Hari Kalidindi

Thanks Kevin!

Favourite paper I have read this year. Check it out! Great work @harikalidindi.bsky.social and @fredcrevecoeur.bsky.social!

awesome paper bridging the gap between RNN and optimal control models of motor control

Enjoyed a lot doing this work with @fredcrevecoeur.bsky.social throughout! Glad it's finally out 🎉🎉🎉

Here is the accompanying code for implementing:
github.com/neurohari/si...

Not playing around: Why neuroscience needs toy models Amid the rise of billion-parameter models, I argue that toy models, with just a few neurons, remain essential—and may be all neuroscience needs.

Amid the rise of billion-parameter models, I argue that toy models, with just a few neurons, remain essential—and may be all neuroscience needs, writes @marcusghosh.bsky.social.

#neuroskyence

www.thetransmitter.org/theoretical-...

Now published in the Journal of Neurophysiology:
journals.physiology.org/doi/full/10....

Get in touch if you think this tool could help in your science! We will be developing improvements and extensions over the next year.

Image of robots struggling with a social dilemma.

1/ Why does RL struggle with social dilemmas? How can we ensure that AI learns to cooperate rather than compete?

Introducing our new framework: MUPI (Embedded Universal Predictive Intelligence) which provides a theoretical basis for new cooperative solutions in RL.

Preprint🧵👇

(Paper link below.)

Why AlexNet Died in AI but Lingers in Neuroscience — Through the Lens of Popper and Kuhn When I talk to any of my machine learning researcher colleagues in 2025, they tell me that a model from 2023 is prehistoric. In AI, a year…

Good piece by @kohitij.bsky.social on why neuroscientists use an "outdated" vision model. Neuroscience is different than AI and that's ok! medium.com/@kohitij_716...

Very interesting work!

The brain computes by processing information over time through interactions between connectivity and dynamics that are hard to model. Here we infer these interactions from data and find they better predict cognitive performance! www.nature.com/articles/s41... w/ @lindenmp.bsky.social

Join us for Fall 2026. In our group, you can run studies from human behavior and neuroimaging, to large-scale NHP ephys, and join them up with a robust computational foundation. Bonus: you can help build the reading list.

European universities leading the way

Thread of French and Dutch research institutes slowly unsubscribing from web of science (and thence impact factors).

0/10 Thanks for the interest in our preprint. Some takes say it negates or fully supports the “manifold hypothesis”, neither quite right. Our results show that if you only focus on the manifold capturing most of task-related variance, you could miss important dynamics that actually drive behavior.

How I contributed to rejecting one of my favorite papers of all time I believe we should talk about the mistakes we make.

How I contributed to rejecting one of my favorite papers of all times, Yes, I teach it to students daily, and refer to it in lots of papers. Sorry. open.substack.com/pub/kording/...

What neuroscience can tell AI about learning in continuously changing environments Nature Machine Intelligence - Durstewitz et al. explore what artificial intelligence can learn from the brain’s ability to adjust quickly to changing environments. By linking neuroscience...

Unlike current AI systems, animals can quickly and flexibly adapt to changing environments.

This is the topic of our new perspective in Nature MI (rdcu.be/eSeif), where we relate dynamical and plasticity mechanisms in the brain to in-context and continual learning in AI. #NeuroAI

Sequence preparation is not always associated with a reaction time cost The extent to which a sequence of movements is prepared before initiating the first movement is a longstanding question in motor neuroscience. The observation that reaction time (RT) increases for lon...

0/7 Excited to 📢 that our (@mkashefi.bsky.social @diedrichsenjorn.bsky.social @andpru.bsky.social) new preprint on sequence preparation and its effect on reaction time is now up: www.biorxiv.org/content/10.1...
Please get in touch if there is anything you'd like to discuss! Brief summary 🧵👇

What's the main factor that prevents us from getting universities and grant committies to pull the plug on these massive profiteering journals and put control back in the hands of researchers ?

Two posts from Bluesky. The first one shows a figure from a paper published in Nature Scientific Reports full of totally incoherent AI fabricated gibberish words. The other a comment on a recently published paper by eLife discussing the paper and its peer reviews which were published along with the paper.

Nature Sci Rep publishes incoherent AI slop. eLife publishes a paper which the reviewers didn't agree with, making all the comments and responses public with thoughtful commentary. One of these journals got delisted by Web of Science for quality concerns from not doing peer review. Guess which one?

13/ 😀Feel free to reach out to discuss this work, or the application of it to your field of study. Or come swing by our poster at #NeurIPS2025. We’d love to chat!

📄 Paper: openreview.net/forum?id=I82...
💾 Code: github.com/adamjeisen/J...
📍 Poster: Thu 4 Dec 11am - 2pm PST (#2111)

Very interesting work!

How do brain areas control each other? 🧠🎛️

✨In our NeurIPS 2025 Spotlight paper, we introduce a data-driven framework to answer this question using deep learning, nonlinear control, and differential geometry.🧵⬇️

looks very cool!

I am not even sure it is a hypothesis.

I mean, it is a certainty that neural activity coding for a behavior is not using the full subspace of coding available.

It is interesting how many dimensions are in use.

But these are almost mathematically guaranteed.

Can you state the hypothesis?

"The inevitability and superfluousness of cell types in spatial cognition". Intuitive cell types are found in random artificial networks using the same selection criteria neuroscientists use with actual data. elifesciences.org/reviewed-pre... 1/2

It looks like networks stripped to their bare minimum end up emulating multiple observations while being analytically tractable. I'm curious to see how learning in networks interacts with this control process

This has made me think that it’s just as important to ask what an ideal actor would do without imposing structure, nonlin or data-training on the networks in the first place...this can allow us to interpret empirical observations in a rule based framework (a useful complementary approach!!)

I like using ANNs for prediction! but one challenge is that we can engineer them in multiple ways to generate similar data. E.g. M1-like patterns emerge from both sensory and autonomous networks with different optimization rules. They depend on training method, and tricky to infer organizing rules

Motor Cortical Output Integrates Distorted Proprioceptive Feedback Proprioceptive feedback from muscles is essential for continuous monitoring and precise control of limb movement, yet how such peripheral feedback is integrated into ongoing descending motor cortical ...

In fact, in some recent work, we showed that transiently blocking some proprioceptive feedback *increases* the dimensionality of dynamics along a direction orthogonal to the task manifold, with only *weak* effects on behavioral trajectories.

Neural dynamics outside task-coding dimensions drive decision trajectories through transient amplification Most behaviors involve neural dynamics in high-dimensional activity spaces. A common approach is to extract dimensions that capture task-related variability, such as those separating stimuli or choice...

“Our findings challenge the conventional focus on low-dimensional coding subspaces as a sufficient framework for understanding neural computations, demonstrating that dimensions previously considered task-irrelevant and accounting for little variance can have a critical role in driving behavior.”