iiinova — where your digital self keeps evolving

Reposted by Subha Nawer Pushpita

Nathan Lambert @natolambert.bsky.social · 11/09/2026

A great read. I have similar feelings about how AI labs approach progress directly and without nurturing of scientific communities & intuition. The math research community went through the transition the fastest, so it was felt most. Other fields next. terrytao.wordpress.com/2026/09/11/a...

terrytao.wordpress.com

A Severe Misalignment of AI in Mathematics

I am proud to be among the list of 25 initial signatories — all Fields Medallists — to the declaration below, which grew out of discussions between ourselves over the last week. We have…

44515 0

Reposted by Subha Nawer Pushpita

Tim Verstynen @tdverstynen.bsky.social · 10/09/2026

It’s easy to see why so many apocalyptic super-AI scenarios being pushed sound more like marketing than genuine risk assessment. Just imagine how other industries would behave if they truly believed one of their products could destroy humanity. A rando 🧵 1/6

132 0

Reposted by Subha Nawer Pushpita

Nancy Kanwisher @nancykanwisher.bsky.social · 08/08/2026

What a joy to visit today with the wonderful Molly Potter, to whom I owe my career. Molly's ideas still motivate much of my research and her example of what it means to be a mentor, scientist, and activist continue to inspire. Yay Molly!

21115 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/08/2026

On the same boat. Flight out of LGA got cancelled. Rebooked myself (the airlines didn't do anything at all other than emailing flight's cancelled and they couldn't rebook me). Ubered to JFK to catch a flight that's been delayed for the third time. Who knows if this one will leave either

120 0

Reposted by Subha Nawer Pushpita

Edvard I Moser @edvardmoser.bsky.social · 06/08/2026

Out now in @science.org : How does the brain rapidly scan space to guide navigation? We found that the brain’s internal map is continuously retuned from moment to moment, focusing neural activity on what matters most. 🧵1/8 👉 www.science.org/doi/10.1126/...

science.org

Adaptive modulation of theta sweeps in the brain’s navigation circuit

Efficient navigation requires prioritizing information from behaviorally relevant locations. In rats, such selective sampling may arise from theta-paced sweeps in grid and place cell populations, whic...

29229 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 03/08/2026

Excited to be at #CCN2026! I will be presenting our Spotlight poster on cortical mechanisms of naturalistic audiovisual processing! If you are interested in how humans process such temporally varying information load across modalities, please come chat! 📍 Poster F13 - Session F 🗓️ Thursday, Aug 6

050 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 15/05/2026

Excited to attend #VSS2026 starting tomorrow! I will also be presenting my work with Leila Wehbe on two cortical mechanisms of natural audiovisual processing on Sunday, 11:15 am, talk room 1 @Multisensory Processing session.

020 0

Reposted by Subha Nawer Pushpita

Eghbal Hosseini @eghbal-hosseini.bsky.social · 04/02/2026

How do diverse context structures reshape representations in LLMs? In our new work, we explore this via representational straightening. We found LLMs are like a Swiss Army knife: they select different computational mechanisms reflected in different representational structures. 1/

13911 0

Reposted by Subha Nawer Pushpita

Sam Gershman @gershbrain.bsky.social · 09/01/2026

With some trepidation, I'm putting this out into the world: gershmanlab.com/textbook.html It's a textbook called Computational Foundations of Cognitive Neuroscience, which I wrote for my class. My hope is that this will be a living document, continuously improved as I get feedback.

16590237 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 22/12/2025

8/n For detailed analyses and results, see our paper: aclanthology.org/2024.conll-1.... For even deeper analyses, see my MEng thesis: dspace.mit.edu/handle/1721....

aclanthology.org

000 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 22/12/2025

7/n We also benchmarked several SOTA VLMs and found evidence that the training objectives and diet of Vision-Language Models (VLMs) significantly impact their psychometric predictive power, more so than their size.

100 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 22/12/2025

6/n We collected human judgments of how grounded each word in an image description is in the image itself. Intuitively, ImageConditionedTextSurprisal-TextSurprisal for a word indicates the reduction of surprisal for presence of image. This measure significantly predicts human groundedness judgments.

100 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 22/12/2025

5/n People make errors during reading (choosing the wrong word in the image conditioned A-maze setup), and we found that it's not arbitrary and in fact, happens for words that are contextually highly surprising(as measured by surprisal of both VLMs and LLMs, note the 5th row).

110 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 22/12/2025

4/n Unlike the strong effect conditions (correct,no,wrong) have on RT when LLM surprisals are used, there's little to no effect of conditions on RT when VLM surprisals are used. Surprisal already encodes image-related information, so condition likely doesn't provide any additional gain.

100 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 22/12/2025

3/n We developed a novel experimental paradigm and ran a study with human subjects in three conditions: correct image, no image, wrong image preview. We found resounding evidence for Comprehensive grounding - there's facilitation for both open and closed-class words in the correct image preview.

100 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 22/12/2025

2/n H1: Lexical Grounding Hypothesis, where facilitation will be observed only for words that are straightforwardly grounded in the image, and Comprehensive Grounding Hypothesis, where Facilitation will be observed for all types of words in the description, regardless of their level of groundedness.

