Marianne de Heer Kloots
@mdhk.net
Linguist in AI & CogSci 🧠👩💻🤖
PhD student @illc-uva.bsky.social
🌐 https://mdhk.net/
🐘 https://scholar.social/@mdhk
🐦 https://twitter.com/mariannedhk
PhD student @illc-uva.bsky.social
🌐 https://mdhk.net/
🐘 https://scholar.social/@mdhk
🐦 https://twitter.com/mariannedhk
Pinned
Marianne de Heer Kloots
@mdhk.net
· Jul 8
✨ Do current neural speech models show human-like linguistic biases in speech perception?
We took inspiration from classic phonetic categorization experiments to explore where sensitivity to phonotactic context emerges in Wav2Vec2 models 🔍
(w/ @wzuidema.bsky.social)
📑 arxiv.org/abs/2407.03005
⬇️
We took inspiration from classic phonetic categorization experiments to explore where sensitivity to phonotactic context emerges in Wav2Vec2 models 🔍
(w/ @wzuidema.bsky.social)
📑 arxiv.org/abs/2407.03005
⬇️
LLMs are everywhere, as are companies advertising LLM-powered products by their “beyond-human” performance levels.
But how do LLMs *actually* fare against humans across the board? Join our effort to organize empirical evidence from the field into a systematic and publicly accessible overview ⬇️
But how do LLMs *actually* fare against humans across the board? Join our effort to organize empirical evidence from the field into a systematic and publicly accessible overview ⬇️
Ever feel like it's too hard to keep track of what LLMs cannot do as well as humans? We're making your life easier over at: what-llms-can-not-do.github.io
We're compiling a list of papers testing the abilities of LLMs against humans. Check it out! And you can help contribute too!
We're compiling a list of papers testing the abilities of LLMs against humans. Check it out! And you can help contribute too!
What LLMs Can(not) Do
A living survey of benchmarks that compare large language models with humans.
what-llms-can-not-do.github.io
September 17, 2026 at 2:42 PM
LLMs are everywhere, as are companies advertising LLM-powered products by their “beyond-human” performance levels.
But how do LLMs *actually* fare against humans across the board? Join our effort to organize empirical evidence from the field into a systematic and publicly accessible overview ⬇️
But how do LLMs *actually* fare against humans across the board? Join our effort to organize empirical evidence from the field into a systematic and publicly accessible overview ⬇️
Reposted by Marianne de Heer Kloots
Ever feel like it's too hard to keep track of what LLMs cannot do as well as humans? We're making your life easier over at: what-llms-can-not-do.github.io
We're compiling a list of papers testing the abilities of LLMs against humans. Check it out! And you can help contribute too!
We're compiling a list of papers testing the abilities of LLMs against humans. Check it out! And you can help contribute too!
What LLMs Can(not) Do
A living survey of benchmarks that compare large language models with humans.
what-llms-can-not-do.github.io
September 17, 2026 at 12:24 PM
Ever feel like it's too hard to keep track of what LLMs cannot do as well as humans? We're making your life easier over at: what-llms-can-not-do.github.io
We're compiling a list of papers testing the abilities of LLMs against humans. Check it out! And you can help contribute too!
We're compiling a list of papers testing the abilities of LLMs against humans. Check it out! And you can help contribute too!
Reposted by Marianne de Heer Kloots
I fear that placing too much emphasis on a specific interpretation of dimensionality, or treating dimensionality as an end-all quantification of some aspect of neural computation, may lead us down the wrong path, writes @mattperich.bsky.social.
#neuroskyence
www.thetransmitter.org/neural-dynam...
#neuroskyence
www.thetransmitter.org/neural-dynam...
Dimensionality—neuroscience’s red herring?
Placing too much emphasis on a specific interpretation of dimensionality may lead neuroscience down the wrong path.
www.thetransmitter.org
September 7, 2026 at 6:42 PM
I fear that placing too much emphasis on a specific interpretation of dimensionality, or treating dimensionality as an end-all quantification of some aspect of neural computation, may lead us down the wrong path, writes @mattperich.bsky.social.
#neuroskyence
www.thetransmitter.org/neural-dynam...
#neuroskyence
www.thetransmitter.org/neural-dynam...
The reading group I've been organizing is entering its 4th academic year of existence!
