推荐阅读！
Generative Representation Learning for Scalable Latent Tokenization
Recommended read on generative representation learning for scalable latent tokenization
AISummary
Dongxi NLP recommends a paper titled "Generative Representation Learning for Scalable Latent Tokenization." The post links to a thread, paper, and blog post, and quotes the authors' claim that the approach works where DINO and LeJEPA collapse.
Post on XView on X
Dongxi NLPVerified on X
@dongxi_nlp
1/ The tokenizer is a new scaling axis! We introduce an approach for diffusable representation learning that's principled, scalable, and works when DINO/LeJEPA collapse. We get joint tokenizer+diffusion scaling laws and it's compute-optimal vs raw diffusion w/ @drfeifei @GordonWetzstein @jcjohnss @ericryanchan @vincesitzmann @KyleSargentAI Thread, Paper, and Blogpost below
Source: Dongxi NLP · x.comPublished · added here