Pinned Tweet
In a meta-analysis of 210 biomedical AI studies that statistically compared models under cross-validation, 97% used invalid statistical tests.
Here's our new preprint doi.org/10.64898/2026.… led by @tianchuzeng @kkli20111 @ZShaoshi @ten_photos 1/N
Here's our new preprint doi.org/10.64898/2026.… led by @tianchuzeng @kkli20111 @ZShaoshi @ten_photos 1/N
9
95
274
122,552
Potential misconduct in a high profile fMRI foundation model study: neuroskepticbsky.substack.com/p/the-sherlock…
3
19
67
12,820
Thomas Yeo retweeted
🚨 New preprint! 🧵
Brain regions 🧠 integrate information over different timescales ⏱️: short in sensorimotor cortex, long in association cortex. But do regions change their timescales when we engage in a task? And how does that flexibility develop? 👇
biorxiv.org/content/10.648…
Brain regions 🧠 integrate information over different timescales ⏱️: short in sensorimotor cortex, long in association cortex. But do regions change their timescales when we engage in a task? And how does that flexibility develop? 👇
biorxiv.org/content/10.648…
2
22
104
6,234
Thomas Yeo retweeted
New collaborative work with @DrYuZhang is out now in Nature Mental Health! 🎉 Our DeCoDE framework two brain–behavior dimensions 🧠 and three biotypes in youth with #ADHD and #anxiety disorders, offering new insights into psychiatric heterogeneity and early risk identification.
🚀 Excited to share our new paper in Nature Mental Health!
🧠 Our Brain-Behavior AI framework identifies 3 transdiagnostic biotypes in youth with ADHD and anxiety.
🔎 Signatures emerge even before diagnosis, offering promise for early risk identification.
#AI #MentalHealth
0
6
22
1,427
Thomas Yeo retweeted
🧠 New in Aperture Neuro:
A new voxel-wise framework combines fMRI connectivity with diffusion-derived structural graphs, offering spatially resolved maps of how local white-matter architecture may support functional relationships.
doi.org/10.52294/001c.…
A new voxel-wise framework combines fMRI connectivity with diffusion-derived structural graphs, offering spatially resolved maps of how local white-matter architecture may support functional relationships.
doi.org/10.52294/001c.…
1
5
15
1,511
Thomas Yeo retweeted
Should Europe leave frontier AI to the Americans and the Chinese?
The idea has gained ground in Europe. It is a fundamental mistake.
Build or surrender: for now, the choice is still ours.
Full article below. x.com/i/article/2110…
The idea has gained ground in Europe. It is a fundamental mistake.
Build or surrender: for now, the choice is still ours.
Full article below. x.com/i/article/2110…
268
765
3,876
440,339
Thomas Yeo retweeted
Some people are confused why mathematicians say OAI's drop was the "death" of a field, or figure that it's death insomuch as a bunch of university professors lost their overpaid hobby of proving theorems. That's a misunderstanding: the damage's real.
Blog: blog.ohaithe.re/post/829972087…
Blog: blog.ohaithe.re/post/829972087…
82
112
778
186,581
Thomas Yeo retweeted
Imaging Neuroscience: The First 3 Years—Editorial
Thanks to our Editorial Board, reviewers, and authors, who trusted that this would work, and moved with us to the new journal. Its success belongs to that community.
doi.org/10.1162/IMAG.a…
Thanks to our Editorial Board, reviewers, and authors, who trusted that this would work, and moved with us to the new journal. Its success belongs to that community.
doi.org/10.1162/IMAG.a…
1
14
39
2,773
Thomas Yeo retweeted
2/11 We fit a biophysically plausible neural mass model to each person’s resting-state fMRI to estimate a putative in vivo E/I ratio marker. In prior work, it tracked pharmacologically induced GABAergic inhibition and declined with age in youth: doi.org/10.1073/pnas.2…
1
1
1
345
Thomas Yeo retweeted
1/11 Our new paper is out in Alzheimer’s & Dementia! 🚨
Does cortical excitation–inhibition (E/I) imbalance link amyloid pathophysiology to memory? We tested this in non-demented older adults (cognitively normal + MCI) across two independent cohorts. 🧵👇
doi.org/10.1002/alz.71…
Does cortical excitation–inhibition (E/I) imbalance link amyloid pathophysiology to memory? We tested this in non-demented older adults (cognitively normal + MCI) across two independent cohorts. 🧵👇
doi.org/10.1002/alz.71…
1
9
28
1,642
Thomas Yeo retweeted
[1/4] We mapped and fractionated large-scale brain networks in the human subcortex using NASCAR, a method built to untangle overlapping, correlated brain networks, recovering structure that conventional methods can't resolve.
Paper: doi.org/10.1073/pnas.2…
Paper: doi.org/10.1073/pnas.2…
3
15
51
3,384
Thomas Yeo retweeted
Check out our new open data resource on myelin sensitive imaging! Includes many image types, test retest data, and fully processed derivatives. Massive congrats to @StevenMeisler Taylor Salo and a superb team!!!!
