Guillaume Bellec

My initial point was about the moving frontier of knowledge in G2.5 and G3. Locating correctly areas was mind blowing to me.

I got fooled by the primate Brian with a mouse olfactory bulb. I cannot hide my poor knowledge of mouse anatomy. But I will continue to work in comp neuro if you allow me

It's not good. I think we agree! Gemini 2.5 and 3 cannot draw a mouse brain.

To show this funny "primate brain bias" in Gemini. I copy pasted your message as a correction prompt (first image). Gyri disappeared which is good. And then i tried to insist to put the cerebellum behind and not below.

It still likes primate brain too much apparently.

Exactly, Gemini loves to draw primate brains. It was ridiculous in G2.5, and cortices were all over the place.

I thought it was corrected in 3. Apparently not. Good there is a long way to fool scientists with good anatomical knowledge.

Imo, area locations were consistent with this image, isn't it?

Tried to correct the big flaws detected by @ctestard.bsky.social and @ackurth.bsky.social .

Well clearly Gemini 3 is still strongly attracted to draw human or primate brains. The teeth look better though

Thanks 🙏! Indeed, I was too convinced by the olfactory bulb but many things are missing. Good to see that the AI models have a long way to go.

My knowledge of mouse anatomy, is quite limited as you can see. The location of the cortices looked good to me. Any opinion about that?

Good catch!

Should probably look more like this ideally:
fr.wikipedia.org/wiki/Fichier...

Now edit capability. Also nailed locations of M2 and S2 (as far as I can fact check myself).

Doing figures will never be the same.

This was the prompt:
"For a scientific paper I want to draw the head of a mouse with the mouse brain visible and dots locating the primary somatosensory S1 the primary visual cortex V1 and the primary motor cortex M1"

I am impressed by the improvement with Gemini 3 and nano 🍌.

I used nano banana to make scientific figures. With 2.5 it was repetitively putting a human brain inside the 🐭 head.

Now it draws an accurate mouse brain anatomy even seems to locate correctly the cortical areas. Big jump imo

Biologically informed cortical models predict optogenetic perturbations A recurrent neural network fitted to large electrophysiological datasets may help us understand the chain of cortical information transmission. In particular, successful network reconstruction methods...

This year, perturbation testing is becoming a thing: Model fitted to neural activity can predict how stimulation changes the animal's behavior

By @mschrimpf.bsky.social & @hannesmehrer.bsky.social :
arxiv.org/abs/2510.03684

Our work (being published at elife)
www.biorxiv.org/content/10.1...

Jaxley: differentiable simulation enables large-scale training of detailed biophysical models of neural dynamics - Nature Methods Jaxley is a versatile platform for biophysical modeling in neuroscience. It allows efficiently simulating large-scale biophysical models on CPUs, GPUs and TPUs. Model parameters can be optimized with ...

Our work on training biophysical models with Jaxley is now out in @natmethods.nature.com. Led by @deismic.bsky.social, with @philipp.hertie.ai, @ppjgoncalves.bsky.social & @jakhmack.bsky.social et al.

Paper: www.nature.com/articles/s41...

Whoopsy

AI and Neuroscience | IVADO

I’m looking for interns to join our lab for a project on foundation models in neuroscience.

Funded by @ivado.bsky.social and in collaboration with the IVADO regroupement 1 (AI and Neuroscience: ivado.ca/en/regroupem...).

Interested? See the details in the comments. (1/3)

🧠🤖

Introducing CorText: a framework that fuses brain data directly into a large language model, allowing for interactive neural readout using natural language.

tl;dr: you can now chat with a brain scan 🧠💬

1/n

Reminder this is happening this Wed/Thu. Free spiking neural network conference - registration required (see below).

This was a fabulous, once in a lifetime colloquium -- and now the videos are available in high quality on the College de France web site @college-de-france.fr
www.college-de-france.fr/fr/agenda/co...
With talks by Edvard Moser, Nancy Kanwisher, Liz Spelke, Manuela Piazza, Luca Bonatti and more!

New paper titled "Tracing the Representation Geometry of Language Models from Pretraining to Post-training" by Melody Z Li, Kumar K Agrawal, Arna Ghosh, Komal K Teru, Adam Santoro, Guillaume Lajoie, Blake A Richards.

LLMs are trained to compress data by mapping sequences to high-dim representations!
How does the complexity of this mapping change across LLM training? How does it relate to the model’s capabilities? 🤔
Announcing our #NeurIPS2025 📄 that dives into this.

🧵below
#AIResearch #MachineLearning #LLM

Assistant/associate Professor of AI-assisted Biophysics at the Ecole polytechnique fédérale de Lausanne (EPFL) and Group Leader at the Paul Scherrer Institute (PSI) The School of Basic Sciences (Physics, Chemistry and Mathematics) at EPFL and the Paul Scherrer Institute (PSI) jointly seek to appoint a Tenure Track Assistant or tenured Associate Professor in AI-As...

#EPFL has an open faculty position - tenure track or tenured!! - in AI driven biophysics. Deadline is Nov 15th. Please spread the word!

www.epfl.ch/about/workin...

Tbf, people did request SE or STD in RL where things are less stable.

But now for big LLMs it is not uncommon (although not ideal) to manually restart to intermediate checkpoints.

If the model you publish is strong and people reproduce your work within 3 months. Your work is very important.

Nop, SE or STD over models do not really capture this.

Depending on learning rate or net size. You can have a model init at 50% acc and another 4 at 98.5%. Making strong STD and SE and the diff with CNN at 99% insignificant. Yet everybody knows it is reproducible like the sunset and the sunrise.

Ok. An example: MNIST

Every ML researcher knows 98.5% means maxing up an MLP and 99% decent CNN. You know the margin is reproducible f you worked with this.

But reporting STD will not easily capture this margin. Not shuffle the training/test split. Not, STD over models.

Indeed, bad example 😅

I agree that inferential versus descriptive is important. But how does STD versus SE address the issue?

In ML for Neuro, the sqrt(n) factor is obscure. n may mean: num of animals, num of models, num of CV shuffles...

The error bar will not represent well reproducibility either way.

Maybe Neuroscience has been over obsessed with statistical tests. Wasting time on rebranding means, std and linear regression behind complicated tests.

Just to find out in the next batch of papers that those statistical test are easy to cheat, or having significant irrelevant results.

Or maybe neuro-statisticians should learn from ML.

1st: Yes, most ML researcher had strong training in applied math including stats and proba.

2nd: When the reproducibility that matters is clear. The ML field quickly agree on a simple metric. So no need to make things complicated.

Okay, sorry I had misunderstood from the context of the our last discussion, cool.

End-to-end back-prop works wonderfully well. So good to study this case too. Cool work

Cool. Does the modulation completely replace the data augmentation via image transformation? (Does your self sup still works up to few percents of acc if you disable data augmentation completely?)

Are all the gradients computed layer wise for weight update?

New #NeuroAI preprint on #ContinualLearning!

Continual learning methods struggle in mostly unsupervised environments with sparse labels (e.g. parents telling their child the object is an 'apple').
We propose that in the cortex, predictive coding of high-level top-down modulations solves this! (1/6)

Cool work !!
I only had a quick read but congrats 🎉

The idea of modeling top-down plasticity as rare labeled 1-class-vs-all classifiers is very elegant.

The ff parameters W are learned via constrastive learning? But I did not understand what the modulation does during unlabeled sample?