Sign in

Satpreet (Sat) Singh

@satpreetsingh.bsky.social
1.1K followers 4.4K following 52 posts

Postdoc @harvard.edu @kempnerinstitute.bsky.social Homepage: satpreetsingh.github.io Twitter: x.com/tweetsatpreet

PostsRepliesMedia
Reposted by Satpreet (Sat) Singh
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
Very excited to announce our most recent paper! TL;DR we made our prior work on comparing dynamics (DSA) much faster and generalizable to diverse data domains and problem settings (1/)
biorxiv.org
A metric for comparing complex systems by their dynamics
Comparisons are fundamental to science: experiment against model, one organism against another, a system against itself across time. Because many systems, from brains to climate, are characterized by ...
17118
Reposted by Satpreet (Sat) Singh
IEEE Spectrum @spectrum.ieee.org · 30/06/2026
The founders of Modos share how they created the Modos Flow, a complete #e-paper monitor. ”We want the whole thing open source, so all our parts need to have open documentation with nothing gated behind an NDA.” spectrum.ieee.org/modos-e-pape...
0123
Reposted by Satpreet (Sat) Singh
Daphne Cornelisse @daphne-cornelisse.bsky.social · 21/08/2026
What are the biggest bottlenecks to applying reinforcement learning and high-performance simulation in biology? In a new series, I’ll explore this question one domain at a time, starting with weakly electric fish. daphnecornelisse.substack.com/p/training-a...
0304
Reposted by Satpreet (Sat) Singh
Tom McCoy @rtommccoy.bsky.social · 20/08/2026
Since many are starting grad school soon, let me re-share my One Big Tip™️ for research! Research involves many skills - collaborating, writing, presenting, etc. But many of these skills can be unified under a single overarching ability: theory of mind Blog post link in reply
Illustration of the blog post's main argument, summarized as: "Theory of Mind as a Central Skill for Researchers: Research involves many skills.If each skill is viewed separately, each one takes a long time to learn. These skills can instead be connected via theory of mind – the ability to reason about the mental states of others. This allows you to transfer your abilities across areas, making it easier to gain new skills."
28019
Reposted by Satpreet (Sat) Singh
Kanaka Rajan @kanakarajanphd.bsky.social · 04/08/2026
Heading to #CCN26? Stop by and see what the Rajan Lab has been working on! Don't miss presentations from John Vastola, Satpreet Singh, and Nathan Wu. Drop by their sessions to chat about our ongoing projects and say hello to the team! @purpleantelope.bsky.social @satpreetsingh.bsky.social
094
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 30/07/2026
I should add that, at some level of complexity, it'll be way more efficient/tractable to imitate/regularize a little like @daphne-cornelisse.bsky.social's work arxiv.org/abs/2403.19648 ... but atm we're trying to see how far we can push the pure-RL approach for a maximally "normative" model.
arxiv.org
Human-compatible driving partners through data-regularized self-play reinforcement learning
A central challenge for autonomous vehicles is coordinating with humans. Therefore, incorporating realistic human agents is essential for scalable training and evaluation of autonomous driving systems...
150
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 30/07/2026
Thanks Eugene! In case you missed it, we're using YOUR (Yu/Velu, Vinitsky et al 2022) implementation of MAPPO. Appreciate you pointing us in the right direction! arxiv.org/abs/2103.01955
arxiv.org
The Surprising Effectiveness of PPO in Cooperative, Multi-Agent Games
Proximal Policy Optimization (PPO) is a ubiquitous on-policy reinforcement learning algorithm but is significantly less utilized than off-policy learning algorithms in multi-agent settings. This is of...
240
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
Congratulations! 🚀🚀
010
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
Yes, that seems like a cool follow-up! This work is a bit more framework focused; many extensions planned, including those that make use more bio-inspired network architectures.
000
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
Thanks Kam! 🙏
110
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
Coauthors: Hanson Lu, @aaronwalsman.bsky.social , Federico Pedraja, @denisturcu.bsky.social , Pratyusha Sharma, @nsaphra.bsky.social 📄 arxiv.org/abs/2511.08436
arxiv.org
Active Electrosensing and Communication in MARL-trained Weakly Electric Fish Collectives
How complex collective behavior emerges from individual interactions is a fundamental scientific question, but experimental cost and difficulty of simultaneous multi-brain recordings limit direct stud...
