Inaccurate architecture description in README: encoder-decoder vs. decoder-only

[shaoyoubo]

I noticed an inaccuracy in the model description between the README and the Technical Report.

README: mentions "...unified encoder-decoder architecture..."
Technical Report: states "...adopts a decoder-only vision–language architecture following the design principles of Qwen3-VL."

To maintain technical accuracy and consistency with Qwen3-VL, it would be better to update the README to reflect the decoder-only nature of the model.

[lixin4ever]

Hi, thank you for pointing out this issue. I think for modern VLMs, calling them "decoder-only" (ViT + projection module as internal image embedding layer) and "encoder-decoder" (ViT as seperate encoder to patchify image inputs) are both fine and technically correct. But your suggestion is great 👍👍, we should keep the model descriptions more clear and consistent between the tech report and the README, will update soon.