Ioana

PhD Student @ University of Amsterdam

[Ioana @ioanasimioni]

@PetarV_93 @MrzSalehi @shawshank_v @cgmsnoek @y_m_asano Definitely! Lets catch up! I’m around for the workshops! ✨

[Ioana @ioanasimioni]

(1/n) What if you could teach a vision model about 3D structure without a single 3D label and just a tiny fraction of its pretraining compute? 👁️ That’s 3DPoV. ✨ Heading to #ICML2026 in Seoul 🇰🇷 to share it 👇

[Ioana @ioanasimioni]

@MrzSalehi @shawshank_v @cgmsnoek @y_m_asano (6/n)
Paper + project page + code below >
📄 https://t.co/UTanZm1SUz
🌐 https://t.co/Uzv19x81In
💻 https://t.co/WNdDrKMNAC

[Ioana @ioanasimioni]

(5/n) Come find me at ICML!
📍Tuesday 2PM -> Hall A, poster #1309

Happy to talk 3D, self-supervised learning, or vision transformers ✨
Huge thanks to my amazing co-authors @MrzSalehi @shawshank_v @cgmsnoek @y_m_asano 🙌

[Floor Eijkelboom @FEijkelboom]

Very excited about our work on finding the right drifting direction 🐎

We tackle a core open question in drifting: when does “no drift left” mean the model really matched the data?

Kernel-gradient drifting is the answer (with natural extensions to manifolds + discrete data)!

[Ioana @ioanasimioni]

Something is brewing 👀👀👀

[Floor Eijkelboom @FEijkelboom]

Discrete diffusion — but fast? ⚡️ Test-time inference — but for discrete data? 🧠

Categorical Flow Maps: continuous transport toward the simplex, turning discrete generation into a single-step problem. Built on Variational FM (CatFlow), we obtain (self-)distillation from scratch.

Language, molecules, and test-time steering — one framework.

Scaling to LLMs and foundation models next. Watch this space 👀

[Oscar Davis @osclsd]

You like discrete diffusion, but it's too slow? 🥀
You like test-time inference, but it's for continuous methods? 😩

We fixed it.

Introducing Categorical Flow Maps: continuously sample discrete data in a single step 🚀💫

How? 🧵⬇️

💪 Co-led with @FEijkelboom, @daan_roos_

[Ioana @ioanasimioni]

@N_Midavaine @ICCVConference Thanks a lot! ✨

[Ioana @ioanasimioni]

🚨 New paper accepted at @ICCVConference

As video foundation models gain momentum, we introduce MoSiC—a method that leverages motion as supervision to learn dense, temporally consistent representations.

Full details in the thread! ✨

[Shashank @shawshank_v]

New paper out - accepted at
@ICCVConference

We introduce MoSiC, a self-supervised learning framework that learns temporally consistent representations from video using motion cues.

Key idea: leverage long-range point tracks to enforce dense feature coherence across time.🧵

[Mohammad Reza Salehi @MrzSalehi]

🎉 NeCo is accepted to @iclr_conf #ICLR25
TL;DR: Self-supervised patch-level neighbor consistency boosts dense features of foundation models like DINOv2—in just 19 GPU hours.
📍 Tomorrow,10:00–12:30@Hall3+Hall2B#133
📄 Paper: https://t.co/er8Y47trVI

[Ioana @ioanasimioni]

@y_m_asano @ELLISforEurope Congratulations! 🎉

[Yuki @y_m_asano]

Wait, we can just ask? 🙋
Here's what we do at @FunAILab in Nuremberg: research on self-supervised learning from images and videos and building /improving (M)LLMs. @nvidia @LambdaAPI @CohereForAI :)? Happy to provide more info.

[Ioana @ioanasimioni]

@PetarV_93 @GoogleDeepMind Congratulations! 👏

[Mohammad Reza Salehi @MrzSalehi]

🚀 Excited to present SIGMA at @eccvconf ! 🎉 We upgrade VideoMAE with Sinkhorn-Knopp on patch-level embeddings, pushing reconstruction to more semantic features. With @mdorkenw.
Let’s connect at today's poster session at 4:30 PM, poster number 256, or send us a DM.

[Floor Eijkelboom @FEijkelboom]

SO happy to announce our work on Variational Flow Matching for Graph Generation has been accepted to 🐳 #NeurIPS2024 🐳!

Thanks again to the amazing @GrigoryBartosh , @canaesseth , @wellingmax , and @jwvdm ❤️

See you in 🇨🇦 for the next edition of @NeurIPSConf

[Yuki @y_m_asano]

First time organising a Tutorial with an amazing team and am very excited 🎉! The topic is learning from videos, which I think will be the new 'Big' paradigm for new vision foundation models. Come to learn, chat and discuss @eccvconf!

[Ioana @ioanasimioni]

Such a clear & accesibile blogpost! 👏
Great addition to the JAX open source space! ✨

[Max Zhdanov @maxxxzdn]

📰 blogpost:
https://t.co/9OWML2JPkX

🕹️ google colab:
https://t.co/vBxnTWhALp

I tried to make the blog post more accessible than the paper and added a lot of supporting visualizations.

Please check it out if you are curious about spacetime-equivariant CNNs 🚀