We'll have our first meeting of this year this Thursday, with @gretatuckute.bsky.social and @klemenkotar.bsky.social presenting recent work on their AuriStream model 🧠🔊
dnn-speech.github.io
We'll have our first meeting of this year this Thursday, with @gretatuckute.bsky.social and @klemenkotar.bsky.social presenting recent work on their AuriStream model 🧠🔊
dnn-speech.github.io
(D)NN-Speech Reading Group
We are a reading group meeting regularly to discuss papers on speech-based deep learning models and their use in modelling human speech processing and acquisition!
dnn-speech.github.io
September 14, 2026 at 8:45 AM
The reading group I've been organizing is entering its 4th academic year of existence!
We'll have our first meeting of this year this Thursday, with @gretatuckute.bsky.social and @klemenkotar.bsky.social presenting recent work on their AuriStream model 🧠🔊
dnn-speech.github.io
We'll have our first meeting of this year this Thursday, with @gretatuckute.bsky.social and @klemenkotar.bsky.social presenting recent work on their AuriStream model 🧠🔊
dnn-speech.github.io
Reposted by Marianne de Heer Kloots
look at my new babylm paper! arxiv.org/abs/2609.11870
basically, i initialize token embeddings with representations from an image encoder rather than randomly and then train text-only as usual. kind of a visual demonstration to start off the word learning process
basically, i initialize token embeddings with representations from an image encoder rather than randomly and then train text-only as usual. kind of a visual demonstration to start off the word learning process
September 11, 2026 at 10:36 AM
look at my new babylm paper! arxiv.org/abs/2609.11870
basically, i initialize token embeddings with representations from an image encoder rather than randomly and then train text-only as usual. kind of a visual demonstration to start off the word learning process
basically, i initialize token embeddings with representations from an image encoder rather than randomly and then train text-only as usual. kind of a visual demonstration to start off the word learning process
Reposted by Marianne de Heer Kloots
A group of 25 Fields Medalists, including myself, have made a joint declaration on Math and AI: mathandai.org . We welcome additional signatories. See also this article in the Economist announcing the declaration: www.economist.com/science-and-...
Declaration — Math and AI
Read the declaration and add your name.
mathandai.org
September 11, 2026 at 5:40 PM
A group of 25 Fields Medalists, including myself, have made a joint declaration on Math and AI: mathandai.org . We welcome additional signatories. See also this article in the Economist announcing the declaration: www.economist.com/science-and-...
I’m presenting this today at #CLIN36 in Brussels! See you at poster 25 in the afternoon session 🎉
Let’s study learning trajectories in self-supervised speech models! 🔊 Do they reflect the hierarchical organization of spoken language?
We have analyzed a lot of training checkpoints to find out 🌠
Preprint: arxiv.org/abs/2604.02043
⬇️
We have analyzed a lot of training checkpoints to find out 🌠
Preprint: arxiv.org/abs/2604.02043
⬇️
September 11, 2026 at 7:26 AM
I’m presenting this today at #CLIN36 in Brussels! See you at poster 25 in the afternoon session 🎉
Reposted by Marianne de Heer Kloots
🤖🧠NEW PAPER🧠🤖
(The result of an 8-year project!)
LLMs seem very different from symbolic systems. Yet LLMs excel in symbolic domains (e.g., language/code/math). How do they do it?
Our finding: LLM representations have implicit symbolic structure!
Link in thread ⬇️
1/n
(The result of an 8-year project!)
LLMs seem very different from symbolic systems. Yet LLMs excel in symbolic domains (e.g., language/code/math). How do they do it?
Our finding: LLM representations have implicit symbolic structure!
Link in thread ⬇️
1/n
![Overview of the paper.
Title: The Emergent Symbolic Structure of Artificial Neural Networks
Authors: Tom McCoy, Paul Soulos, Tal Linzen, Paul Smolensky
Left: Neural networks encode information in vectors (there is then an image of a vector), yet they excel at tasks long thought to require symbolic structure (there is then an image of a symbolic representation, specifically a syntax tree). How do LLMs do it?
Right: We find that LLM representations can be closely approximated with symbolic structures. This approximation lets us edit the structure of an LLM’s output by editing the structure of its internal representations, as shown. There is then an image of two edits to LLMs. In the first one, the original input is 3 + 6 * 8, with an answer of 51. But if we swap the positions of the 3 and the 6, the output becomes 30. In the second one, the original input is a Python command repeating the list [Z, U] three times, producing [Z, U, Z, U, Z, U]. But if we edit the input in a way that adds a Q at the end of the input, the output becomes [Z, U, Q, Z, U, Q, Z, U, Q].](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:bcsvlv7ubeagj5virdvatadd/bafkreic3zm2rpc4vtqrz2t3y4vcl3t2owp2sxay42vuh5zm5zkg4jhr64e)
September 1, 2026 at 7:21 PM
🤖🧠NEW PAPER🧠🤖
(The result of an 8-year project!)