040
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
15/n Co-first authored with Sonja Johnson-Yu; during postdoc with @kanakarajanphd.bsky.social at @kempnerinstitute.bsky.social al + @harvardmed.bsky.social ed.bsky.social, in collaboration with Nate Sawtell's group at @zuckermanbrain.bsky.social
120
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
14/n Limitations aside, we now have a tractable, fully observable model for generating and prioritizing hypotheses about real weakly electric fish. It should adapt easily to other social animals, where simultaneous multi-brain recordings remain hard. 💻 github.com/KempnerInstitute/wef
github.com
GitHub - KempnerInstitute/wef
Contribute to KempnerInstitute/wef development by creating an account on GitHub.
110
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
13/n 🧪Virtual fish allow interventions that are hard in vivo. Falsifiable predictions for future experiments include: - Short-range sensor ablations impair foraging, - Social sensor ablation increases pulse rates, - Muting changes spacing, - Food layout drives inequality
110
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
12/n 🧠Inside the "brains": RNN activity has far lower effective rank than the network width. It robustly encodes task and social variables, e.g., the field-line homing angle decodes better than the naive straight-line angle. Two fish align their activity more at short range than long.
110
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
11/n To isolate the role of signaling, we built a 2-agent, 1-patch assay: Agent A starts on a patch as Agent B approaches. Size dictates the outcome: - Larger B displaces A - Equal-sizes jostle for access - Smaller B hesitates Overall, B finds the patch more easily with A around.
110
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
10/n A pecking order therefore emerges: Larger fish bite and chase smaller fish more often, consistently winning confrontations with more bites on average. In groups, higher size inequality correlates with larger inequalities in food consumption.
120
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
9/n ⚔️ We also see emergent aggression, like in real fish! Although agents are not explicitly programmed with social rules, beyond incurring a higher penalty when bitten by larger fish, they spontaneously exhibit competitive behavior, including chasing and biting during foraging.
161
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
8/n 📈 Agent EOD signaling statistics also resemble biology: - Heavy-tailed inter-discharge-interval (IDI) distributions - EOD rate acceleration during feeding - EOD suppression around larger fish
120
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
7/n ⚡ Did we need to model electric fields? Yes! In a homing task, electric-field sensing produced smooth, curved approaches aligned with field geometry, qualitatively resembling real fish data (Hopkins ’97). Conversely, a ray-casting/LIDAR baseline followed nearly straight paths.
130
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
6/n With these pieces in place, MARL training produces agents that forage using active electro-sensing in a variety of virtual settings, both alone and socially. Here, four agents explore a patchy arena; the flashes are their EODs.
120
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
5/n We use multi-agent reinforcement learning (MARL): N recurrent-network agents share a 2D arena. Each can move, emit pulses, and 'bite' — abilities modeled on real fish. Rewards for foraging and penalties for being bitten; other dynamics emerge naturally.
110
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
4/n 📡 Our virtual fish have three receptor types, modeled on real fish: M: active, short range (~5 cm for food; ~10 cm for fish) A: passive, short range K: passive, long range (~100 cm), for social signals only
110
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
3/n We built a physics engine to simulate agent movements and the dynamic electric landscape: ⚡ Electric sources on each fish (the EOD pulses) 🌇 Distortions from surrounding objects, including other agents, food and boundary walls.
120
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
2/n We model G. petersii, a weakly electric fish that emits electric organ discharges (EODs) to both sense their surroundings and to communicate with other fish, a rich setting for studying complex social behavior.
130
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 29/07/2026
How do you "see" with electric eyes? How does collective behavior emerge from individual interactions? Nocturnal weakly electric fish evolved to do this, but studying naturalistic social behavior is very hard. Our solution? Virtual 'fish' 🤖🐟⚡ 📄 arxiv.org/abs/2511.08436
17925
Reposted by Satpreet (Sat) Singh
The Transmitter @thetransmitter.bsky.social · 15/07/2026
In the latest entry of Frameshift, our series about alternative careers in neuroscience, Soha Ashrafi, executive director of research at Harvard Medical School, shares how supporting a department full of neuroscientists brings her “scientific joy.” www.thetransmitter.org/frameshift/m...
thetransmitter.org
Making an impact through academic administration
As executive director of research at Harvard Medical School, Soha Ashrafi supports more than 300 scientists, students and staff members.
1125
Reposted by Satpreet (Sat) Singh
Simons Foundation @simonsfoundation.org · 16/06/2026
In a new @cp-neuron.bsky.social article, members of the Simons Collaboration on Ecological Neuroscience (SCENE) outline their plans to investigate how the brain links perception, cognition and action by encoding environmental affordances: bit.ly/49YfgaE
cell.com
The Simons Collaboration on Ecological Neuroscience: Studying how the brain interacts with the world
The Simons Collaboration on Ecological Neuroscience (SCENE) seeks to uncover general principles of brain function through an ecological perspective: studying perception, cognition, and action in the c...
