[KongchangAI]
ConceptPixel-Space Diffusion

Pixel-Space Diffusion

A diffusion model approach that performs denoising directly in the raw high-resolution pixel space. It avoids the lossy compression of a tokenizer but faces difficulties in scaling.

Timeline (last 90 days)

Oct 14

Existing pixel-space models (such as JiT) adopt a dense scaling strategy, where each increase in model capacity increases inference computation by the same amount, making it difficult to improve generation quality by scaling up parameters under a fixed inference budget.

Unverified50%

All Facts (1)

Unverified

Existing pixel-space models (such as JiT) adopt a dense scaling strategy, where each increase in model capacity increases inference computation by the same amount, making it difficult to improve generation quality by scaling up parameters under a fixed inference budget.

50%

Source Articles

PixelMoE: LLM-Style MoE Sparse Scaling for Pixel-Space Diffusion ModelsOct 14