Seed-llama

📌 核心贡献

这篇论文的核心贡献在于通过引入“SEED Tokenizer”,成功扩展了LLaMA模型,使其能够同时进行图像理解(“SEE”)和图像生成(“Draw”)任务。这一方法创新旨在弥合现有多模态大型语言模型在理解与生成间的差距,并推动模型在开放世界中展现出涌现能力,向通用人工智能(AGI)迈进。

📖 摘要

In the past several years, Howe curves have been studied actively in the field of algebraic curves over fields of positive characteristic. Here, a Howe curve is defined as the desingularization of the fiber product over a projective line of two hyperelliptic curves. In this paper, we construct an explicit plane sextic model for non-hyperelliptic Howe curves of genus five. We also determine singularities of our sextic model.

🔍 关键信息

字段内容
机构ARC Lab
发表2023-10-02
分类cs.CV
链接arXiv

📝 我的笔记

🔗 相关论文

基于/改进自: VQGAN 同方向: Chameleon