about alignment evaluation or prompt following study?

[BruceLeeeee]

Hi, thanks for the amazing work.
I have read your technical report about T2I part, but no found about alignment evaluation or prompt following study.

[gaopengpjlab]

Thanks for your suggestion. We will add more T2I evaluation soon.

[BruceLeeeee]

And I noticed your model also supports other languages (e.g. Chinese), although the image looks a little worse. I'd like to know if you used other languages text-image pairs in training?

[gaopengpjlab]

The prompts contain only English. However, after english-only training, the Lumina-T2X model developed emerging multilingual capabilities, including the ability to understand and generate Chinese, emoji, Japanese, and other languages.

[gaopengpjlab]

We will include the discuss of emerging multi-lingual ability into our report in the future.

[BruceLeeeee]

amazing!
it seems no discussion of multilingual capabilities when using large language models as text encoders in current research.

Looking forward to training script support (doesn't work properly right now)

[gaopengpjlab]

We will update the training script soon.