Tag
PixelDense reframes dense prediction tasks as representation alignment targets for pixel-space diffusion, routing semantic (DINOv2, SAM2) and geometric (DepthAnythingv2, Metric3Dv2) teachers through separate projection streams with an orthogonality penalty. The method improves GenEval, DPG-Bench, and HPSv2.1 scores on PixelGen and DeCo, achieves up to 53.1% PQ gain in partial-noise reconstruction, and trains 1.23x faster than baseline.
This paper presents the first systematic study of adversarial post-training for pixel diffusion models, showing it effectively corrects high-frequency content deficiencies to improve distribution fidelity, coverage, and perceptual quality without altering model architecture.
Introduces Representation Forcing (RF), a technique that enables unified multimodal models to perform both perception and generation end-to-end without external VAE latent spaces, matching state-of-the-art VAE-based models in image generation while improving understanding.
NVIDIA introduces PiD, a Pixel Diffusion Decoder that replaces traditional VAE/RAE decoders in latent diffusion models, enabling fast, high-resolution decoding with up to 6× speedup and improved visual fidelity.
NVIDIA Spatial Intelligence Lab proposes PiD, which redesigns the decoding stage of latent diffusion models as a conditional pixel diffusion process, unifying decoding and upsampling to achieve low-latency, high-resolution decoding.
PiD introduces a pixel diffusion decoder that reformulates latent decoding as conditional pixel diffusion, enabling fast and high-quality image synthesis at high resolutions with reduced computational requirements. It decodes latents into 4x or 8x upscaled images in under a second on consumer hardware.
L2P proposes an efficient transfer paradigm that leverages pre-trained latent diffusion models to build pixel-space diffusion models, enabling high-quality generation with minimal computational overhead and data requirements, and supporting native 4K resolution.