Jay Hennig
@jhennig.bsky.social
Computational neuroscientist interested in how we learn, and dad to twin boys
Asst prof at Baylor College of Medicine
https://www.henniglab.org/
Asst prof at Baylor College of Medicine
https://www.henniglab.org/
Pinned
Jay Hennig
@jhennig.bsky.social
· May 5
www.biorxiv.org
🥳 Super excited to share our lab's first preprint, led by @hungyun.bsky.social
"Distinct tasks engage shared neural subspaces in human hippocampus and anterior cingulate cortex"
www.biorxiv.org/content/10.6...
🧵(1/9)
"Distinct tasks engage shared neural subspaces in human hippocampus and anterior cingulate cortex"
www.biorxiv.org/content/10.6...
🧵(1/9)
Reposted by Jay Hennig
Granule cells predict future dopamine reward. Delayed climbing-fiber signals can reinforce the action that precedes it.
Our paper, led by Benjamin Filio, is out today in Nature Neuroscience.
www.nature.com/articles/s41...
Public PDF: rdcu.be/RscCNeXvkXAa
Our paper, led by Benjamin Filio, is out today in Nature Neuroscience.
www.nature.com/articles/s41...
Public PDF: rdcu.be/RscCNeXvkXAa
September 16, 2026 at 11:17 AM
Granule cells predict future dopamine reward. Delayed climbing-fiber signals can reinforce the action that precedes it.
Our paper, led by Benjamin Filio, is out today in Nature Neuroscience.
www.nature.com/articles/s41...
Public PDF: rdcu.be/RscCNeXvkXAa
Our paper, led by Benjamin Filio, is out today in Nature Neuroscience.
www.nature.com/articles/s41...
Public PDF: rdcu.be/RscCNeXvkXAa
Reposted by Jay Hennig
Here is a graph.
September 15, 2026 at 10:12 AM
Here is a graph.
Reposted by Jay Hennig
New preprint! 🧠 How do reward, risk & information preferences relate to psychopathology?
How do they vary across decisions about gains vs. losses?
1,954 people! Hundreds of economic choices each + transdiagnostic assessment of mental health symptoms online.
www.biorxiv.org/content/10.6...
How do they vary across decisions about gains vs. losses?
1,954 people! Hundreds of economic choices each + transdiagnostic assessment of mental health symptoms online.
www.biorxiv.org/content/10.6...
A dimensional link between gain-loss economic preferences and psychopathology
Human decision-makers differ in how they value and trade off reward, risk, and information when evaluating future gains and losses, and these preferences are central to well-being. However, how multi-...
www.biorxiv.org
September 15, 2026 at 7:28 PM
New preprint! 🧠 How do reward, risk & information preferences relate to psychopathology?
How do they vary across decisions about gains vs. losses?
1,954 people! Hundreds of economic choices each + transdiagnostic assessment of mental health symptoms online.
www.biorxiv.org/content/10.6...
How do they vary across decisions about gains vs. losses?
1,954 people! Hundreds of economic choices each + transdiagnostic assessment of mental health symptoms online.
www.biorxiv.org/content/10.6...
Reposted by Jay Hennig
Are virtual flies playing Beat Saber? SM64? Doom? No.
The new fly connectome is an incredible achievement! Less is going on in the sim videos than it seems. I'm stoked to see people engage with neuroscience though. My breakdown:
www.neuroai.science/p/are-flies-...
The new fly connectome is an incredible achievement! Less is going on in the sim videos than it seems. I'm stoked to see people engage with neuroscience though. My breakdown:
www.neuroai.science/p/are-flies-...
Are flies playing Beat Saber?
What a time to be alive!
www.neuroai.science
September 14, 2026 at 3:21 PM
Are virtual flies playing Beat Saber? SM64? Doom? No.
The new fly connectome is an incredible achievement! Less is going on in the sim videos than it seems. I'm stoked to see people engage with neuroscience though. My breakdown:
www.neuroai.science/p/are-flies-...
The new fly connectome is an incredible achievement! Less is going on in the sim videos than it seems. I'm stoked to see people engage with neuroscience though. My breakdown:
www.neuroai.science/p/are-flies-...
This is such cool work, especially the idea that on-manifold geometry is "imposed by the circuit's intrinsic dynamics" and not just stim encoding. We think something similar happens in human hippocampus—though tbf we haven't characterized the stimulus encoding dims yet
bsky.app/profile/jhen...
bsky.app/profile/jhen...
After a long journey through review, our paper is finally out in Nature Communications.
