Developers need to convert Chinese prompts into English that is better understood by LLMs to reduce output bias.
Currently developers may manually translate or use general translation tools, which may not be precise or optimized for prompts.
Differences between Chinese and English expression may cause model misunderstanding, and input language may affect regional judgment.