Show-o and Discrete-Diffusion Unified Models
Transfusion mixes continuous and discrete representations. Show-o (Xie et al., August 2024) goes the other way: text tokens use causal next-token prediction, image tokens use masked discrete diffusion in the spirit of MaskGIT. Both sit inside one transformer with a hybrid attention mask. The result unifies VQA, text-to-image, inpainting, and mixed-modality generation on one backbone, one tokenizer per modality, one loss formulation (next-token extended to masked prediction). This lesson walks the Show-o design — why masked discrete diffusion is a parallel, few-step image generator — and contrasts with Transfusion and Emu3. Explain masked discrete diffusion: the schedule that masks tokens uniformly then asks the transformer to recover them. Compare parallel image decoding (Show-o, MaskGIT) to autoregressive image decoding (Chameleon, Emu3) on speed and quality. Name the three tasks Show-o handles in one checkpoint: T2I, VQA, image inpainting. Pick a masking schedule (cosine, linear, truncated) and reason about its effect on sample quality. Transfusion's two-loss training works but has trickier dynamics — the continuous diffusion loss lives on a different numerical scale from the discrete NTP loss. Balancing loss weights is a hyperparameter search. The architecture is effective but complex. Show-o's answer: keep both modalities discrete (like Chameleon), but generate images in parallel via masked discrete diffusion instead of sequentially. The training objective becomes a single masked-token-prediction that generalizes next-token-prediction naturally.…
Show-o and Discrete-Diffusion Unified Models: Transfusion mixes continuous and discrete representations. Show-o (Xie et al., August 2024) goes the other way:…
This free lesson is part of the AI Engineering from Scratch curriculum. Read the full explanation, run the lesson code, and verify the result in the interactive reader or from the repository source.
Browse the complete course catalog or open this lesson on GitHub.