crowdin can do to prevent this.
Hi @trevorweng we will have to double-check with the team
Thanks @Ira , also just wanted to noted that this is not 100% reproducible, but when I did a mass translation it did happen on a small percent of strings.
Hi @trevorweng ! can you please try switching to gpt-4o model?
1 Like