LLMs seem very different from symbolic systems. Yet LLMs excel in symbolic domains (e.g., language/code/math). How do they do it?
Our finding: LLM representations have implicit symbolic structure!
Link in thread ⬇️
1/n
(The result of an 8-year project!)
LLMs seem very different from symbolic systems. Yet LLMs excel in symbolic domains (e.g., language/code/math). How do they do it?
Our finding: LLM representations have implicit symbolic structure!
Link in thread ⬇️
1/n
Reposted by Marianne de Heer Kloots
Are brains and artificial neural networks converging onto universal representations?
There is a seductive idea making the rounds in NeuroAI / machine learning: train systems well enough, and they all converge on the same representation of reality (i.e. a unique world model).
We have thoughts™
1/n
There is a seductive idea making the rounds in NeuroAI / machine learning: train systems well enough, and they all converge on the same representation of reality (i.e. a unique world model).
We have thoughts™
1/n
The Umwelt Representation Hypothesis: rethinking Universality
Recent studies reveal striking representational alignment between artificial neural
networks (ANNs) and biological brains, leading to proposals that all sufficiently
capable systems converge on univer...
www.cell.com
August 28, 2026 at 2:59 PM
Are brains and artificial neural networks converging onto universal representations?
There is a seductive idea making the rounds in NeuroAI / machine learning: train systems well enough, and they all converge on the same representation of reality (i.e. a unique world model).
We have thoughts™
1/n
There is a seductive idea making the rounds in NeuroAI / machine learning: train systems well enough, and they all converge on the same representation of reality (i.e. a unique world model).
We have thoughts™
1/n
Reposted by Marianne de Heer Kloots
New preprint 🐝🧵 arxiv.org/abs/2608.25779
Bee species signal food via their waggle dance in different ways. Species with horizontal combs point straight at food. Species with vertical combs dance relative to gravity instead. How could a population evolve from one system to the other?
Bee species signal food via their waggle dance in different ways. Species with horizontal combs point straight at food. Species with vertical combs dance relative to gravity instead. How could a population evolve from one system to the other?
August 28, 2026 at 7:10 AM
New preprint 🐝🧵 arxiv.org/abs/2608.25779
Bee species signal food via their waggle dance in different ways. Species with horizontal combs point straight at food. Species with vertical combs dance relative to gravity instead. How could a population evolve from one system to the other?
Bee species signal food via their waggle dance in different ways. Species with horizontal combs point straight at food. Species with vertical combs dance relative to gravity instead. How could a population evolve from one system to the other?
Reposted by Marianne de Heer Kloots
New perspective piece out in @cp-neuron.bsky.social with Zaid Zada, @adelegoldberg.bsky.social, and Uri Hasson! We try to articulate some of our excitement about LLMs and discuss what kinds of insights they might provide into the neural computations supporting natural language in the human brain.
August 18, 2026 at 1:04 AM
New perspective piece out in @cp-neuron.bsky.social with Zaid Zada, @adelegoldberg.bsky.social, and Uri Hasson! We try to articulate some of our excitement about LLMs and discuss what kinds of insights they might provide into the neural computations supporting natural language in the human brain.
Reposted by Marianne de Heer Kloots
New Blog post: Which Data Repository Should you Use?
In light of OSF closing down, I compare Zenodo, Dataverse, ResearchBox, PsychArchive, and local repositories on six important dimensions. If you want to know which to pick: It depends!
daniellakens.blogspot.com/2026/08/whic...
In light of OSF closing down, I compare Zenodo, Dataverse, ResearchBox, PsychArchive, and local repositories on six important dimensions. If you want to know which to pick: It depends!
daniellakens.blogspot.com/2026/08/whic...
Which Data Repository Should You Use?
The Center for Open Science has announced that from November 16, 2026, no new projects can be created on the Open Science Framework. After F...
daniellakens.blogspot.com
August 15, 2026 at 4:24 PM
New Blog post: Which Data Repository Should you Use?
In light of OSF closing down, I compare Zenodo, Dataverse, ResearchBox, PsychArchive, and local repositories on six important dimensions. If you want to know which to pick: It depends!
daniellakens.blogspot.com/2026/08/whic...
In light of OSF closing down, I compare Zenodo, Dataverse, ResearchBox, PsychArchive, and local repositories on six important dimensions. If you want to know which to pick: It depends!
daniellakens.blogspot.com/2026/08/whic...
