Excited for my lab's first @cogcompneuro.bsky.social ! We are presenting a mix of projects, ranging from infant modeling and neuroimaging to one-shot category learning #CCN2026

Finally, the developmental literature provides a rich set of benchmarks for model comparison Evaluating the developmental trajectory of models may offer a new metric for biological alignment, and also improve model efficiency by uncovering the minimal conditions necessary for a capacity to arise

Third, the objectives of children are open-ended and best described by a desire to seek novelty within an understandable "Goldilocks zone" This can be modeled using predictive frameworks where models prioritize data that progressively reduces error, i.e., where they make the most progress

Second, children learn from many fewer examples than current models However, their experiences, starting in the womb, provide an excellent curriculum for learning. And, indeed, training models on a child-like sequence of experience improves their robustness and efficiency

First, the brain develops along a sensory-association gradient, with early developing areas providing a foundation for the development of higher level ones This maturation process may prevent catastrophic forgetting, and could be implemented in AI by training model layers hierarchically

He's playing "putting babies in the MRI at work" My heart ❤️‍🔥

Barenholtz & Tarr (2007) - Reconsidering the Role of Structure in Vision The real structural descriptions are the friends we make along the way.

I think it's important for neuroscientists to reinforce this too. It's the first thing I teach in my cog neuro class. A thing isn't more real just because it's found in the brain. Every psychological thing must be in the brain at some level

However, we also found differences in maturity among the three pathways. Across gestational time points, the organization of the dorsal pathway was consistently the most adult-like and the ventral pathway was consistently the least adult-like

This large-scale organization was present by term birth, but was markedly less clear in early- and pre-term infants.

We found that on the very first day of life infants already exhibit an adult-like hierarchical three-pathway visual organization, corresponding to ventral, lateral, and dorsal pathways I love how clear this pattern is, you can literally zoom in and follow the hierarchy of each pathway area to area

Excited to be at my first @cogscisociety.bsky.social meeting! Come check out my talk tomorrow morning on the organization of the human visual system at birth in @noranewcombe.bsky.social Rumelhart award symposium #CogSci2025

Indeed, when we plot human and model performance using these estimates, we see a massive discrepancy between the two. Children achieve much strong performance with far fewer examples.

We generally found that kids outperformed most models *except* for models trained on massive datasets like CLIP. However, these large models may be given more training than a child could realistically experience in their lifetime.

To our surprise, children performed extremely well on this task, even at the fastest speed and when the contours of the objects were disrupted.

Here we administered what we thought would be a challenging task for kids. Three- to five-year-olds were asked to identify object outlines, that were both forward and backward masked, could be presented as quickly as 100 ms, and could even have their contours perturbed or deleted.

Come check out my poster with @mikearcaro.bsky.social tomorrow afternoon in the pavilion where we describe the cortical and subcortical organization of the newborn visual system! @vssmtg.bsky.social #VSS2025

Vlad-Lab: Year 0 Starting with a fresh board

I'm hiring a full-time research assistant to start this Summer! The lab studies perception and cognition from infancy to adulthood using neuroimaging, behavioral, and computational approaches For more details about the lab and the position, see: vlad-lab.com/join #psychjobs #neurojobs

However, we also found a differential pattern of maturity across the pulvinar connectivity maps, such that the pattern of connections with the ventral visual pathway was less adult-like than for other visual areas.

Remarkably, the pattern of connections with V1 was spatially correlated with *functional* retinotopic responses of the adult pulvinar. That is, areas of the pulvinar most connected with dorsal V1 in neonates responded to stimuli in the lower-visual-field of adults (and vice versa for ventral V1)

This structured connectivity was even found when splitting a single area, V1, into dorsal and ventral components, revealing distinct patterns of connectivity corresponding to each component

Zooming into the pulvinar, we found that that each cortical area exhibited a distinct peak of connectivity. Collectivity, visual areas showed gradients of connectivity in the pulvinar corresponding to the hierarchical organization of ventral, lateral, and dorsal visual pathways.

First, we found that the large-scale organization of pulvino-cortical connections was already in place at birth, such that the neonate pulvinar exhibited connections to each individual visual area.

But how adult like is the pulvinar in neonates and to what extent is it already wired up with cortex? We tested this question by examining the organization and maturity of *structural connections* between the pulvinar and areas of occipital, ventral, lateral, and dorsal visual cortices.

Of course huge thanks to all my amazing mentors that helped me get to this point: Stella Lourenco, Danny Dilks, Marlene Behrmann, Mike Arcaro, @noranewcombe.bsky.social, and Justin Harris Bonus: baby vlad presenting his first poster as an undergrad

By working at the intersection of developmental psychology, neuroscience, and artificial intelligence, our goal is to shed light on deep philosophical questions regarding the origins of cognition, as well as the processes by which we may 'grow' more human-like AI.

Super excited to announce that I'll be joining Temple University as an assistant professor of Psychology and Neuroscience! The Vision Learning and Development Lab will officially open its doors in Summer 2025! vlad-lab.com

Now out in Developmental Cognitive Neuroscience! Here we examined the degree to which dorsal and ventral visual pathways can functionally reorganize after children have half of their brain surgically removed (!!) doi.org/10.1016/j.dc... #cogsci #devpsy #neuroscience