We started with a simple problem:
The visual cortex is strongly modulated by movement. So how can it represent the visual world without movement interfering with everything?
www.nature.com/articles/s41...
We started with a simple problem:
The visual cortex is strongly modulated by movement. So how can it represent the visual world without movement interfering with everything?
www.nature.com/articles/s41...
September 14, 2026 at 1:42 PM
This is such cool work, especially the idea that on-manifold geometry is "imposed by the circuit's intrinsic dynamics" and not just stim encoding. We think something similar happens in human hippocampus—though tbf we haven't characterized the stimulus encoding dims yet
bsky.app/profile/jhen...
bsky.app/profile/jhen...
Reposted by Jay Hennig
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-...
Reposted by Jay Hennig
We derived six laws of psychophysics from a single efficient-coding equation. The laws include Weber's law; scaling laws in visual working memory (which had been noted before, but with unclear theoretical justification); Wei & Stocker's law of human perception;
arthurprat.com/pdfs/Prat-Ca...
arthurprat.com/pdfs/Prat-Ca...
arthurprat.com
September 9, 2026 at 4:43 PM
We derived six laws of psychophysics from a single efficient-coding equation. The laws include Weber's law; scaling laws in visual working memory (which had been noted before, but with unclear theoretical justification); Wei & Stocker's law of human perception;
arthurprat.com/pdfs/Prat-Ca...
arthurprat.com/pdfs/Prat-Ca...
Reposted by Jay Hennig
UCSD Psychology is hiring!! We’re seeking a new assistant professor colleague who studies high-level human cognition, broadly construed. Please share and/or apply! apol-recruit.ucsd.edu/JPF04646
Assistant Professor in High-level Human Cognition
University of California, San Diego is hiring. Apply now!
apol-recruit.ucsd.edu
September 9, 2026 at 5:16 PM
UCSD Psychology is hiring!! We’re seeking a new assistant professor colleague who studies high-level human cognition, broadly construed. Please share and/or apply! apol-recruit.ucsd.edu/JPF04646
Reposted by Jay Hennig
New paper from the Neurosurgery Research Team at BCM! “Neural geometry in the human hippocampus enables generalization across spatial position and gaze” led by @assiachericoni.bsky.social! 🧵
www.cell.com/neuron/abstr...
www.cell.com/neuron/abstr...
Neural geometry in the human hippocampus enables generalization across spatial position and gaze
Chericoni et al. show that neurons in the human hippocampus track the positions of
multiple agents in a simple pursuit-based video game. They identified place codes
for each agent and for gaze. The co...
www.cell.com
September 9, 2026 at 3:34 PM
New paper from the Neurosurgery Research Team at BCM! “Neural geometry in the human hippocampus enables generalization across spatial position and gaze” led by @assiachericoni.bsky.social! 🧵
www.cell.com/neuron/abstr...
www.cell.com/neuron/abstr...
Reposted by Jay Hennig
How fast can mice learn complex mazes? With @mameister4.bsky.social we reveal memory, generalization, latent learning, and remote credit assignment of mice in the Manhattan Maze. Do mice born without a cortex or hippocampus have them too? Check it out!
www.biorxiv.org/content/10.6...
www.biorxiv.org/content/10.6...
Rapid spatial cognition in mice, with and without neocortex and hippocampus
Rapid learning, memory, and generalization are often attributed to circuits of the neocortex and hippocampus, but their specific role remains unclear. To examine these cognitive abilities together in ...
www.biorxiv.org
September 4, 2026 at 1:56 AM
How fast can mice learn complex mazes? With @mameister4.bsky.social we reveal memory, generalization, latent learning, and remote credit assignment of mice in the Manhattan Maze. Do mice born without a cortex or hippocampus have them too? Check it out!
www.biorxiv.org/content/10.6...
www.biorxiv.org/content/10.6...
Reposted by Jay Hennig
Terence Tao says: there's an emerging consensus on how to solve one of the $1,000,000 Millennium prize problems.
He says AI may help and lead to new insights - but it may solve the problem in a way that gives no new insights! Read what he wrote here:
mathstodon.xyz/@tao/1172078...
He says AI may help and lead to new insights - but it may solve the problem in a way that gives no new insights! Read what he wrote here:
mathstodon.xyz/@tao/1172078...
September 3, 2026 at 4:04 PM
Terence Tao says: there's an emerging consensus on how to solve one of the $1,000,000 Millennium prize problems.
He says AI may help and lead to new insights - but it may solve the problem in a way that gives no new insights! Read what he wrote here:
mathstodon.xyz/@tao/1172078...
