Excited to be in Amsterdam for #CCN2025! If you're here, check out the presentations from our lab 👇@qinwenshuo.bsky.social @ziruichen.bsky.social @manasimalik.bsky.social

@cogcompneuro.bsky.social

If you're heading to @cogscisociety.bsky.social check out the presentations from our lab

And congratulations again to @emaliemcmahon.bsky.social on her Glushko dissertation prize 🥳

I’m happy to be at #VSS2025 and share what our lab has been up to this year!

I’m also honored to receive this year’s young investigator award and will give a short talk at the awards ceremony Monday

We found distinct neural signatures for physical interactions (eg fighting, dancing) and communicative interactions along posterior to anterior regions of the recently proposed lateral pathway. A distinction that has been difficult to find with hypothesis-driven experiments

2/n

While we observed marginal differences between species, these differences could not be attributed to language model prediction alone, suggesting they arise from other inter-species or recording modality factors 5/n

We compared vision, language, and multimodal models to ventral stream responses and found strikingly similar results across species. Notably, pure language models (evaluated on image captions) predicted ventral stream responses almost as well as pure vision models in both humans AND macaques 4/n

We collected electrophysiology data from the ventral visual stream of rhesus macaques while they viewed the same 1,000 natural scene images as human fMRI participants from the Natural Scenes Dataset (the "NSD1000") 3/n

New preprint “Monkey See, Model Knew: LLMs accurately predict visual responses in humans AND NHPs”
Led by Colin Conwell with @emaliemcmahon.bsky.social Akshay Jagadeesh, Kasper Vinken @amrahs-inolas.bsky.social @jacob-prince.bsky.social George Alvarez @taliakonkle.bsky.social & Marge Livingstone 1/n

Monkey See, Model Knew: LLMs accurately Predict Human AND Macaque Visual Brain Activity
Colin Conwell, @emaliemcmahon.bsky.social, Akshay Vivek Jagadeesh, Kasper Vinken, @amrahs-inolas.bsky.social, @jacob-prince.bsky.social, George Alvarez, @taliakonkle.bsky.social and Marge Livingstone

Vision and language representations in multimodal AI models and human social brain regions during natural movie viewing led by
@hsmall.bsky.social, with @hleemasson.bsky.social and Stewart Mostofsky

Our paper "Relational visual representations underlie human social interaction recognition" led by @manasimalik.bsky.social is now out in Nature Communications
www.nature.com/articles/s41...

Our paper "Hierarchical organization of social action features in the lateral visual stream" led by @emaliemcmahon.bsky.social with Mick Bonner is now out in @currentbiology.bsky.social

www.sciencedirect.com/science/arti...

📣 Now out in J Neuro, led by @hleemasson.bsky.social: Using EEG-fMRI fusion we find that observing others' touch events (social vs. non-social) is processed in a feedforward manner through social perceptual brain regions
www.jneurosci.org/content/earl...
#neuroskyence #PsychSciSky

Great examples! I get the sense that the internal prompt engineering does help somewhat, and possibly even within a chat...

For example, if I first ask for a "closed, empty box", it seems to do better with "closed box with sheep inside, not visible"(though still only 50-50!)