Reposted by Marianne de Heer Kloots
I've been unsettled lately when reading messages and papers. It feels like I'm dissociating. Everything seems a bit alien, even if it's completely human. I've had a realization:
When our simulations finally exited the Uncanny Valley, they brought the Uncanny with them.
When our simulations finally exited the Uncanny Valley, they brought the Uncanny with them.
Life on the Uncanny Precipice | Naomi Saphra
We were wrong about the Uncanny Valley.
nsaphra.net
August 9, 2026 at 6:02 PM
I've been unsettled lately when reading messages and papers. It feels like I'm dissociating. Everything seems a bit alien, even if it's completely human. I've had a realization:
When our simulations finally exited the Uncanny Valley, they brought the Uncanny with them.
When our simulations finally exited the Uncanny Valley, they brought the Uncanny with them.
Reposted by Marianne de Heer Kloots
Today is the only day you can experience Ray Bradbury’s “There Will Come Soft Rains” on the day it is set.
Audio: archive.org/details/brad...
PDF: thephilosopher.net/bredberi/wp-...
Audio: archive.org/details/brad...
PDF: thephilosopher.net/bredberi/wp-...
August 4, 2026 at 10:51 AM
Today is the only day you can experience Ray Bradbury’s “There Will Come Soft Rains” on the day it is set.
Audio: archive.org/details/brad...
PDF: thephilosopher.net/bredberi/wp-...
Audio: archive.org/details/brad...
PDF: thephilosopher.net/bredberi/wp-...
Reposted by Marianne de Heer Kloots
Let’s study learning trajectories in self-supervised speech models! 🔊 Do they reflect the hierarchical organization of spoken language?
We have analyzed a lot of training checkpoints to find out 🌠
Preprint: arxiv.org/abs/2604.02043
⬇️
We have analyzed a lot of training checkpoints to find out 🌠
Preprint: arxiv.org/abs/2604.02043
⬇️
June 30, 2026 at 11:01 AM
Let’s study learning trajectories in self-supervised speech models! 🔊 Do they reflect the hierarchical organization of spoken language?
We have analyzed a lot of training checkpoints to find out 🌠
Preprint: arxiv.org/abs/2604.02043
⬇️
We have analyzed a lot of training checkpoints to find out 🌠
Preprint: arxiv.org/abs/2604.02043
⬇️
Reposted by Marianne de Heer Kloots
1/5 Over a decade of comparing deep neural networks to the human brain—but what have we actually learned? Our new @cp-trendscognsci.bsky.social Feature Review synthesizes a decade of brain–DNN comparisons, asking what they reveal about brain function across vision and language.
July 21, 2026 at 5:58 PM
1/5 Over a decade of comparing deep neural networks to the human brain—but what have we actually learned? Our new @cp-trendscognsci.bsky.social Feature Review synthesizes a decade of brain–DNN comparisons, asking what they reveal about brain function across vision and language.
Reposted by Marianne de Heer Kloots
Our ability to speak involves remarkable feats of motor sequencing that most of us take completely for granted. The new issue of @science.org has a fascinating essay by Sergey Stavisky, covering his team's use of brain-computer interfaces to restore speech for people with neurological injuries.
🧠🗣️🧪👇
🧠🗣️🧪👇
Regaining your voice
AI speech neuroprostheses can restore day-to-day communication after neurological injury
www.science.org
July 16, 2026 at 6:22 PM
Our ability to speak involves remarkable feats of motor sequencing that most of us take completely for granted. The new issue of @science.org has a fascinating essay by Sergey Stavisky, covering his team's use of brain-computer interfaces to restore speech for people with neurological injuries.
🧠🗣️🧪👇
🧠🗣️🧪👇
Reposted by Marianne de Heer Kloots
Our paper "Rarely categorical, highly separable representations along the cortical hierarchy" is now out in Nature! www.nature.com/articles/s41...
In this paper, we studied whether the brain is well-organized and interpretable, both globally and locally. (See thread below.)
In this paper, we studied whether the brain is well-organized and interpretable, both globally and locally. (See thread below.)
www.nature.com
July 15, 2026 at 3:28 PM
Our paper "Rarely categorical, highly separable representations along the cortical hierarchy" is now out in Nature! www.nature.com/articles/s41...
In this paper, we studied whether the brain is well-organized and interpretable, both globally and locally. (See thread below.)
In this paper, we studied whether the brain is well-organized and interpretable, both globally and locally. (See thread below.)
Reposted by Marianne de Heer Kloots
What's more nonsensical: smashing a pumpkin using a number, or growing flowers inside a sneeze?
Our paper on graded inconceivability is out now in Cognition!
Come for the cognitive science 🧠🔍, stay for the whimsy 🌼🧚!
