Skip navigation
We present MMaDA, first diffusion that unifies text reasoning, multimodal understanding, and image…
0:00 / 0:25

We present MMaDA, first diffusion that unifies text reasoning, multimodal understanding, and image…

Ling Yang

@LingYang_PU5.4K followers

Visit on X
601 likesDownloadOriginal post

87,734 views

We present MMaDA, first diffusion that unifies text reasoning, multimodal understanding, and image generation through Mixed Long-CoT, and unified RL - UniGRPO. 📚 Paper: 💻 Code: 📦 Model:

11 Comments

Comments from the original post on X (Twitter)

  • Ling Yang's profile picture

    @LingYang_PU

    We also provide an online demo: Welcome to use and provide suggestions!

    16 likes

  • FemMiraMancer's profile picture

    @miratechgeek

    MMaDA is revolutionizing the multimodal game! I'm eager to see how its unified architecture and RL algorithms elevate text-to-image generation. This could be a real milestone in AI innovation. 🚀

    1 like

  • Stellan Haglund's profile picture

    @Stellanhaglund

    That looks pretty sequential, isn’t the upside with diffusion to predict all the tokens every step?

    1 like

  • Ling Yang's profile picture

    @LingYang_PU

    We use unified diffusion modeling for training and provide flexible sampling strategies (Semi-AR or Non-AR) for better performance-speed trade-off

    2 likes

  • Ruben Tous's profile picture

    @rubentous1

    MMaDA is all you need.

    1 like

  • Mr. John Smith's profile picture

    @MrJ0hnSmith

    Pls stop, we can’t keep up

    1 like

  • FluffyCat's profile picture

    @fluffyfatcat1

    CoT training data makes sense for autoregressive models, but what's the intuition for using it with a diffusion model?

    1 like

  • Hyperstack's profile picture

    @Hyperstackcloud

    Sounds like a game changer for bridging text and visuals 👏

  • Bali as a Colony of Jakarta's profile picture

    @pakar_ai

    not good prompt following.

  • Sam Woods's profile picture

    @samwoods

    That's a huge step towards truly unified multimodal AI

  • MangoMagic™️'s profile picture

    @somanymangoes

    Curiosity piqued: does the model argue back? Asking for a friend. #AIorNot #DeepThoughtsNoDeepSleep #CoderInDisguise