iiinova — where your digital self keeps evolving

Reposted by Abdulkadir Gokce

Martin Schrimpf @mschrimpf.bsky.social · 01/09/2026

Hiring postdocs @EPFL: please reach out and/or apply to the AI Center fellowship www.epfl.ch/research/fun.... My #NeuroAI group is excited about testing and building models of brain and behavior, from sensory through language to cognition, and from healthy to dysfunction (treatment)! 🧠🤖🧪

epfl.ch

EPFL AI Center and Swiss AI Initiative Postdoctoral Fellowships

The 3rd call is now open with a submission deadline on 9 November 2026 17:00 (CET).

0158

Reposted by Abdulkadir Gokce

Badr AlKhamissi @bkhmsi.bsky.social · 03/08/2026

Excited to be at #CCN2026 in NYC! I'll be presenting our spotlight poster, Topo-Omni, with @hannesmehrer.bsky.social 🧠 📍 Poster F14 — Session F 🗓️ Thursday, Aug 6 Come by if you're around, we'd love to talk about how our multimodal model can discover functionally selective brain regions!

1256

Reposted by Abdulkadir Gokce

Hannes Mehrer @CCN2026 @hannesmehrer.bsky.social · 04/08/2026

"So, you have a highly predictive model, now what?" At this #CCN2026 event on Wed, 5 Aug, 4pm at NYU Skirball Theatre I will present work on using ANN brain models for visual prostheses (tinyurl.com/modelguidedm...) and for dyslexia (tinyurl.com/neuroaidysle..., first author: Melika Honarmand).

0124

Reposted by Abdulkadir Gokce

Martin Schrimpf @mschrimpf.bsky.social · 04/08/2026

Come say hi at #CCN2026! Scaling laws for visual cortex, topographic multimodal models, video generation for the lateral stream, model-enabled applications, and a GAC on whether #NeuroAI is asking the right questions

0234

Reposted by Abdulkadir Gokce

Yingtian (David) Tang @davidtyt.bsky.social · 07/07/2026

🚨 NEW PREPRINT Videos strongly shape activity across the visual cortex. But can we design videos that maximally drive specific brain regions? We present NEvo 🧬🧠 — a neural-guided evolutionary framework that synthesizes videos to maximally activate target visual ROIs. (1/10)

13010

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

11/ 🔗 Resources: 🌐 multimodal-brain-scaling.epfl.ch 📄 openreview.net/forum?id=OQ6jQHJPTT 💻 github.com/epflneuroailab/multimodal-brain-scaling 🤗 huggingface.co/datasets/epfl-neuroai/multimodal-brain-scaling 🎤 icml.cc/virtual/2026/poster/64356

010

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

10/ Come say hi at #ICML2026 in Seoul! 🇰🇷 📍 Poster Session 4 Wed Jul 8, 2:30–4:15 PM KST Hall A – Poster #901

100

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

9/ Lots more in the appendix: • alternative alignment metrics (RSA/CKA) • per-ROI alignment vs. accuracy plots across datasets • adversarially fine-tuned models • Qwen3-VL vision–language model results • LoRA …and more.

110

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

8/ Where does task accuracy fit? Better ImageNet/Ecoset accuracy tracks higher alignment — but with diminishing returns. And neural fine-tuning trades a small accuracy drop for an alignment gain: brain-likeness and task performance don't always move together.

130

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

7/ Readout matters too. Better mapping design leaves room for further gains. We introduce a subject-shared cross-attention readout. It tops linear readout on average while using ~10× fewer parameters — sharing a mapping across subjects + lightweight per-subject heads.

100

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

6/ The biggest lever? The neural data used to fit the model→brain mapping. Alignment grows log-linearly with the number of stimulus–response pairs, reliably across all modalities — and several datasets show no plateau yet. More neural data still pays off.

100

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

5/ These gains transfer: fine-tuning on one neural dataset improves alignment on others, even across modalities. Neural supervision nudges models toward broadly brain-like features; strikingly, even noisier EEG can improve predictivity for cleaner electrophysiology.