He says AI may help and lead to new insights - but it may solve the problem in a way that gives no new insights! Read what he wrote here:
mathstodon.xyz/@tao/1172078...
Reposted by Jay Hennig
Pouring a cup of coffee looks trivial. It isn’t. Humans effortlessly control nonlinear liquid dynamics across different vessels and speeds—while robots still struggle. In our new preprint, we ask how the human motor system solves this everyday control problem. 1/5
September 3, 2026 at 3:17 PM
Pouring a cup of coffee looks trivial. It isn’t. Humans effortlessly control nonlinear liquid dynamics across different vessels and speeds—while robots still struggle. In our new preprint, we ask how the human motor system solves this everyday control problem. 1/5
Reposted by Jay Hennig
CSHL is hiring a new BioAI faculty member, part of a multi-year hiring process for the Foundations of the Future.
We need to fill a new BioAI building, come join!
www.cshl.edu/about-us/car...
We need to fill a new BioAI building, come join!
www.cshl.edu/about-us/car...
September 1, 2026 at 9:45 PM
CSHL is hiring a new BioAI faculty member, part of a multi-year hiring process for the Foundations of the Future.
We need to fill a new BioAI building, come join!
www.cshl.edu/about-us/car...
We need to fill a new BioAI building, come join!
www.cshl.edu/about-us/car...
Reposted by Jay Hennig
🤖🧠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 Jay Hennig
We are recruiting an Assistant Professor in Cog Neuro/Cog Psych to join us at Rice! Please spread the word! apply.interfolio.com/192080
Apply - Interfolio
{{$ctrl.$state.data.pageTitle}} - Apply - Interfolio
apply.interfolio.com
September 1, 2026 at 6:07 PM
We are recruiting an Assistant Professor in Cog Neuro/Cog Psych to join us at Rice! Please spread the word! apply.interfolio.com/192080
Lesson learned: If you want double spacing in Overleaf/latex just like MS Word, \doublespacing is not it! You want:
\usepackage{setspace}
\setstretch{2}
\usepackage{setspace}
\setstretch{2}
Just realized that creating a double-spaced document in Latex gives you more lines-per-page than double-spacing in Word. If a grant says your application must be "8 pages double-spaced"...will I get penalized if I didn't use Word's stricter definition? 😬😬😬
August 31, 2026 at 9:29 PM
Lesson learned: If you want double spacing in Overleaf/latex just like MS Word, \doublespacing is not it! You want:
\usepackage{setspace}
\setstretch{2}
\usepackage{setspace}
\setstretch{2}
Just realized that creating a double-spaced document in Latex gives you more lines-per-page than double-spacing in Word. If a grant says your application must be "8 pages double-spaced"...will I get penalized if I didn't use Word's stricter definition? 😬😬😬
August 31, 2026 at 5:14 PM
Just realized that creating a double-spaced document in Latex gives you more lines-per-page than double-spacing in Word. If a grant says your application must be "8 pages double-spaced"...will I get penalized if I didn't use Word's stricter definition? 😬😬😬
Reposted by Jay Hennig
How can we design experiments that make computational models disagree? One section of our new @natrevneuro.nature.com Review with @kriegeskorte.bsky.social and @heikoschuett.bsky.social examines studies that used stimulus sets designed to elicit distinct predictions from competing models. 1/16
August 28, 2026 at 7:10 PM
How can we design experiments that make computational models disagree? One section of our new @natrevneuro.nature.com Review with @kriegeskorte.bsky.social and @heikoschuett.bsky.social examines studies that used stimulus sets designed to elicit distinct predictions from competing models. 1/16
Reposted by Jay Hennig
🧵 What is the briefest visual response we can elicit? Intuitively, that's the response to the briefest stimulus, aka a flash, right? 📸
No, we can go even briefer! Using a deceptively simple technique, we designed stimuli that elicit briefer-than-brief responses
www.biorxiv.org/content/10.6...
No, we can go even briefer! Using a deceptively simple technique, we designed stimuli that elicit briefer-than-brief responses
www.biorxiv.org/content/10.6...
www.biorxiv.org
August 28, 2026 at 2:25 PM
🧵 What is the briefest visual response we can elicit? Intuitively, that's the response to the briefest stimulus, aka a flash, right? 📸
No, we can go even briefer! Using a deceptively simple technique, we designed stimuli that elicit briefer-than-brief responses
www.biorxiv.org/content/10.6...
No, we can go even briefer! Using a deceptively simple technique, we designed stimuli that elicit briefer-than-brief responses
www.biorxiv.org/content/10.6...