02310
Reposted by Satpreet (Sat) Singh
Shervin Safavi @neuroprinciplist.bsky.social · 09/06/2026
1/11 Happy to share our TICS paper on using the flexibility of one of the most basic cognitive functions, perception, to understand one of the most complex cognitive dysfunctions, psychiatric conditions (also my first formal work in computational psychiatry 🎉) 📄: www.cell.com/trends/cogni... 🧵 : 👇
cell.com
Perceptual multistability: a multifaceted window into brain dysfunctions
Perceptual multistability, observed across species and sensory modalities, offers valuable insights into numerous cognitive functions and dysfunctions. For instance, differences in temporal dynamics a...
320966
Reposted by Satpreet (Sat) Singh
Roxana Zeraati @roxana-zeraati.bsky.social · 07/07/2026
Our paper "Neural timescales from a computational perspective" is finally out in @natneuro.nature.com: www.nature.com/articles/s41... We discuss how computational models and methods can distill empirical observations on neural timescales into quantitative, testable theories of brain computation.
nature.com
Neural timescales from a computational perspective - Nature Neuroscience
This review integrates computational approaches to provide a unified view on how data analysis methods, biophysical mechanistic models and machine learning approaches can help to uncover the origins a...
422986
Reposted by Satpreet (Sat) Singh
Nima Dehghani @neurovium.bsky.social · 16/06/2026
new paper, #NeuroAI 📣📜 Can measured cortical organization be used as an inductive bias for artificial recurrent neural networks? In this work, we ask whether cortical geometry, wiring, and function can push RNNs learn. Not as metaphor, but as measurable structure! 1/n🧵👇 arxiv.org/abs/2606.14975
arxiv.org
Harnessing cortical geometry, wiring, and function as inductive biases for recurrent neural networks
How the wiring and functional organization of cortex shape recurrent computation remains a central question in both neuroscience and machine learning. Here, we leverage data released through the Machi...
12715
Reposted by Satpreet (Sat) Singh
Wei-Chung Allen Lee @darbly.bsky.social · 09/06/2026
Now published - the #BANC! A full central nervous system (CNS) connectome of a limbed animal at single-synapse resolution, enabling us to follow sensory-motor arcs and understand how the CNS controls the body. rdcu.be/fncjS. #neuroscience. Video by @quorumetrix.bsky.social 1/18
214984
Reposted by Satpreet (Sat) Singh
Noga Zaslavsky @nogazs.bsky.social · 24/04/2025
Excited to share our new paper "Towards Human-Like Emergent Communication via Utility, Informativeness, and Complexity" direct.mit.edu/opmi/article... @rplevy.bsky.social And looking forward to speaking about this line of work tomorrow at @nyudatascience.bsky.social!
direct.mit.edu
Towards Human-Like Emergent Communication via Utility, Informativeness, and Complexity
Abstract. Two prominent, yet contrasting, theoretical views are available to characterize the underlying drivers of language evolution: on the one hand, task-specific utility maximization; on the othe...
0376
Reposted by Satpreet (Sat) Singh
Talmo Pereira @talmo.bsky.social · 05/05/2026
We tracked mice in space! This video shows our estimate of their self-induced centripetal force during circling in 3D 👩‍🚀 Preprint thread below 👇
37733
Reposted by Satpreet (Sat) Singh
Will Redman @wtredman.bsky.social · 28/04/2026
New preprint out 🚨 “Predictive pursuit emerges in high-dimensional recurrent neural networks”! This was an awesome collaboration with co-first author @fatihdinc.bsky.social , @andyalexander.bsky.social, @xiaoxiao-lin.bsky.social, and May Chen biorxiv.org/content/10.6... 1/
biorxiv.org
12912
Reposted by Satpreet (Sat) Singh
Ann Huang @annhuang42.bsky.social · 23/04/2026
[ #ICLR2026 ] How do we know if two systems are performing the same computation when they are constantly driven by different external inputs? 🧠🤖 I’ll be presenting our novel method InputDSA tomorrow April 23 (2:15pm-4:45pm EDT in Pavilion 3 P3-#1614)📍 Come swing by our poster! I’d love to chat!
0248
Reposted by Satpreet (Sat) Singh
Kanaka Rajan @kanakarajanphd.bsky.social · 22/04/2026
✍️ In the @kempnerinstitute.bsky.social blog: our new tool built to compare the dynamics of complex systems when both internal circuitry and external inputs shape their behavior. Catch @annhuang42.bsky.social presenting this work at #ICLR! kempnerinstitute.harvard.edu/research/dee...
kempnerinstitute.harvard.edu
InputDSA: Demixing then comparing recurrent and externally driven dynamics in complex systems - Kempner Institute
We explored how to measure the similarity between two complex systems when they are driven by external inputs, like biological neural circuits or reinforcement learning agents. Our novel method, calle...
