🌙 Lune Bellec

I find AI coding most useful to comment / suggest on what I do. Your disastrous experience with code generation matches mine. But as a side kick it's incredibly positive IMO.

🥁... we are SO happy to officially announce that registration is now OPEN for our OHBM Virtual Satellite Meeting, taking place September 10-12!

This has been a major goal of the SEA-SIG for a while now and we're so excited to show you what we've been working on!

🌱🌎✨🧠

four brain maps showing noise ceiling estimates in response to image presentation

New CNeuroMod-THINGS open-access fMRI dataset: 4 participants · ~4 000 images (720 categories) each shown 3× (12k trials per subject)· individual functional localizers & NSD-inspired QC . Preprint: arxiv.org/abs/2507.09024 Congrats Marie St-Laurent and @martinhebart.bsky.social !!

Precision functional mapping reveals less inter-individual variability in the child vs. adult human brain Human brain organization shares a common underlying structure, though recent studies have shown that features of this organization also differ significantly across individual adults. Understanding the...

1/11 Very excited to say that our preprint, Precision functional mapping reveals less inter-individual variability in the child vs. adult human brain, is up on biorxiv!
www.biorxiv.org/content/10.1...

Google’s Gemini 2.5 paper has 3295 authors

arxiv.org/abs/2507.06261

Advancing neural decoding with deep learning - Nature Computational Science A recent study introduces a neural code conversion method that aligns brain activity across individuals without shared stimuli, using deep neural network-derived features to match stimulus content.

Excited to share our News&Views on Kamitani Lab's NatComputSci paper! Their neural code converter enables transformation of brain activity patterns across individuals, and it doesn't need shared stimuli or connectivity information!
www.nature.com/articles/s43...

Excited to co-organize our NeurIPS 2025 workshop on Foundation Models for the Brain and Body!
We welcome work across ML, neuroscience, and biosignals — from new approaches to large-scale models. Submit your paper or demo! 🧠 🧪 🦾

This is why I think the platonic rep hypothesis doesn’t apply to brain-ANN alignment, since most existing (functional?) models are implicitly or explicitly trained to mimic humans.

The assumption of PRH is that the networks are trained independently which doesn’t hold in brain-ANN comparisons.

Brain2Model Transfer: Training sensory and decision models with human neural activity as a teacher Transfer learning enhances the training of novel sensory and decision models by employing rich feature representations from large, pre-trained teacher models. Cognitive neuroscience shows that the hum...

Using intracranial brain recordings to guide representations in a vision-action AI model leads to faster and better training arxiv.org/abs/2506.208...

Just wrapped up my first real foray into analyzing brain data at Brainhack School 2025 a couple weeks ago 🧠💻

I focused on comparing fMRI techniques on a single subject, using fear as a case study.

school-brainhack.github.io/project/many...

Benchmarking methods for mapping functional connectivity in the brain | doi.org/10.1038/s415...

What is the best measure of functional connectivity (FC)?

led by @zhenqi.bsky.social in @natmethods.nature.com ⤵️

Research assistant job posting is live! Come join us in Calgary and be part of a supportive, interdisciplinary team.

careers.ucalgary.ca/jobs/1631661...

Few originals available
Anyone interested can DM me.

an absolute gain on a relative measure. Confusing indeed.

DeepMReye DeepMReye is a software package for magnetic resonance-based eye tracking for (f)MRI experiments. Contact: m.nau[at]vu.nl & markus.frey1[at]gmail.com - DeepMReye

#OSCAwards2025 | open access, data, materials & software: OpenMReye: camera-free magnetic resonance-based eye tracking for research and clinical applications 👀🩻 by @matthiasnau.bsky.social
Find out more: github.com/DeepMReye/

Hi friends, I have news!

Giga connectome, a BIDS app for post-fMRIPrep connectome extraction, is now in @joss-openjournals.bsky.social 🎉

Thanks to @remigau.bsky.social, @clarken.bsky.social, Quentin Dessain, and @lune-bellec.bsky.social for working on this together

Text-to-LoRA: Instant Transformer Adaption

arxiv.org/abs/2506.06105

Generative models can now produce text, image, video. They should also be able to generate models! We trained a Hypernetwork to generate new task-specific LoRA models by simply giving it a description of the task as a text prompt.