Reposted by Jay Hennig
The URH: representations are shaped by the slice of reality a system can sense, act on, and cares about: its Umwelt.
Alignment is explained by overlap of ecological constraints, not convergence onto a veridical world model. Universality cannot explain both similarity *and* systematic difference.
6/n
Alignment is explained by overlap of ecological constraints, not convergence onto a veridical world model. Universality cannot explain both similarity *and* systematic difference.
6/n
August 28, 2026 at 2:59 PM
The URH: representations are shaped by the slice of reality a system can sense, act on, and cares about: its Umwelt.
Alignment is explained by overlap of ecological constraints, not convergence onto a veridical world model. Universality cannot explain both similarity *and* systematic difference.
6/n
Alignment is explained by overlap of ecological constraints, not convergence onto a veridical world model. Universality cannot explain both similarity *and* systematic difference.
6/n
Reposted by Jay Hennig
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 Jay Hennig
Our task-switching paper is now out at Current Biology!
www.cell.com/current-biol...
We find that that our brains reset to a task-neutral state between trials, providing flexibility when the upcoming task is uncertain.
RNNs also learn this strategy, but only when trained to switch tasks.
www.cell.com/current-biol...
We find that that our brains reset to a task-neutral state between trials, providing flexibility when the upcoming task is uncertain.
RNNs also learn this strategy, but only when trained to switch tasks.
August 27, 2026 at 3:39 PM
Our task-switching paper is now out at Current Biology!
www.cell.com/current-biol...
We find that that our brains reset to a task-neutral state between trials, providing flexibility when the upcoming task is uncertain.
RNNs also learn this strategy, but only when trained to switch tasks.
www.cell.com/current-biol...
We find that that our brains reset to a task-neutral state between trials, providing flexibility when the upcoming task is uncertain.
RNNs also learn this strategy, but only when trained to switch tasks.
Reposted by Jay Hennig
Interesting piece in which @itsneuronal.bsky.social suggests dropping the common bit-per-spike metric & reporting the duration of held-out data to validate a model
Seems to be just 120ms of data for a sharply tuned, high firing rate head-direction cell, i.e. just 3-4 spikes
arxiv.org/abs/2607.28779
Seems to be just 120ms of data for a sharply tuned, high firing rate head-direction cell, i.e. just 3-4 spikes
arxiv.org/abs/2607.28779
Bits per Spike as a Betting Game: An Interpretable Unit for Held-Out Log-Likelihood in Neural Data Analysis
Held-out log-likelihood is the standard currency for comparing statistical models of neural spike trains, and is often reported as bits per spike relative to a homogeneous Poisson baseline. The units ...
arxiv.org
August 27, 2026 at 11:07 PM
Interesting piece in which @itsneuronal.bsky.social suggests dropping the common bit-per-spike metric & reporting the duration of held-out data to validate a model
Seems to be just 120ms of data for a sharply tuned, high firing rate head-direction cell, i.e. just 3-4 spikes
arxiv.org/abs/2607.28779
Seems to be just 120ms of data for a sharply tuned, high firing rate head-direction cell, i.e. just 3-4 spikes
arxiv.org/abs/2607.28779
Reposted by Jay Hennig
How does the brain continually adapt? Fresh in @natureportfolio.nature.com our super exciting collaboration with @mark-j-wagner.bsky.social showing that the cerebellum contextualises cortical dynamics enable multi-task learning:
www.nature.com/articles/s41...
www.nature.com/articles/s41...
Granule cells reorient cortical trajectories to separate contexts - Nature
Simultaneous imaging of premotor cortex and cerebellar granule cells in mice learning two skills in parallel shows that trajectories generalize in cortex but coherently reorient apart in granule cells...
www.nature.com
August 27, 2026 at 9:29 PM
How does the brain continually adapt? Fresh in @natureportfolio.nature.com our super exciting collaboration with @mark-j-wagner.bsky.social showing that the cerebellum contextualises cortical dynamics enable multi-task learning:
www.nature.com/articles/s41...
www.nature.com/articles/s41...
Reposted by Jay Hennig
Now out at PNAS! The final version includes a smart analysis suggested by a reviewer showing that biomarkers collected in blood track our neural population metrics. A very exciting connection between basic neurophys and translational approaches. Congrats to all! www.pnas.org/doi/10.1073/...
August 25, 2026 at 4:52 PM
Now out at PNAS! The final version includes a smart analysis suggested by a reviewer showing that biomarkers collected in blood track our neural population metrics. A very exciting connection between basic neurophys and translational approaches. Congrats to all! www.pnas.org/doi/10.1073/...