Michael Tschannen
261
posts
@mtschannen
Research Scientist
@GoogleDeepMind. Representation learning for multimodal understanding and generation. Personal account.
Zurich, Switzerland
Joined June 2012
Pinned
For the past years my research focus was on unifying models and training paradigms across modalities. Today I'm excited that we're releasing our latest model aligned with this theme: Gemma 4 12B, a dense encoder-free model which processes raw text, image, and audio inputs! 1/
For the past years my research focus was on unifying models and training paradigms across modalities. Today I'm excited that we're releasing our latest model aligned with this theme: Gemma 4 12B, a dense encoder-free model which processes raw text, image, and audio inputs! 1/
Have you ever wondered how to train an autoregressive generative transformer on text and raw pixels, without a pretrained visual tokenizer (e.g. VQ-VAE)? We have been pondering this during summer and developed a new model: JetFormer 🌊🤖 arxiv.org/abs/2411.19722 A thread 👇 1/
We are presenting JetFormer at ICLR this morning, poster #190. Stop by if you’re interested in unified multimodal architectures!
Have you ever wondered how to train an autoregressive generative transformer on text and raw pixels, without a pretrained visual tokenizer (e.g. VQ-VAE)? We have been pondering this during summer and developed a new model: JetFormer 🌊🤖 arxiv.org/abs/2411.19722 A thread 👇 1/
4o native image generation is confirmed to be some sort of autoregressive model. Maybe this is a good moment for AR skeptics to catch up on the recent literature on multimodal AR models.
