Fran Castro

Maths & CS @CanalUGR

[Fran Castro @franblueee]

So happy to share that Conditional Diffusion Sampling has been accepted to #ICML 2026!! 🎉

I loved working on this: it has a great mix of theory + experiments, and I learned a lot while working on it! ✨

[MatLab crashes @memecrashes]

outdated not even 3hs later by a human, not everything is lost

https://t.co/TUZpvQ8cn4

[Maria Esteban @Maria__Esteban]

🏎️Drift in the right direction🏎️

Introducing kernel-gradient drifting models: a reformulation of drifting models where the kernel itself defines the direction of motion through its gradient.

📜Paper: https://t.co/1ILDAZj66Y
💾Notebook: https://t.co/9tzqaXAyz7

[Eric Jang @ericjang11]

For the last few months I've been working on a from-scratch implementation of AlphaGo, a 2016 AI breakthrough that inspired me to get into deep learning. My casual understanding of AlphaGo was "search-augmented deep neural networks trained with self-play", but I wanted to go deeper and understand it by creating it.

Frontier deep learning research has always been expensive, but any given capability gets cheaper very quickly. In 2026, you no longer need DeepMind's resources to train a strong Go AI - you can vibe code all of it yourself for just a few thousand dollars of rented compute.

It was a huge honor to be invited to teach this with @dwarkesh_sp on @dwarkeshpodcast

I am an AlphaGo & Go apprentice, not a master, so all factual errors in the podcast are mine.

Web version of tutorial: https://t.co/Xkf9VsgtuT
Code: https://t.co/rWKOwclPDg
Play the go bot here: https://t.co/aVglJXldVX

[Andreas Bergmeister @AndBergmeister]

Pretraining diffusion/flow-matching is simple and scales: noise a clean sample, regress against a closed-form target.

RL post-training can be made just as simple.

Reinforce Adjoint Matching (RAM) matches Flow-GRPO on SD3.5M in 50× fewer steps. Highest reward, no reward hacking.

[Fran Castro @franblueee]

Hopefully you will like it as much as I do. It is available at: https://t.co/O8rLdwsmaz.

w/ Saifuddin Syed, Pablo Morales Álvarez, Daniel Hernández Lobato, Rafael Molina, and @jmhernandez233

[Fran Castro @franblueee]

So happy to share that Conditional Diffusion Sampling has been accepted to #ICML 2026!! 🎉

I loved working on this: it has a great mix of theory + experiments, and I learned a lot while working on it! ✨

[Fran Castro @franblueee]

We evaluate the approach across a wide range of settings, from synthetic Gaussian mixtures to particle systems, molecular systems, and Bayesian posterior inference. 🧪🧬🔬 We obtain gains in performance under reduced computational budgets!

[Fran Castro @franblueee]

@JulioGonzalo1 @ANECAinfo Lo cual repercute en el talento que producimos. Ejemplo: @CanalUGR (top en IA según los rankings), las publicaciones en congresos A+ son completamente ignoradas en las tesis doctorales. Curiosamente, estas publicaciones son las que se valoran para trabajar fuera de España.

[Mikel Artetxe @artetxem]

El sinsentido de @ANECAinfo y del sistema académico español a la hora de valorar los congresos de IA es incomprensible y tremendamente dañino. Quien defiende estos criterios es, en el mejor de los casos, un ignorante completo de los estándares de publicación de nuestro campo, cuando no miembro de una mafia que se beneficia del sistema.

El Premio Nobel de Física y Premio Turing @geoffreyhinton, considerado el padre de la IA moderna, ha publicado 20 artículos en congresos A* y solamente 3 en revistas en los últimos 10 años. Según los criterios de ANECA, debe ser un investigador mediocre que todavía no se ha enterado dónde hay que publicar.

[Julio Gonzalo @JulioGonzalo1]

La ANECA valora los congresos CORE A+ -- los lugares más competitivos y de más prestigio donde publicar en IA -- como a las revistas del ultimo tercil en JCR. ¿Alguien en @ANECAinfo puede explicarlo? ¿Se dan cuenta de hasta qué punto fomentan nuestra irrelevancia científica?

[Nikita Morozov @nvimorozov]

Got asked in a review for my ICML paper whether "there are realistic tasks where one needs to sample from a probability distribution given by its unnormalized density" (rephrased for anonymity). Are we cooked?

[Alejandro García @algarciacast]

New paper accepted at @GRaM_org_ !!! 🥳🤩

Here is a small infomercial about it 🤠

If you are interested, you can find extra info in the next tweet 👇

[Ignacio Peis @IgnacioPeis]

Incredibly honored to share that I have been selected by #laCaixaFoundation as Postdoctoral Junior Leader Fellow (Incoming). Over the next years, I will lead a research line on flow-based generative modelling and meta-learning at @IMAG_UGR, @CanalUGR .

[MICCAI Society @MICCAI_Society]

👏 Thank you Francisco M Castro-Macías for presenting torchmil in the most recent @RMiccai Tutorial session.
#ICYMI: You can watch the recording here:
🔗 https://t.co/KuryIuVWkn
@MiccaiStudents @WomenInMICCAI #MICCAI

[MICCAI Society @MICCAI_Society]

⭐ Join @RMiccai at the next tutorial titled "Multiple Instance Learning for Histopathology: An Introduction to torchmil" with Francisco M. Castro-Macías, Creator, Lead-developer
🗓️ March 3 - 16:00 CET
Register:
🔗 https://t.co/626OrO5L4D #MICCAI @MiccaiStudents @WomenInMICCAI

[arXiv.org @arxiv]

#HBD to arXiv!🎈

On August 14, 1991, the very first paper was submitted to arXiv. That's 34 years of sharing research quickly, freely & openly!

Some baby pictures to show how far we've come . . . when we were just a computer under desk . . . & in our 1994 punk phase . . . 👶💾

[Fran Castro @franblueee]

@itsbautistam @IgnacioPeis

[LOGML Summer School @LogmlSchool]

And that's a wrap on the LOGML 2025 Summer School! 🚀

We’ve been quiet on here, but that’s because we were busy with 100 brilliant minds diving deep into geometry, ML, and social activities. 😉

[Universidad Granada @CanalUGR]

🔴 La #UGR manifiesta su desacuerdo ante los informes desfavorables de la @ACCUAndalucia para algunas titulaciones estratégicas en #Granada y #Melilla

ℹ️ Lee el Comunicado completo: https://t.co/bgcJrP3M74