14315
Reposted by Satpreet (Sat) Singh
Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 15/04/2026
Still hiring for PhD candidates who are *specifically* excited in building and deploying RL systems for self-driving vehicles and other multi-agent planning settings. Shoot me an email if you think this is you and please help spread the word!
24214
Reposted by Satpreet (Sat) Singh
Amos Posner @amosposner.bsky.social · 10/04/2026
There is no McTry.
A McDonald’s where the sign has been reduced to McDo.
368222083731
Reposted by Satpreet (Sat) Singh
Reinforcement Learning Conference @rl-conference.bsky.social · 31/03/2026
We have the keynote speakers for RLC2026 now! Thrilled to welcome Rika Antonova, Sheila McIlraith, Marc G. Bellemare, Danijar Hafner, Balaraman Ravindran! Details: rl-conference.cc/index.html The RL community is coming together this August in Montréal, Québec, Canada. Hope you make it!
rl-conference.cc
RLC 2026
02410
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 28/03/2026
Speakers: Omri Barak, Cristina Savin, @lilweb.bsky.social @reecedkeller.bsky.social Caroline Haimerl, Hannah Choi, @xaqlab.bsky.social Srini Turaga, Yanan Sui, @trackingskills.bsky.social Virtual: Eva A Naumann, Justin N Wood
000
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 28/03/2026
Big thanks to our speakers, panelists, and everyone who participated in the discussions. We hope these talks are useful to the broader NeuroAI community!
100
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 28/03/2026
Talk recordings from our CoSyNe 2026 workshop on Agent-based Models in Neuroscience are now online! Playlist: www.youtube.com/playlist?lis... RL agents and theory, interoception, biomechanical models connectome-constrained models, and more. #CoSyNe #NeuroAI #CompNeuro #RL #EmbodiedAI
youtube.com
CoSyNe 2026 Agents in Neuroscience Workshop - YouTube
https://neuro-agent-models.github.io/
1276
Reposted by Satpreet (Sat) Singh
Ilenna Jones @ilennaj.bsky.social · 25/03/2026
My memory sucks. Note-taking is my superpower. I decided to upgrade this superpower by making a living knowledge base for research, powered by Claude Code (or any context agent). Here's a github template so you can do the same too. 🧵 github.com/ilennaj/obsidian-scientist-knowledge-template
github.com
GitHub - ilennaj/obsidian-scientist-knowledge-template: Template for keeping a living scientific knowledge database using obsidian and claude code.
Template for keeping a living scientific knowledge database using obsidian and claude code. - ilennaj/obsidian-scientist-knowledge-template
4192
Reposted by Satpreet (Sat) Singh
Dare Obasanjo @carnage4life.bsky.social · 21/03/2026
Steve is an engineering VP at Microsoft and I find a similar pattern is showing up for product managers. I see lots of PMs who were already good now multiplying their impact with AI while weaker PMs are now using AI to generate work slop which is a waste of everyone’s time.
59215
Reposted by Satpreet (Sat) Singh
Sibo Wang-Chen @wangchen.bsky.social · 14/01/2026
🧠💪🤖 Check out our new review article on how neuromechanical "digital twins" are used to study the neuroscience of embodied control, and how they may contribute to future #NeuroAI! arxiv.org/abs/2601.08056
Experimental neurobiology informs better designs for animal-like AI systems, while advances in physics simulation, robotics, and NeuroAI help validate hypotheses of brain functioning.
141
Reposted by Satpreet (Sat) Singh
AmiyaJLO @amiyajlo.bsky.social · 15/03/2026
#IdesOfMarch
A white bust sculpture is used as a knife holder, with multiple knives and a pair of scissors stuck into it, placed on a kitchen island.
8772292056
Reposted by Satpreet (Sat) Singh
Kempner Institute at Harvard University @kempnerinstitute.bsky.social · 12/03/2026
🧠🤖 Are you at #COSYNE2026? Check out the #KempnerInstitute's presentations! 👇 #neuroscience #NeuroAI
166
Satpreet (Sat) Singh @satpreetsingh.bsky.social · 12/03/2026
Thank you for sharing and the kind words!
010
Reposted by Satpreet (Sat) Singh
Shervin Safavi @neuroprinciplist.bsky.social · 12/03/2026
Don't miss @maxschwabe.bsky.social's poster this evening (our master's student and presenter awardee 🎉) on decomposing RNN dynamics during naturalistic decision-making! It was a joint work with @armanbehrad.bsky.social and @roxana-zeraati.bsky.social!
1207