Wei Liu

FSP-Diff effectively eliminates content drift in image super-resolution, resulting in HQ images that retain both visual detail and semantic integrity.

Pixel-level restoration can dramatically improve image quality by addressing specific degradation types and severities, outperforming traditional uniform methods.

CoSA achieves nearly 5× faster attention computation without sacrificing accuracy, revolutionizing long-context processing in LLMs.

UNIFUSION achieves unprecedented performance in generative tasks by seamlessly adapting autoregressive models to uniform-noise diffusion, outperforming all evaluated models on key metrics.