🔗Journal link: bit.ly/gradedInconCog
Our paper on graded inconceivability is out now in Cognition!
Come for the cognitive science 🧠🔍, stay for the whimsy 🌼🧚!
🔗Journal link: bit.ly/gradedInconCog
Now out (for realz) in Cognition:
"People Make Graded Judgments About The Inconceivable"
(by Hu, Sosa, & me)
Free preprint: www.tomerullman.org/papers/grade...
Journal link: bit.ly/gradedInconCog
@jennhu.bsky.social
@cognitionjournal.bsky.social
"People Make Graded Judgments About The Inconceivable"
(by Hu, Sosa, & me)
Free preprint: www.tomerullman.org/papers/grade...
Journal link: bit.ly/gradedInconCog
@jennhu.bsky.social
@cognitionjournal.bsky.social
July 2, 2026 at 2:33 PM
What's more nonsensical: smashing a pumpkin using a number, or growing flowers inside a sneeze?
Our paper on graded inconceivability is out now in Cognition!
Come for the cognitive science 🧠🔍, stay for the whimsy 🌼🧚!
🔗Journal link: bit.ly/gradedInconCog
Our paper on graded inconceivability is out now in Cognition!
Come for the cognitive science 🧠🔍, stay for the whimsy 🌼🧚!
🔗Journal link: bit.ly/gradedInconCog
Now out in BBS, as commentary on @futrell.bsky.social & @kmahowald.bsky.social's "How linguistics learned to stop worrying and love the language models"!
Humans learn much of spoken language structure from speech (not text), & we can study models that do the same.
www.cambridge.org/core/journal...
Humans learn much of spoken language structure from speech (not text), & we can study models that do the same.
www.cambridge.org/core/journal...
July 3, 2026 at 10:03 AM
Now out in BBS, as commentary on @futrell.bsky.social & @kmahowald.bsky.social's "How linguistics learned to stop worrying and love the language models"!
Humans learn much of spoken language structure from speech (not text), & we can study models that do the same.
www.cambridge.org/core/journal...
Humans learn much of spoken language structure from speech (not text), & we can study models that do the same.
www.cambridge.org/core/journal...
Reposted by Marianne de Heer Kloots
The full BBS treatment from me and @futrell.bsky.social on "How linguistics learned to stop worrying and love the LMs" is now out, with all the commentaries and our response. If you "Save PDF", it will give you the whole target article + commentary + response pdf:
www.cambridge.org/core/journal...
www.cambridge.org/core/journal...
How linguistics learned to stop worrying and love the language models | Behavioral and Brain Sciences | Cambridge Core
How linguistics learned to stop worrying and love the language models - Volume 49
www.cambridge.org
July 2, 2026 at 3:07 PM
The full BBS treatment from me and @futrell.bsky.social on "How linguistics learned to stop worrying and love the LMs" is now out, with all the commentaries and our response. If you "Save PDF", it will give you the whole target article + commentary + response pdf:
www.cambridge.org/core/journal...
www.cambridge.org/core/journal...
Reposted by Marianne de Heer Kloots
Linguists should learn to love speech-based deep learning models www.cambridge.org/core/journal... "Once the bottleneck of text can be replaced...modelling more human-like linguistic processes..[& handle] the vast majority of local & regional language varieties that are rarely or never written down"
https://www.cambridge.org/core/journals/behavioral-and-brain-sciences/article/linguists-should-learn-to-love-speechbased-deep-learning-models/3049F8D55B9FE1C1246819A1D0A69C24"Once
July 1, 2026 at 10:25 PM
Linguists should learn to love speech-based deep learning models www.cambridge.org/core/journal... "Once the bottleneck of text can be replaced...modelling more human-like linguistic processes..[& handle] the vast majority of local & regional language varieties that are rarely or never written down"
Let’s study learning trajectories in self-supervised speech models! 🔊 Do they reflect the hierarchical organization of spoken language?
We have analyzed a lot of training checkpoints to find out 🌠
Preprint: arxiv.org/abs/2604.02043
⬇️
We have analyzed a lot of training checkpoints to find out 🌠
Preprint: arxiv.org/abs/2604.02043
⬇️
June 30, 2026 at 11:01 AM
Let’s study learning trajectories in self-supervised speech models! 🔊 Do they reflect the hierarchical organization of spoken language?
We have analyzed a lot of training checkpoints to find out 🌠
Preprint: arxiv.org/abs/2604.02043
⬇️
We have analyzed a lot of training checkpoints to find out 🌠
Preprint: arxiv.org/abs/2604.02043
⬇️
