Kohitij Kar
@kohitij.bsky.social
Assistant Professor, Canada Research Chair in Visual Neuroscience, @yorkuniversity, Visiting Scientist @MIT, Previously: Postdoc(@MIT), PhD (@RutgersU)
🌟New Preprint
Work led by @sabinemuzellec.bsky.social, in collab with
Yousif Alghetaa, and Simon Kornblith
We introduce a neurally validated strategy "MAPS" leveraging current ANNs and XAI tools to approximate human visual strategies. Also, validated against Bubbles!
arxiv.org/abs/2510.12141
Work led by @sabinemuzellec.bsky.social, in collab with
Yousif Alghetaa, and Simon Kornblith
We introduce a neurally validated strategy "MAPS" leveraging current ANNs and XAI tools to approximate human visual strategies. Also, validated against Bubbles!
arxiv.org/abs/2510.12141
October 15, 2025 at 1:32 AM
🌟New Preprint
Work led by @sabinemuzellec.bsky.social, in collab with
Yousif Alghetaa, and Simon Kornblith
We introduce a neurally validated strategy "MAPS" leveraging current ANNs and XAI tools to approximate human visual strategies. Also, validated against Bubbles!
arxiv.org/abs/2510.12141
Work led by @sabinemuzellec.bsky.social, in collab with
Yousif Alghetaa, and Simon Kornblith
We introduce a neurally validated strategy "MAPS" leveraging current ANNs and XAI tools to approximate human visual strategies. Also, validated against Bubbles!
arxiv.org/abs/2510.12141
🧠 New preprint!
Recognition-trained ANNs miss a key dimension of IT coding: memorability. Jointly optimizing for both recognition and memorability yielded more brain-like visual models.
Another step toward multi-goal modeling of the ventral stream.
www.biorxiv.org/content/10.1...
Recognition-trained ANNs miss a key dimension of IT coding: memorability. Jointly optimizing for both recognition and memorability yielded more brain-like visual models.
Another step toward multi-goal modeling of the ventral stream.
www.biorxiv.org/content/10.1...
October 9, 2025 at 4:01 PM
🧠 New preprint!
Recognition-trained ANNs miss a key dimension of IT coding: memorability. Jointly optimizing for both recognition and memorability yielded more brain-like visual models.
Another step toward multi-goal modeling of the ventral stream.
www.biorxiv.org/content/10.1...
Recognition-trained ANNs miss a key dimension of IT coding: memorability. Jointly optimizing for both recognition and memorability yielded more brain-like visual models.
Another step toward multi-goal modeling of the ventral stream.
www.biorxiv.org/content/10.1...
Two new preprints from our lab!
@sabinemuzellec.bsky.social leads work on reverse(Brain➡ ANN) predictivity showing gaps in forward(ANN ➡ Brain) metrics shorturl.at/bnoFl
@marenwehrheim.bsky.social leads work on facial expression emerging from shared not segregated neural subspaces. shorturl.at/QAIl4
@sabinemuzellec.bsky.social leads work on reverse(Brain➡ ANN) predictivity showing gaps in forward(ANN ➡ Brain) metrics shorturl.at/bnoFl
@marenwehrheim.bsky.social leads work on facial expression emerging from shared not segregated neural subspaces. shorturl.at/QAIl4
August 28, 2025 at 7:17 PM
Two new preprints from our lab!
@sabinemuzellec.bsky.social leads work on reverse(Brain➡ ANN) predictivity showing gaps in forward(ANN ➡ Brain) metrics shorturl.at/bnoFl
@marenwehrheim.bsky.social leads work on facial expression emerging from shared not segregated neural subspaces. shorturl.at/QAIl4
@sabinemuzellec.bsky.social leads work on reverse(Brain➡ ANN) predictivity showing gaps in forward(ANN ➡ Brain) metrics shorturl.at/bnoFl
@marenwehrheim.bsky.social leads work on facial expression emerging from shared not segregated neural subspaces. shorturl.at/QAIl4
We are excited to share our latest research at #vss2025 @vssmtg.bsky.social and learn about all things vision! Please come to our talks and posters. Looking forward to all your feedback!
May 16, 2025 at 1:48 PM
We are excited to share our latest research at #vss2025 @vssmtg.bsky.social and learn about all things vision! Please come to our talks and posters. Looking forward to all your feedback!
Our lab had a great time sharing our work at the NETI Workshop at @utaustin.bsky.social. Lots of new, exciting work is going on in systems neuroscience!
April 30, 2025 at 3:38 PM
Our lab had a great time sharing our work at the NETI Workshop at @utaustin.bsky.social. Lots of new, exciting work is going on in systems neuroscience!
If you are at @cosynemeeting.bsky.social #cosyne2025 , please check out two posters from our lab today, one by Mualla (2-62), and the other by Jean de Dieu (2-127).
March 28, 2025 at 3:31 PM
If you are at @cosynemeeting.bsky.social #cosyne2025 , please check out two posters from our lab today, one by Mualla (2-62), and the other by Jean de Dieu (2-127).
(3/3) Many ANN models of untrained IT show similar changes at the level of ANN-IT (which is typically not at the very end of the ANN hierarchy) with additional category training (implemented via varied learning approaches) that also generalize to category orthogonal representational shifts.
December 30, 2024 at 5:58 PM
(3/3) Many ANN models of untrained IT show similar changes at the level of ANN-IT (which is typically not at the very end of the ANN hierarchy) with additional category training (implemented via varied learning approaches) that also generalize to category orthogonal representational shifts.
(2/3) Untrained animals can see objects but can’t attach labels—so we don’t expect ventral stream to reformat with learning. But does it change at all, or everything happens downstream? Trained vs. untrained monkeys show minor but robust IT changes -- not a complete reformat, but meaningful tweaks.
December 30, 2024 at 5:58 PM
(2/3) Untrained animals can see objects but can’t attach labels—so we don’t expect ventral stream to reformat with learning. But does it change at all, or everything happens downstream? Trained vs. untrained monkeys show minor but robust IT changes -- not a complete reformat, but meaningful tweaks.
This new article perfectly concludes my time & lessons in Jim’s lab.@JamesJDiCarlo and I propose +review SMART models of object recognition
✅ Sensory computable
✅ Mechanistic
✅ Anatomically Referenced
✅ Testable
Coming in Annual Reviews 2024
Preprint: bit.ly/3tk7u8D
✅ Sensory computable
✅ Mechanistic
✅ Anatomically Referenced
✅ Testable
Coming in Annual Reviews 2024
Preprint: bit.ly/3tk7u8D
December 12, 2023 at 3:20 AM
This new article perfectly concludes my time & lessons in Jim’s lab.@JamesJDiCarlo and I propose +review SMART models of object recognition
✅ Sensory computable
✅ Mechanistic
✅ Anatomically Referenced
✅ Testable
Coming in Annual Reviews 2024
Preprint: bit.ly/3tk7u8D
✅ Sensory computable
✅ Mechanistic
✅ Anatomically Referenced
✅ Testable
Coming in Annual Reviews 2024
Preprint: bit.ly/3tk7u8D