There are many MRI approaches proposed to study myelin. But which ones are reliable, how do they relate to each other, and how do they behave in gray vs white matter?
In our new preprint, we introduce Myelin Imaging Reliability and ReprOducibility Resource (MIRROR) 🧠🪞
1
6
36
2,502
Thomas Yeo retweeted
There are many MRI approaches proposed to study myelin. But which ones are reliable, how do they relate to each other, and how do they behave in gray vs white matter?
In our new preprint, we introduce Myelin Imaging Reliability and ReprOducibility Resource (MIRROR) 🧠🪞
In our new preprint, we introduce Myelin Imaging Reliability and ReprOducibility Resource (MIRROR) 🧠🪞
1
3
18
2,840
Thomas Yeo retweeted
1/ 🚨 Just out in @NatureNeuro! 🧠
We profiled >6,000 features of brain dynamics across humans + monkeys, mice, fish and worms 🧑🐒🐒🐁🐟🪱 to find a shared neural endpoint of anaesthesia: spatiotemporal isolation of local activity 🧵
🔗 doi.org/10.1038/s41593…
We profiled >6,000 features of brain dynamics across humans + monkeys, mice, fish and worms 🧑🐒🐒🐁🐟🪱 to find a shared neural endpoint of anaesthesia: spatiotemporal isolation of local activity 🧵
🔗 doi.org/10.1038/s41593…
8
58
240
13,828
Thomas Yeo retweeted
Excited to share our new work now out in Nature Neuroscience!! We use a whole-brain framework to show that aging and Alzhiemer's disease affect the brain's functional organization along completely separate axes. Paper and thread below 👇
The first paper from my PhD is out in @NatureNeuro!
Aging and Alzheimer's disease both reshape how the brain communicates. But not in the same way: aging vs AD related alterations follow two distinct axes of brain organization
Paper → nature.com/articles/s4159…
🧵⬇️
0
16
54
3,710
Thomas Yeo retweeted
Maybe a bit late, but I wanted to comment on this: I think the AI boom made people forget what a PhD was meant for. Now it's going back to its original purpose: not short-term profit, but long-term research that might not have a practical use for another 10 or 20 years -> Backprop existed decades before the deep learning boom.
A PhD isn't for everyone, and that's okay. Industry may suit you better if you want to build something useful now, or academia if you believe in a problem industry doesn't care about yet. The AI boom made a PhD look like preparation for an industry job, but that was never its main purpose.
A PhD isn't for everyone, and that's okay. Industry may suit you better if you want to build something useful now, or academia if you believe in a problem industry doesn't care about yet. The AI boom made a PhD look like preparation for an industry job, but that was never its main purpose.
Students sometimes ask me if it still makes sense, in this accelerating age, to pursue a PhD in AI.
Perhaps counterintuitively, I think it's a great time to do so.
I wrote up some thoughts on this here: web.mit.edu/phillipi/www/w…
13
77
870
67,187
Thomas Yeo retweeted
Can an LM, starting from random init (!!), learn to generate all of its pretraining data?
Introducing Self-Play Pretraining with Zero Data. Two models start from random initialization: a generator proposes programs for a universal Turing machine and a learner trains on their on any real data, but see predictable scaling on natural datasets: zero-shot val loss on images, text, audio, and melodies decreases predictably with self-play compute. And the learner develops in-context learning capabilities.
A fun proof-of-concept, co-led with @AdityaCowsik and @KfirDolev and co-authors @gbruno_dl, @ANourya @noahdgoodman, and @YoavLevine.
Introducing Self-Play Pretraining with Zero Data. Two models start from random initialization: a generator proposes programs for a universal Turing machine and a learner trains on their on any real data, but see predictable scaling on natural datasets: zero-shot val loss on images, text, audio, and melodies decreases predictably with self-play compute. And the learner develops in-context learning capabilities.
A fun proof-of-concept, co-led with @AdityaCowsik and @KfirDolev and co-authors @gbruno_dl, @ANourya @noahdgoodman, and @YoavLevine.
85
459
3,562
752,717
Thomas Yeo retweeted
netneurotools: a trainee-oriented approach to network neuroscience | doi.org/10.1038/s41596…
Our lab’s internal toolkit for accomplishing everyday tasks in brain imaging @NatureProtocols ⤵️
Our lab’s internal toolkit for accomplishing everyday tasks in brain imaging @NatureProtocols ⤵️
2
31
133
8,342
Thomas Yeo retweeted
I wrote an essay about the imminent death of all science.
Go read it, it's fun!
doomslide.substack.com/p/mathematics-…
Go read it, it's fun!
doomslide.substack.com/p/mathematics-…
WE DID IT BOIS!!! WE KILLED MATH!!!
49
59
483
143,850