Jacob Bamberger

Looking for geometry where it shouldn’t be. PhD student @UniofOxford. Interested in Geometric Deep Learning

[Jacob Bamberger @jacobbamb]

Flow Matching models often struggle to balance memorization and generalization. 😱
We set out to fix this — by using the geometry of the data manifold.

Introducing Carré du Champ Flow Matching (CDCFM)🧑‍🎨🥖 — improving generalization without sacrificing sample quality.

[Xingyue Huang @hxyscott]

Now that the NeurIPS deadline is over, consider submitting your work to the ICML 2026 Workshop on Graph Foundation Models (GFM)!

We would love to see more exciting work from the community!

🗓️ Submission Deadline: May 8, 2026 AoE

[Michael Bronstein @mmbronstein]

Multiple postdoc positions in geometric ML and generative modeling are available at Oxford in collaboration with Aithyra and Imperial @bose_joey

https://t.co/JXt5O4509w

[Jacob Bamberger @jacobbamb]

It was a pleasure to participate in this great workshop!

[GRaM Workshop at ICLR 2026 @GRaM_org_]

Wonderful in-person contributed talks from @Riccardo_Ali_, @LauraPGomez_ , and @jacobbamb on approximate equivariance, choosing optimal optimizer steps, and learning the Riemannian geometry of data!

[Lars Holdijk @HoldijkLars]

Did @wellingmax’s keynote at #iclr2026 leave you wanting more stochastic thermodynamics and generative AI?

Come join Max and me at poster #1023 in Pavilion 3 in the afternoon to chat about how we can learn escorted protocols for multistate free-energy estimation!

[Jacob Bamberger @jacobbamb]

The second is Riemannian Metric Matching, an Oral at the GRAM workshop, on scaling tools from diffusion geometry with neural networks!

The oral is at 10:10 on April 26th and then poster at 16:05.

Paper: https://t.co/4ttPDgRt2o

[Jacob Bamberger @jacobbamb]

I’m at #ICLR2026 and will be presenting two papers on geometry and generative models.

[Jacob Bamberger @jacobbamb]

CDCFM at the main conference on using geometry to regularize flow matching
April 23rd at 10:30 P3 #701

Paper: https://t.co/3FusxOPziQ
Code: https://t.co/98YjdbH3Ko

[Floor Eijkelboom @FEijkelboom]

Flow-LLM Blogpost :D https://t.co/0HiyNPJHsk

In the last few weeks, a bunch of work on flows for language came out 🌊

That is exciting, because it makes truly parallel text generation feel real: generation where models can keep refining the whole response during inference, instead of committing token by token.

I wrote an intuitive and animated introduction to the area — why autoregression has a structural ceiling, why discrete diffusion only partly escapes it, and why flows may be the first genuinely parallel alternative.

Here's an overview of the key parts of the blog - and let's chat at #ICLR2026 :)

[Michael Bronstein @mmbronstein]

Come and see our papers at #ICLR2026 in Rio!

[Hugh Blayney @HughBlayney]

1/8 Taking a look inside looped language models! We've released a new preprint on looped LLMs, an exciting new direction for scaling test-time compute. 🔁 Thanks to wonderful collaborators @arroyo_alvr @johanobandoc @pcastr @AaronCourville @mmbronstein @epomqo 🧵

[Alex Tong @AlexanderTong7]

Join us at #AITHYRA. It's a fantastic research environment.

If you are considering applying for the AI/ML PI position and will be attending #ICLR2026, send me a message. I’d love to meet up and answer any questions you have!

[Jacob Bamberger @jacobbamb]

@cryptexis @iolo_jones @AGosztolai @duncdennis @mmbronstein @trekkinglemon You can find it here :)
https://t.co/98YjdbH3Ko

[Jacob Bamberger @jacobbamb]

Flow Matching models often struggle to balance memorization and generalization. 😱
We set out to fix this — by using the geometry of the data manifold.

Introducing Carré du Champ Flow Matching (CDCFM)🧑‍🎨🥖 — improving generalization without sacrificing sample quality.

[Jacob Bamberger @jacobbamb]

@ZizuResearch @iolo_jones @AGosztolai @duncdennis @mmbronstein @trekkinglemon Here it is!
https://t.co/98YjdbH3Ko

[Jacob Bamberger @jacobbamb]

Code is now public: https://t.co/98YjdbH3Ko

[Jacob Bamberger @jacobbamb]

Flow Matching models often struggle to balance memorization and generalization. 😱
We set out to fix this — by using the geometry of the data manifold.

Introducing Carré du Champ Flow Matching (CDCFM)🧑‍🎨🥖 — improving generalization without sacrificing sample quality.

[Ahmed Elhag @Ahmed_AI035]

Excited to finally share TransIP: a scalable and open-source Transformer model for force fields! TransIP learns symmetry in embedding space and doesn't require a pretrain-finetuning framework!

project led together w @arun_raja007 and many great collab.! Happy to be small part!

[Alex Tong @AlexanderTong7]

AITHYRA is recruiting an AI/ML Starting PI!

Come join us in Vienna! 🇦🇹 Couldn't ask for a better place to do AI/ML research.

Don't hesitate to reach out if you're thinking of applying and have questions!

Application Deadline: 30 April 2026

More info: https://t.co/aKA6aQ03fG

[Kieran Didi @DidiKieran]

📢 We’re launching Proteina-Complexa — and after the Jensen keynote mention, we definitely had to post this thread now ;)
Atomistic binder design with generative pretraining + test-time compute, plus large-scale wet-lab validation.
Project page: https://t.co/aT8Lz2VhSJ
🧵 1/n

[Jacob Bamberger @jacobbamb]

Still think persistent homology lacks a killer app?
Turns out TDA can steer diffusion model for drug design 🤯⬇️

[Alicja (Ala) Maksymiuk @AlaMaksymiuk]

🚨We present MacroGuide: the first model to generate arbitrary macrocycles in 3D 🎉

How? Topological guidance with persistent homology!🧵⬇️ [1/6]

With @duplessis_alex @mmbronstein @AlexanderTong7 @fjduarteg @ismaililkanc

[Lars Holdijk @HoldijkLars]

Struggling with minibatch noise in Stochastic Gradient Bayesian Inference? Want your chains to naturally run on stochastic hardware?

Introducing — Stochastic Gradient Lattice Random Walk!

New work from @NormalComputing in collab with @zierhjmensch, @adnarim066, @wellingmax, and the dream team at normal computing, @Sam_Duffield, @MaxAifer and @ColesThermoAI.

[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_