The development of unified multimodal large language models (MLLMs) is fundamentally challenged by the granularity gap between visual understanding and generation: understanding requires high-level ...

"Given a text string (e.g., `\"tiktoken is great!\"`) and an encoding (e.g., `\"cl100k_base\"`), a tokenizer can split the text string into a list of tokens (e.g ...

Abstract: Training deep neural networks (DNNs) with altered data, known as adversarial training, is essential for improving their robustness. A significant challenge emerges as the robustness ...

This project presents UniFlow (Unified Pixel Flow Tokenizer), a unified continuous visual tokenizer that eliminates the conflict between semantic understanding and high-fidelity reconstruction. By ...