Language Models in Plato's Cave Why language models succeeded where video models failed, and what that teaches us about AI

AI may still need some neuroscience:

"AI systems will not acquire the flexibility and adaptability of human intelligence until they can actually learn like humans do, shining brightly with their own light rather than observing a shadow from ours."

🧠🤖

sergeylevine.substack.com/p/language-m...

so sorry to hear 🫂💜

Darwin Gödel Machine: A Commentary on Novelty and Implications – Random Bits of Knowledge Autonomous, self-improving artificial intelligence has long been a theoretical aspiration, yet practical implementations have remained elusive because formal proof–based self-modification is computati...

Darwin Gödel Machine: A Commentary on Novelty and Implications.

antomon.github.io/posts/darwin...

With Mila’s support, our Founder and Scientific Advisor @yoshuabengio.bsky.social is launching @law-zero.bsky.social, a nonprofit organization committed to advancing research and developing technical solutions for safe-by-design AI systems.

Tout système d'IA de pointe devrait reposer sur un engagement essentiel : celui de préserver la joie et les aspirations humaines. Aujourd’hui, nous lançons LoiZéro, une organisation à but non lucratif visant à concevoir des systèmes d'IA sécuritaires. loizero.org

New preprint! 🧠🤖

How do we build neural decoders that are:
⚡️ fast enough for real-time use
🎯 accurate across diverse tasks
🌍 generalizable to new sessions, subjects, and even species?

We present POSSM, a hybrid SSM architecture that optimizes for all three of these axes!

🧵1/7

Brain-tuned Speech Models Better Reflect Speech Processing Stages in the Brain Pretrained self-supervised speech models excel in speech tasks but do not reflect the hierarchy of human speech processing, as they encode rich semantics in middle layers and poor semantics in late la...

More evidence on the fact the fact that fine tuning AI sound processing models reshapes representations in a meaningful way. Perhaps unsurprising- it's just distillation of a big model in a small model, but i often get a lot of skepticism on this idea. arxiv.org/abs/2506.038...

A Brain Graph Foundation Model: Pre-Training and Prompt-Tuning for Any Atlas and Disorder As large language models (LLMs) continue to revolutionize AI research, there is a growing interest in building large-scale brain foundation models to advance neuroscience. While most existing brain fo...

new brain foundation model study, this time with a graph neural net architecture and fMRI, looking at a range of disorders and downstream tasks. arxiv.org/abs/2506.020... #neuroAI

New preprint from the Hasson Lab🪄Delineates coupling b/w regions within the language network AND shows the elements of language that connect them. Led by
@ahmadsamara.bsky.social; thanks @samnastase.bsky.social
Zaid Zada & Uri Hasson. 💥

www.biorxiv.org/content/10.1...

Challenges in multi-task learning for fMRI-based diagnosis: Benefits for psychiatric conditions and CNVs would likely require thousands of patients Abstract. There is a growing interest in using machine learning (ML) models to perform automatic diagnosis of psychiatric conditions; however, generalising the prediction of ML models to completely in...

whoa can't wait to read in details!! We tried very hard (for years) to do something like that with static connectivity, and no success doi.org/10.1162/imag...

🚨Preprint!🚨 So excited to share my 1st major study as a postdoc at @holmeslab-bhi.bsky.social!🤘

www.biorxiv.org/content/10.1...

Using a transdiagnostic framework, we found that brain network dynamics were able to discriminate individualized psychiatric symptom profiles with high accuracy 🧵1/9👇

YouTube video by Jeremy Tuloup Jupyter Frontends & JupyterLite updates | Jupyter Open Studio Day 2025

just watched a very great overview of the latest updates in Jupyterlab and Jupyterlite

- Improved code consoles, RTC,
- Support for R via xeus-R; consoles
- Improvements with xeus-python with some pip magic
- JupyterAI and Jupyterlite-AI, with options for in-browser LLMs
- hybrid kernels 🤯 (cont ⬇️)

Precise individual measures of inhibitory control Nature Human Behaviour - Collecting an extensive publicly available dataset on four inhibitory control tasks, Gratton et al. show that more than 1,000 trials per participant are necessary to reduce...

Excited to announce that our paper (led by @hyejinjadelee.bsky.social & Derek Smith) is out in @nathumbehav.nature.com
go.illinois.edu/Lee2025_NHB
We use 36 days of testing in 4 inhibitory control tasks to examine the precision of inhibitory control measures in *individuals* and how they vary
🧵👇