Marcelo Mattar

Our new paper is out! When navigating through an environment, how do we combine our general sense of direction with known landmark states? To explore this, @denislan.bsky.social used a task that allowed subjects (or neural networks) to choose either their next action or next state at each step.

NYU Application Support Group Matching Form Dear Prospective Neuroscience & Psychology PhD Students, We are a group of current NYU Neuroscience & Psychology PhD students who would like to help you with your PhD applications. We want to support...

Thinking of applying to US-based Ph.D. programs in neuroscience, psychology, or cognitive science? We’ve got your back! NYU’s Application Support Group is a student-led, free mentorship program offering 1-on-1 support and guidance. Apply now! docs.google.com/forms/d/e/1F...

🚨 Fully Funded PhD positions

Gonda Brain Institute, Sharp Lab

We will explore how people build and deploy world models efficiently for planning and decision making. We will also seek to characterize how world models construal and use is biased in anxiety.

Deadline: 1 Sept 2025. Please share!

📣 I'm looking for a postdoc to join my lab at NYU! Come work with me on a principled, theory-driven approach to studying language, learning, and reasoning, in humans and AI agents.
Apply here: apply.interfolio.com/170656
And come chat with me at #CogSci2025 if interested!

Fantastic work by our (now former) lab manager Liv Christiano. We assess the test-retest reliability of OPM and compare it to fMRI and iEEG. 🧠📄🧵

Reliability and signal comparison of OPM-MEG, fMRI & iEEG in a repeated movie viewing paradigm Optically pumped magnetometers (OPMs) offer a promising advancement in noninvasive neuroimaging via magnetoencephalography (MEG), but establishing their reliability and comparability to existing metho...

How reliable is OPM-MEG, and how does it compare to other neuroimaging modalities? 🤔

In a new preprint with ‪@s-michelmann.bsky.social‬, we evaluate the reliability of OPM-MEG within & between individuals, and compare it to fMRI & iEEG during repeated movie viewing. 🧠

📄 doi.org/10.1101/2025...

Beautiful work by @neurozz.bsky.social and @annaschapiro.bsky.social !

A gradient of complementary learning systems emerges through meta-learning Long-term learning and memory in the primate brain rely on a series of hierarchically organized subsystems extending from early sensory neocortical areas to the hippocampus. The components differ in t...

Excited to share a new preprint w/ @annaschapiro.bsky.social! Why are there gradients of plasticity and sparsity along the neocortex–hippocampus hierarchy? We show that brain-like organization of these properties emerges in ANNs that meta-learn layer-wise plasticity and sparsity. bit.ly/4kB1yg5

Considering What We Know and What We Don’t Know: Expectations and Confidence Guide Value Integration in Value-Based Decision-Making Abstract. When making decisions, we often have more information about some options than others. Previous work has shown that people are more likely to choose options that they look at more and those t...

Our newest paper, led by Romy Froemer
and @fredcallaway.bsky.social‬, is now out in Open Mind: “Considering What We Know and What We Don’t Know: Expectations and Confidence Guide Value Integration in Value-Based Decision-Making”

direct.mit.edu/opmi/article...

Want to know what kinds of studies webcam-based eye tracking can be used for? Here's our take on the current tech. This certainly isn't the first paper on this topic, but it provides some converging evidence about the viability of eye tracking with online methods. online.ucpress.edu/collabra/art...

Very cool. Will the talk be recorded?

Looks super interesting! Can't wait to read!

YAY!!! I'm so happy for you, Anna! 🎉 Can't wait to see what your post-tenure research looks like! I'll be watching closely.

Thanks!

And yes, the reason our tiny models did so well is that the tasks we studied were so simple (like most tasks studied in neuroscience!)

In more complex tasks, large models will definitely outperform, assuming you have enough data to train them.

Discovering cognitive strategies with tiny recurrent neural networks - Nature Modelling biological decision-making with tiny recurrent neural networks enables more accurate predictions of animal choices than classical cognitive models and offers insights into the underlying cog...

Here's the link to our paper: doi.org/10.1038/s415...
Here's the link to the Centaur paper: www.nature.com/articles/s41...
And here's Marcel Binz' thread on Centaur: bsky.app/profile/marc...

Also thrilled that this work appears in the same @nature issue as another paper I contributed to, Centaur! Like TinyRNNs, it excels at predicting human behavior. Impressively, Centaur works across many tasks, but this comes with a trade-off in model interpretability.