100

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

4/ Can we beat saturation? Yes — with neural data. We fine-tune a backbone on ONE dataset using a hybrid task+neural objective, then benchmark on ALL. Alignment keeps climbing with neural samples, with no saturation yet at today's dataset sizes.

110

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

3/ A consequence: model differences melt away with scale. At low scale, architecture matters a lot (CORnet-S leads, ViTs lag). With enough data & compute, families converge to the same alignment band — ViTs can even overtake. Pretraining washes out inductive biases.

110

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

2/ Scaling task pretraining helps — alignment rises with compute, data & model size across electrophysiology, fMRI, EEG & MEG. But it saturates. We fit scaling curves on 600+ models and they extrapolate to held-out large models (DINOv2/v3, CLIP). More scale ≠ endless gains.

110

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

1/ We disentangle 3 main stages in brain-modeling pipelines: 1) Task pretraining: learn representations from tasks 2) Neural fine-tuning: update the backbone on neural data 3) Neural mapping: fit readouts from model features to brain responses Then ask: how does each scale?

100

Abdulkadir Gokce @akgokce.bsky.social · 02/07/2026

🧠 What's the best way to spend compute & neural data to build brain-aligned models? Our new ICML 2026 paper charts the scaling laws across 8 neural datasets & 600+ vision models. 👇 🌐 multimodal-brain-scaling.epfl.ch w/ @davidtyt.bsky.social & @mschrimpf.bsky.social

194

Reposted by Abdulkadir Gokce

Badr AlKhamissi @bkhmsi.bsky.social · 16/06/2026

🚨 New Preprint! 🧠 We gave an AI model one simple rule: rearrange your neurons so that nearby ones respond alike. We never told it what a face, a voice, or a sentence was. It grew brain-like maps for all three anyway. 🧵👇 🌐 Website: topo-omni.epfl.ch

14817

Reposted by Abdulkadir Gokce

Badr AlKhamissi @bkhmsi.bsky.social · 04/06/2026

So why build toward a brain-encoding foundation model? → Simulate fMRI responses to sensory stimuli → Insights about the brain → Path toward clinical applications MIRAGE is our first step. 🧠 More on the project site: mirage-brain.epfl.ch w/ @akgokce.bsky.social & @mschrimpf.bsky.social

093

Reposted by Abdulkadir Gokce

Badr AlKhamissi @bkhmsi.bsky.social · 01/06/2026

+ we've a demo: play a clip and watch MIRAGE's predicted whole-brain activity light up in sync 🧠🍿 📄 Paper: arxiv.org/abs/2605.29850 💻 Code: github.com/epflneuroail... 🤗 Model: huggingface.co/epfl-neuroai... 🌐 Demo: mirage-brain.epfl.ch Joint work w/ @akgokce.bsky.social & @mschrimpf.bsky.social

011

Reposted by Abdulkadir Gokce

Badr AlKhamissi @bkhmsi.bsky.social · 01/06/2026

🧠 When you watch a movie, your brain blends sight, sound, and speech into a single experience. Should models of the brain blend them too, or keep the senses separate until the very end? We built MIRAGE to find out. It sets a new SOTA for predicting whole-brain fMRI from movies. 🧵

1299

Reposted by Abdulkadir Gokce

Hannes Mehrer @CCN2026 @hannesmehrer.bsky.social · 07/10/2025

🧠 New preprint: we show that model-guided microstimulation can steer monkey visual behavior. Paper: arxiv.org/abs/2510.03684 🧵

13715

Reposted by Abdulkadir Gokce

benlonnqvist.bsky.social @benlonnqvist.bsky.social · 22/04/2025

AI vision is insanely good nowadays—but is it really like human vision or something else entirely? In our new pre-print, we pinpoint a fundamental visual mechanism that's trivial for humans yet causes most models to fail spectacularly. Let's dive in👇🧠 [arxiv.org/abs/2504.05253]

2123