Kosta Derpanis
@csprofkgd
#CS Assoc Prof @YorkUniversity, #ComputerVision Scientist Samsung #AI, @VectorInst Faculty Affiliate, TPAMI AE, @ELLISforEurope Member #CVPR2025 Publicity Chair
ID: 3439990697
https://csprofkgd.github.io 25-08-2015 20:57:12
32,32K Tweet
62,62K Followers
201 Following
I always found it puzzling how language models learn so much from next-token prediction, while video models learn so little from next frame prediction. Maybe it's because LLMs are actually brain scanners in disguise. Idle musings in my new blog post: sergeylevine.substack.com/p/language-mod…
Slides from my talk on "Language as a Visual Format" at the Visual Generative Modeling workshop at CVPR (mostly derived from slides made by Hyojin Bahng and Caroline 🐳): dropbox.com/scl/fi/5ok46te…
Just touched down from the air back to solid ground😄 We're honored to receive the Best Paper Award this year, while I want to give a special shout-out to the authors of the brilliant paper Dust3R: Shuzhe Wang (Shuzhe Wang), Vincent Leroy (Vincent Leroy), Yohann Cabon, Boris
![1991: first neural network distillation [1-3]. I called it "collapsing," back then, not “distilling."
References
[1] J. Schmidhuber (1991). Neural sequence chunkers. Tech Report FKI-148-91, Tech Univ. Munich. Sec. 3.2.2. & Sec. 4 are about "collapsing" or "distilling" or](https://pbs.twimg.com/media/GtkrMXZaIAAJ5g-.png "Jürgen Schmidhuber (@schmidhuberai) on Twitter photo 1991: first neural network distillation [1-3]. I called it \"collapsing,\" back then, not “distilling.\"
References
[1] J. Schmidhuber (1991). Neural sequence chunkers. Tech Report FKI-148-91, Tech Univ. Munich. Sec. 3.2.2. & Sec. 4 are about \"collapsing\" or \"distilling\" or")