Crucially, we can discover these patterns for each individual! This is a game-changer for computational psychiatry, as it lets us pinpoint exactly how cognitive processes differ across people, without the constraints of a prespecified learning/decision model like RL.

This approach uncovered several patterns in the animal's decisions that previous models missed. Eg we found state-dependent learning rates, novel patterns of perseveration, and a peculiar type of forgetting where an action value decays towards the value of the alternative action.

But prediction is only half the story; we also need interpretability! We viewed tiny RNNs as dynamical systems with inputs (observations/rewards) and outputs (actions). Given their small size, we could visualize how their states evolved and discover the strategies they learned.

Across six different reward-learning tasks, tiny RNNs consistently outperform dozens of classical cognitive models in predicting the choices of individual animals and humans. Surprisingly, networks with just 2-4 units often performed the best in those simple lab tasks.

Our solution was to use very small RNNs, composed of 1-4 units. Those models are still great at modeling biological behavior without the need for pre-specified assumptions, but are small enough for us to interpret their mechanisms, combining the best of both worlds.

Cognitive modeling has always faced a dilemma. Traditional approaches based on normative or mechanistic assumptions are simple to understand but often cannot capture behavioral complexity. Large AI models are excellent predictors but are "black boxes" that are hard to interpret.

Discovering cognitive strategies with tiny recurrent neural networks - Nature Modelling biological decision-making with tiny recurrent neural networks enables more accurate predictions of animal choices than classical cognitive models and offers insights into the underlying cog...

Thrilled to see our TinyRNN paper in @nature! We show how tiny RNNs predict choices of individual subjects accurately while staying fully interpretable. This approach can transform how we model cognitive processes in both healthy and disordered decisions. doi.org/10.1038/s415...

Yay, amazing news! Congrats Lucy!

Excited to share our upcoming workshop on neuroscience, reinforcement learning, and decision making at RLDM 2025 in Dublin, Ireland — June 11–14!

Check out the terrific speaker lineup:

🔗 sites.google.com/view/neurorl...

Co-organized with @angelaradulescu.bsky.social

@rldmdublin2025.bsky.social

Neural dynamics of an extended frontal lobe network in goal-subgoal problem solving Complex behavior calls for hierarchical representation of current state, goal, and component moves. In the human brain, a network of “multiple-demand” (MD) regions underpins cognitive control. We reco...

Now out 🚨 🧪 : our preprint describing dynamics of an extended frontal lobe network (4 cortical regions) in monkeys solving complex multi-step spatial problems! We observe distributed codes for goals, states, and planned moves across PFC!

www.biorxiv.org/content/10.1...
#neuroscience #compneuro

🧵👇

📢 I'm happy to share the preprint: _Reward-Aware Proto-Representations in Reinforcement Learning_ ‼️

My PhD student, Hon Tik Tse, led this work, and my MSc student, Siddarth Chandrasekar, assisted us.

arxiv.org/abs/2505.16217

Basically, it's the SR with rewards. See below 👇

A schematic of our method. On the left are shown Bayesian inference (visualized using Bayes’ rule and a portrait of the Reverend Bayes) and neural networks (visualized as a weight matrix). Then, an arrow labeled “meta-learning” combines Bayesian inference and neural networks into a “prior-trained neural network”, described as a neural network that has the priors of a Bayesian model – visualized as the same portrait of Reverend Bayes but made out of numbers. Finally, an arrow labeled “learning” goes from the prior-trained neural network to two examples of what it can learn: formal languages (visualized with a finite-state automaton) and aspects of English syntax (visualized with a parse tree for the sentence “colorless green ideas sleep furiously”).

🤖🧠 Paper out in Nature Communications! 🧠🤖

Bayesian models can learn rapidly. Neural networks can handle messy, naturalistic data. How can we combine these strengths?

Our answer: Use meta-learning to distill Bayesian priors into a neural network!

www.nature.com/articles/s41...

1/n

Congratulations, Fred! Fred has been an outstanding postdoc in my lab. He combines exceptional experimental and modeling skills and is a caring and attentive mentor. This is an excellent opportunity for prospective students and postdocs looking for a lab to join!

Huge congratulations to Q on becoming Presidential Assistant Professor at City U Hong Kong! Q is truly one of the best researchers building neural network models of human memory. So excited to see all of the amazing things his lab will do!