bsky.app Open in urlscan Pro
3.13.124.211  Public Scan

Form analysis
0 forms found in the DOM

Text Content

Follow

Martin Schrimpf
@mschrimpf.bsky.social
2.2K followers52 following
25 posts
NeuroAI Prof @EPFL 🇨🇭. ML + Neuro 🤖🧠. Brain-Score, CORnet, Vision, Language.
Previously: PhD @MIT, ML @Salesforce, Neuro @HarvardMed, & co-founder
@Integreat. go.epfl.ch/NeuroAI



Posts

Media


Reposted by
Martin Schrimpf
‪Badr AlKhamissi‬ ‪@bkhmsi.bsky.social‬
·
10d
🚨 New Paper! Can neuroscience localizers uncover brain-like functional
specializations in LLMs? 🧠🤖 Yes! We analyzed 18 LLMs and found units mirroring
the brain's language, theory of mind, and multiple demand networks! w/
@gretatuckute.bsky.social
,
@abosselut.bsky.social
,
@mschrimpf.bsky.social
🧵👇

2
27
84

‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
1mo
One month left to apply to this faculty position in neuromodulation:
www.epfl.ch/about/workin... The EPFL NeuroX institute is a truly unique
environment combining neuro-technology, -science, and -computation with the
ultimate goal of applications in medicine. Come be our colleague!
#neuro
#tech
🧠
Faculty Position in Neuromodulation
The Schools of Engineering (STI), Life Sciences (SV) and Computer and
Communication Sciences (IC) at EPFL invite applications for a tenure-track
assistant faculty position to hold the Medtronic chair ...
www.epfl.ch
5
9

‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
1mo
A few people asked why my slides looked jumbled up at the NIH BRAIN workshop --
I promise they were fine when I sent them in last week 😄 Copy of them here for
reference: mschrimpf.altervista.org/wp-content/u...
#NeuroAI
🧠🤖🧪
https://mschrimpf.altervista.org/wp-content/uploads/2024/11/20241112-NIH.pdf
mschrimpf.altervista.org
1

‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
1mo
For a while, gains in AI have translated into gains in modeling the brain. We
test if that will continue to be the case with recent advances in scaling.
Surprisingly we get mixed results: while increased scale advances model
alignment to behavior, neural alignment saturates: arxiv.org/abs/2411.05712

1
15
32

View full thread
‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
1mo
Beyond the distinction of behavioral and neural alignment, there is actually a
striking gradient: scaling helps more for higher-level visual regions but only
barely for early visual areas.

1
2
5

‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
1mo
To me this suggests that to really model the brain, we (
#NeuroAI
) need to start building our own models that respect intricacies of brain
processing and push for higher and higher neural alignment
1
1
9

‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
1mo
NIH NeuroAI workshop happening today! Our instructions as speakers were to
imagine what the next 10 years would bring for the field. Will be streamed.
n4solutionsllc.com/brainneuroai/
#BRAIN
#NeuroAI
#neuroscience
#AI
BRAIN NeuroAI Workshop – Home - N4 Solutions
We would like to invite you to register at the upcoming NIH BRAIN NeuroAI
Workshop to be held on November 12th & 13th, 2024 at the NIH Campus in Bethesda,
Maryland.
n4solutionsllc.com
3
6

‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
1mo
How can we train LLMs with <100M words? In our
#BabyLM
paper, we introduce a new language+vision self-synthesis training recipe to
tackle this question: Our model learns over 4 phases -- most crucially
self-captioning unseen images to generate synthetic language data
arxiv.org/abs/2411.00828

1
7

‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
1mo
Applications now open for the Summer@EPFL program summer.epfl.ch -- 3-month
fellowship for Bachelor/Master students to immerse yourself in research

1

‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
2mo
Functional responses in the brain to linguistic inputs are spatially organized
-- but why? We show that a simple smoothness loss added to language model
training explains a range of topographic phenomena in neuroscience:
arxiv.org/abs/2410.11516
#NeuroAI
#neuroscience
#language

6
14

‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
6mo
In 2021 we were surprised to find that untrained language models are already
decent predictors of activity in the human language system
(doi.org/10.1073/pnas...). In arxiv.org/abs/2406.15109, we identify the core
components underlying the alignment of untrained models: tokenization and
aggregation
The neural architecture of language: Integrative modeling converges on
predictive processing | PNAS
The neuroscience of perception has recently been revolutionized with an
integrative modeling approach in which computation, brain function, and beh...
doi.org
1
4
10

‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
6mo
With these findings, we built a simple untrained network "SUMA" with
state-of-the-art alignment to brain and behavioral data -- this feature encoder
provides representations that are then useful for efficient language modeling 2/

1

2

‪Martin Schrimpf‬ ‪@mschrimpf.bsky.social‬
·
6mo
Directly mapping our model onto the brain, these results characterize the human
language system as a generic feature encoder that aggregates incoming sensory
representations for downstream use. If you disagree we hope you consider
breaking our model (soon on Brain-Score/Github)
2

Reposted by
Martin Schrimpf
‪Greta Tuckute‬ ‪@gretatuckute.bsky.social‬
·
1/3/2024
1/ Our paper is published in Nature Human Behavior! tinyurl.com/p28jy3bx Driving
and suppressing brain activity in the human language network with model
(GPT)-selected stimuli With A. Sathe, S. Srikant, M. Taliaferro, M. Wang,
@mschrimpf.bsky.social
, K. Kay,
@evfedorenko.bsky.social

1
4
23


End of feed

Join the conversation
Sign up
Sign in
English
EnglishAragonés – AragoneseAsturianu – AsturianCatalà – CatalanDeutsch –
GermanEnglish (UK)Español – SpanishSuomi – FinnishFrançais – FrenchGaeilge –
IrishGalego – Galicianहिंदी – Hindimagyar – HungarianBahasa Indonesia –
IndonesianItaliano – Italian日本語 – Japaneseភាសាខ្មែរ – Khmer한국어 –
KoreanNederlands – Dutchनेपाली – NepaliPolski – PolishPortuguês (BR) –
Portuguese (BR)Română – RomanianРусский – Russianภาษาไทย – ThaiTürkçe –
TurkishУкраїнська – UkrainianTiếng Việt – Vietnamese简体中文 – Simplified
Chinese繁體中文 – Traditional Chinese粵文 – Cantonese
Privacy • Terms • Help



JAVASCRIPT REQUIRED

This is a heavily interactive web application, and JavaScript is required.
Simple HTML interfaces are possible, but that is not what this is.

Learn more about Bluesky at bsky.social and atproto.com.


PROFILE

Martin Schrimpf

mschrimpf.bsky.social

did:plc:4ukss6526aqby24q5w2yds76

NeuroAI Prof @EPFL 🇨🇭. ML + Neuro 🤖🧠. Brain-Score, CORnet, Vision, Language.
Previously: PhD @MIT, ML @Salesforce, Neuro @HarvardMed, & co-founder
@Integreat. go.epfl.ch/NeuroAI