It should also be the authors' choice whether the reviews are published before or after a version of record is authorized. Otherwise it's not authors' choice.

I miss 2024

Worst take I've seen on here in awhile.

Being able to Swifty educate a wrong reviewer is an upside that outweighs any logistical concerns.

How does eLife let you do what you want with the reviews? Aren't they public? Whether you click, "make this the version of record" it is stays as a "reviewed preprint" seems semantic.

I've been saying this for years! Just let the authors and reviewers have communicate directly, asynchronously through the eLife forum for a week of two after each round of review and rebuttal for clarifications.

a man is standing in a doorway with the words i 'll see myself out ALT: a man is standing in a doorway with the words i 'll see myself out

Sorry lol

Try a BAKS filter with trial wraparound.

Doesn't "encoding" mean during memory writing, aka learning and plasticity?

Not even mentioning BTSP here is wild

Single author flips table. Wait does it do XOR

It's myself

Ok

Holy shit

That's the best part of research, too. Banging your head against things that no one gets yet!

I was devastated to learn someone else had already put Axolittle axolotl on a t-shirt

I wrote and posted this to Instagram a couple of days ago and it became my most-read poem ever. I’m honored I get to feel big feelings alongside you all. I’m posting it here too and want to use this space more consistently. Hello friends ❤️

No, instead reviewers will just completely misunderstand everything about it but at least one person will read it and love it

Trump administration drops court fight to cap NIH payments for research overhead costs The Trump administration will not be asking the Supreme Court to take up its fight to slash NIH support for research indirect costs

Oh good, Universities can continue to educate doctors and cure diseases, because why would that ever have been in doubt in a modern civilization.

www.statnews.com/2026/04/08/t...

Entorhinal cortex represents task-relevant remote locations independently of CA1 - Nature Neuroscience Neurons in medial entorhinal cortex collectively represent discrete nonlocal positions during immobility. During this nonlocal coding, CA1 is uncoupled from entorhinal cortex. These representations ar...

doi.org/10.1038/s415...

Pfizer has ended its large Phase III Trial of an updated COVID19 vaccine.

They were not able to enroll enough people to meet their goal.

A major problem is the FDA’s wholly unreasonable demand that people with chronic conditions, who benefit most from the vaccine, not be allowed to participate.

Big companies that train big models on big datasets and require big server farms to offer big inference services for big subscription fees.

The novel finding is the brain somehow has a circuit to deliver a delta to each dendrite that is related to the causal effect of that neuron on behavior.

Again, dw = pre * post_soma * mod where post_soma is driven to a target firing rate is not consistent with the data they collected. It's more consistent with dw = pre * (post_dend-post_soma), where post_soma is not at the target.

So I would find it interesting for a model with the same components to develop the same signals over the course of learning. But yes it's an approximation of a gradient descent, for which there are many proposed implementations.

And surprisingly, the brain does appear to do that. The somas are just reporting their firing rates, which are not yet at their targets. The dendrites are getting either more or less average input than the soma, telling it to either increase or decrease its synaptic weights.

This study was more asking the question, does the brain find a way to deliver that type of cell-specific feedback when the mapping of which neurons are supposed to increase or decrease was arbitrarily imposed by the experimenter?

I think it's uncontroversial that if you provide to every neuron either a target signal or a signed error signal to tell it whether it's supposed to increase or decrease its firing rate to reduce performance error, it will work.

Dendritic excitatory-inhibitory balance and branch-specific gating enable selective recall of associative memories www.biorxiv.org/content/10.64898/2026.03...

Because then you have a system that only needs the external teaching signal to be hand delivered to each soma, no need for dendrites, and won't reproduce the somatodendritic residual-based learning mechanism they characterized.