Amir Ozhan Dehghani

I am excited to share that my undergraduate research has now been accepted to #ICLR2025. 📎: openreview.net/pdf?id=wMgr7wB… 🧑‍💻:github.com/BashivanLab/CB… Tweeprint🧵⬇️

New paper 🚨 #ICLR26 Most world models predict the future from a past trajectory. But neuroscience suggests that such inference can instead be made from temporally independent experiences. We built the Episodic Spatial World Model (ESWM), a model that does exactly this:

So proud of this work co-led by @aspa1313 and @johanobandoc Fantastic collaboration with @AaronCourville & @pcastr Tldr; Constraining representations within isotropic Gaussian distribution massively improves performance during RL!

Excited to give a talk at #Cosyne2026 about my PhD work! We show that RNNs trained on visual search converge on brain-like solutions, producing primate-like behavior and neural representations. Happy to chat if you're at Cosyne! 📅 March 15, 10:45–11 AM 📎 Preprint 👇

🔴 Live: Our MAIN 2025 Keynote #3 Yoshua Bengio @Yoshua_Bengio (Université de Montréal & Mila)  “Scientist AI: disentangling agency and causal understanding” #AI #Neuroscience #AIsafety #MAIN2025

We’ll be at CCN next week with two posters, one contributed talk and one satellite talk! If you are interested in neural coding properties in primate hippocampus, modeling visual search, or cortical topography come and talk to us! See👇for details:

McGill has been named the #1 university in Canada — and ranked 27th in the world — in the QS World University Rankings 2026. 🏆 This achievement belongs to every student, professor, researcher, and staff member who shapes McGill. 🔗 Read more: mcgill.ca/x/iPf

This sounds very interesting!!

🚨 Preprint Alert 🚀 📄 seq-JEPA: Autoregressive Predictive Learning of Invariant-Equivariant World Models arxiv.org/abs/2505.03176 Can we simultaneously learn both transformation-invariant and transformation-equivariant representations with self-supervised learning (SSL)? TL;DR Yes! This is possible via simple predictive learning and architectural inductive biases – without extra loss terms and predictors! 👇🧵 (1/7)

“A network of face patches in human prefrontal cortex for social processing of faces” biorxiv.org/content/10.110… Let us characterize a novel network of four frontal face patches (FFPs) arranged dorsoventrally in the human lateral prefrontal cortex! 🔢 @rajimehr @ElaheYargholi

Very cool talk on Human AI alignment workshop @iclr_conf 🇸🇬🔥 Never thought I would get chess lessons here

Before ICLR 2025 comes to and end tomorrow, a few #NeuroAI impressions from Singapore. First, very happy to present our work on TopoLM as an oral, here with @neil_rathi x.com/HannesMehrer/s… paper: doi.org/10.48550/arXiv… code: github.com/epflneuroailab…

Follow Mila researchers on day 2 of #ICLR2025 Full schedule of Mila-affiliated presentations at @iclr_conf here: mila.quebec/en/news/follow…

Together with @neil_rathi I will present our #ICLR2025 Oral paper on TopoLM, a topographic language model! Oral: Friday, 25 Apr 4:18 p.m. (session 4C) Poster: Friday, 25 Apr 10 a.m. --> Hall 3 + Hall 2B Paper: arxiv.org/abs/2410.11516 Code and weights: github.com/epflneuroailab…

📌 Poster Session: When📅: Friday, April 25 from 10:00 AM to 12:30 PM Where: Hall 3 + Hall 2B #60 Check out our project page: amirozhan.github.io/CBSOM/ See you soon!

I’ll be at @iclr_conf presenting our work “Credit-based self organizing maps: training deep topographic networks with minimal performance degradation” — if you're into #NeuroAI, cortical topography, vision, or just up for a good convo, drop by and DM me! 🧠✨

@SKordasti سلام، درسته که فعلاً نمی‌شه از این مدل‌ها به عنوان ابزار قطعی برای تشخیص پزشکی استفاده کرد، اما به نظرم می‌تونن کمک خیلی خوبی برای تکمیل تشخیص پزشک باشن و حتی در کشف درمان‌های جدید هم نقش مفیدی داشته باشن.

هر وقت اجازه دادید چت‌جی‌پی‌تی و مشابهش برایتان شیمی‌درمانی و رادیوتراپی تجویز کند، در خصوص سایر تشخیص و درمانهای مهم پزشکی هم ازش نظر بخواهید.

@ajaysub110 Thank you Ajay!

@great_caster Congrats!

I am excited to share that my undergraduate research has now been accepted to #ICLR2025. 📎: openreview.net/pdf?id=wMgr7wB… 🧑‍💻:github.com/BashivanLab/CB… Tweeprint🧵⬇️

People who might be interested: @talia_konkle, @martin_schrimpf, @nmblauch, @apurvaratan, @Nancy_Kanwisher,@dyamins

11/Last (but not least), this work wouldn’t have been possible without the help of my collaborators @qianxinyu8, @Asa_Farahani and research advisor @PouyaBashivan ⭐️