Oumar Kaba

Ph.D. Student at @Mila_Quebec / @mcgillu • ML, Physics, Math • Science Comms Enthusiast. Former intern @MSFTResearch AI4Science

[Oumar Kaba @sekoumarkaba]

@mjzellinger @MillionInt That's not quite right. Clearest example is AI becoming much better than humans at Go. It didn't just learn from human data, it also learned by playing against itself and discovering new strats. The same is done with other types of problems, e.g. math.

[Oumar Kaba @sekoumarkaba]

@KushaSareen @rish2k1 @agarwl_ @Devvrit_Khatri @LakshyAAAgrawal @inderjit_ml @profjoeyg @KurtKeutzer Amazing work, congrats @KushaSareen !

[Rishabh Agarwal @agarwl_]

I gave a talk at ICLR 2026 about how we are scaling RL on frontier LLMs with 1T+ parameters, on experimental data from our physical lab at Periodic!

Here's a rough recording of the talk:

[Montreal Expos @Montreal_Expos]

The Montreal Expos are exiting the baseball space. During Q2 and Q3 2026, we will transition to acquiring high-performance GPU assets. This is all part of our long-term vision to become a fully integrated GPU-as-a-Service (GPUaaS) and AI-native cloud solutions provider.

[Oumar Kaba @sekoumarkaba]

@Michael_D_Moor @francoisfleuret I think the surprising thing is that you can’t represent (and not approximate) all these functions with finite (and not infinite) compositions of these functions.

[Andrej Karpathy @karpathy]

- Drafted a blog post
- Used an LLM to meticulously improve the argument over 4 hours.
- Wow, feeling great, it’s so convincing!
- Fun idea let’s ask it to argue the opposite.
- LLM demolishes the entire argument and convinces me that the opposite is in fact true.
- lol

The LLMs may elicit an opinion when asked but are extremely competent in arguing almost any direction. This is actually super useful as a tool for forming your own opinions, just make sure to ask different directions and be careful with the sycophancy.

[Julian Suk @sukjulian]

One month left 'til @iclr_conf, about time to launch our ✨@GRaM_org_ Competititon✨ The theme is geometry x AI4science with a dataset kindly provided by @BeyondMathLtd .

Deadline: April 22, 2026 (AoE)

🔗 https://t.co/S5KaCpGtcU

[Sara Hooker @sarahookr]

this is honestly why the most important measure of progress in building true intelligence is our ability to eliminate prompt engineering.

[Oumar Kaba @sekoumarkaba]

@zhuci19 Way to go! An interesting fact about canonicalization is that it’s probably more expressive than equivariance for distributions. That’s because of data symmetry breaking. Another solution is relaxed equivariance. Theorems A.1 and A.2 of our paper here: https://t.co/1uvIJrryew

[Sharvaree Vadgama @SharvVadgama]

We have updated the submission deadline for GRaM workshop @iclr_conf.

Submit your works on everything geometry + generative modeling to GRaM workshop (@GRaM_org_ ) at #ICLR2026.

🗓️ Rio | April 26-27
⏰Deadline: Feb 5th AOE
Submission site: https://t.co/65aK41a6t8

https://t.co/u7qw94Q9jI

[Oumar Kaba @sekoumarkaba]

@Sylvestre_II @BibiFock_ @ToineWotan31514 @MonsieurPhi Si vous étiez confiant dans votre propos, vous ne vous sentirez pas le besoin d’insulter vos interlocuteurs et de parler de leur QI. Bonne journée à vous.

[Oumar Kaba @sekoumarkaba]

@Sylvestre_II @BibiFock_ @ToineWotan31514 @MonsieurPhi Tu n’as juste pas compris le but de ce montage. Le point est que le fait qu’il se répète constamment est en contradiction totale avec son argument. Le montage est excellent.

[Oumar Kaba @sekoumarkaba]

@EBerrier38 @A_Moatti @MonsieurPhi Écoutez bien ce qu’il dit dans les extraits. Qu’il ne sait pas ce qu’il va dire, que c’est «indéterministe», alors qu’il répète 10 fois la même chose mot pour mot. L’ironie est incroyable et mérite d’être soulignée.

[Oumar Kaba @sekoumarkaba]

Excited to announce the second edition of the GRaM Workshop at #ICLR26 🚀
If you work on geometry-grounded representations or generative modeling, we’d love to see your submissions!

[GRaM Workshop at ICLR 2026 @GRaM_org_]

📢The second edition of ✨GRaM workshop✨ is here this time at #ICLR26.

🌟Submit your exciting works in Geometry-grounded representations.

We welcome submissions in multiple tracks i.e.
📄 Proceedings
📝extended abstract
👩‍🏫Tutorial/blogpost
as well as an exciting challenge!

[Oumar Kaba @sekoumarkaba]

The GRaM workshop is coming back at #ICLR2016!

Very much looking forward to exciting discussions on geometry, representation learning and generative modeling!

[GRaM Workshop at ICLR 2026 @GRaM_org_]

📢 GRaM workshop is back!! This time at #ICLR2026
🌟Excited to be in Brazil 🇧🇷!

More information coming soon!

@iclr_conf @GRaM_org_

[Oumar Kaba @sekoumarkaba]

This work is a joint effort with my incredible collaborators @KushaSareen @dnllvy and Siamak!
Check out the paper on arxiv: https://t.co/8laCAApYE0
And come say hello at our poster on Friday afternoon!

[Oumar Kaba @sekoumarkaba]

Prediction of physical configurations is almost always trained with proxy losses like MSE and cross-entropy.
But physics already gives us the right objective: (free-)energy minimization.
In our NeurIPS paper, we show that using approximate energies as losses can drastically boost performance at no extra cost 👇🏾
🧵 1/7

[Oumar Kaba @sekoumarkaba]

We have tested this in practice on training diffusion models to generate molecules.
We find that training models with energy loss leads to faster convergence, better optima. It is also more data efficient and has greater benefits than equivariant models at lower cost.
In principle, the resulting score estimator can be biased at high times.
But in practice this yielded significantly higher molecular stabilities at essentially no cost.