100 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 22/12/2025

1/n If people see an image before reading the description of that image, we analyzed two hypotheses regarding how the presence of the right image (as opposed to no image or image irrelevant to the description) facilitates the reading of the description.

100 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 22/12/2025

Better late than never. This is a slightly old work I did with @rplevy.bsky.social during my MEng at MIT. I presented this work at @conll-conf.bsky.social last year. The key scientific question we ask here was what is the underlying cognitive mechanism of image description comprehension?

110 0

Reposted by Subha Nawer Pushpita

Jewish Voice for Peace @jvp.bsky.social · 14/12/2025

We grieve the loss of life in Sydney, Australia, where gunmen attacked a celebration of the first night of Hanukkah, killing at least 11 people. This was a terrifying incident of hate-fueled violence that no community should face.

2387114 0

Reposted by Subha Nawer Pushpita

Melanie Mitchell @melaniemitchell.bsky.social · 14/12/2025

Thinking of all the colleagues and friends I have who are connected to Brown University. What a devastating day for Brown, and for all of us.

112012 0

Reposted by Subha Nawer Pushpita

Anna (Anya) Ivanova @neuranna.bsky.social · 11/12/2025

The last chapter of my PhD (expanded) is finally out as a preprint! “Semantic reasoning takes place largely outside the language network” 🧠🧐 www.biorxiv.org/content/10.6... What is semantic reasoning? Read on! 🧵👇

biorxiv.org

Semantic reasoning takes place largely outside the language network

The brain's language network is often implicated in the representation and manipulation of abstract semantic knowledge. However, this view is inconsistent with a large body of evidence suggesting that...

29126 0

Reposted by Subha Nawer Pushpita

Apurva Ratan Murty @apurvaratan.bsky.social · 04/12/2025

Need more fMRI data (beyond the amazing NSD)? Introducing MOSAIC! Incredible effort led expertly by Ben Lahner, with help from grad student Mayukh Deb. Work in collaboration with the amazing Aude Oliva! @neurosky.bsky.social. More below..

1269 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 21/11/2025

We have studied how human brains process naturalistic audiovisual stimuli. The statistical frameworks we introduced- DIP and PIP- go beyond audio and video, they’re general tools for quantifying how two feature-based models differ or converge in predicting a common target over time.

020 0

Reposted by Subha Nawer Pushpita

Anna (Anya) Ivanova @neuranna.bsky.social · 23/10/2025

🤖📈 How do LLMs use their depth? Akshat Gupta led a fun project to find out! We leverage TunedLens (~linear decoding of tokens) to explore how LLMs' internal representations change from layer to layer. Preprint: arxiv.org/abs/2510.18871 1/

36917 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/11/2025

The statistical frameworks we introduced- DIP and PIP- go beyond audio&video, they’re general tools for quantifying how two feature-based models differ or converge in predicting a common target over time, revealing patterns of dominance and joint encoding across brains🧠, models🤖, and modalities🎬🎧.

010 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/11/2025

We uncover two complementary cortical mechanisms underlying how the brain processes naturalistic audiovisual information - one system flexibly switches what sensory modality it encodes; another one jointly encodes auditory + visual inputs. ⚡ Quick thread with the main findings 👇

000 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/11/2025

9/n The coexistence of these systems points to a cortical architecture that flexibly reweights sensory inputs while maintaining balanced multimodal representations, supporting robust comprehension of complex natural events.

010 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/11/2025

8/n Notably, our multimodal transformer independently rediscovered the brain’s modality organization, assigning higher visual weight to occipital regions, higher auditory weight to temporal areas, variable modality weight to switching areas, and balanced weight to jointly predicted areas.

000 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/11/2025

7/n The circular belts of regions that alternate their dominant sensory modality over time(shades of red below) , and the axis of regions along the occipito-temporal lobe that jointly track both modalities(shades of blue below), together form a bow-and-arrow-like pattern.

000 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/11/2025

6/n PIP reveals an axis of regions that are jointly predicted more often than other cortical regions - they extend from the lateral occipital cortex to the temporal cortex

000 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/11/2025

5/n Analogously, the Performance Indication Paradigm (PIP) captures periods of statistically reliable performance by both modalities. This allows us to identify regions that are well predicted by both modalities for a substantial number of TRs.

000 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/11/2025

4/n DIP reveals a pair of “bows” of modality switching - one posterior bow encircling category-selective visual cortex and another anterior bow spanning dorso-lateral frontal areas (purple to pink in the plot are switching regions)

000 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/11/2025

3/n Using the Dominance Indication Paradigm (DIP), we capture periods of sustained dominance of one modality’s prediction performance over the other for an ROI. This allows us to identify regions that both switch modalities they encode and do so for a substantial number of TRs.

000 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/11/2025

2/n Using large-scale fMRI data collected while participants watched movies, we developed two computational approaches relying on prediction performance to analyze temporal dynamics of sensory processing across cortical regions

000 0

Subha Nawer Pushpita @pushpita1729.bsky.social · 07/11/2025

Excited to share our work on mechanisms of naturalistic audiovisual processing in the human brain 🧠🎬!! www.biorxiv.org/content/10.1...

biorxiv.org

965 0