Whether support multi-language compression

[FuJiaJie123]

I am wondering whether LLMingua-2 can be used to compress Chinese prompt, or only for english?

[aniruddhaadak80]

From my point of view, the xlm-roberta base strongly suggests multilingual capability is possible, but the practical question is not only whether it runs on Chinese input. It is whether the compression preserves the right information density and semantic faithfulness across languages with different tokenization behavior.

I would treat multilingual support as something to verify empirically per language and domain rather than assume from the encoder name alone.