Xiang Li

Image Tokenizer Needs Post-Training

On the Diversity of Synthetic Data and its Impact on Training Large Language Models

Masked Autoencoders Are Effective Tokenizers for Diffusion Models

Scalable Benchmarking and Robust Learning for Noise-Free Ego-Motion and 3D Reconstruction from Noisy Video

ImageFolder: Autoregressive Image Generation with Folded Tokens

On the Diversity of Synthetic Data and its Impact on Training Large Language Models

ControlVAR: Exploring Controllable Visual Autoregressive Modeling