Lorenzo Posani

Lorenzo Posani

@lorenzoposani.com

Coding and de-coding brains👨‍💻 ⮂ 🧠 Posani Lab at ICM Paris Brain Institute & CNRS Co-founder of Cubbit

758 followers715 following71 posts

Pinned

Lorenzo Posani reposted

Excited to see this work published! We find that the brain codes semantic relationships similar to contextual LLMs (like GPT2) but unlike LLMs, uses contrastive coding to prevent confusion of highly similar words. Many thanks to the reviewers and BCM neurosurgery team! www.nature.com/articles/s41...

nature.comA population code for semantics in human hippocampus - Nature NeuroscienceFranch et al. show that human hippocampal neurons encode the meanings of the words we hear through distributed, context-sensitive population activity that mirrors some features of large language model...

If you are at Bernstein, come by at poster 60 session IV (Wed 2pm), Charlotte will present our results on compositional and abstract representations of social stimuli in the monkey face patch system. Particularly proud of this as it’s the first poster from the first student of the lab 😊

Lorenzo Posani reposted

The superposition of these two studies — both incredible — reflect a snapshot of "one big thing" systems neuroscience has yet to resolve. In mice: 1. Task-related signals are everywhere. 2. Absent large swaths of "everywhere," suprisingly sophisticated learning & behavior remains.

Lorenzo Posani reposted

🚨Update: Our analysis of global APC expenditure has now been expanded to cover more publishers (7 -> 14) and more years (2019-2025). We estimate $15B in APCs paid over seven years, $3.7B in 2025 alone. Elsevier crosses the $1B threshold by itself in 2025. arxiv.org/abs/2608.16322 #ScholComm

Eric Schares@eschares.bsky.social

🚨 Preprint! We combine our recent open dataset of #APC prices with the article counts per journal-year from #OpenAlex to estimate how much the academic community has paid in APCs over the last 5 years. A. $8.349 billion ($8.968 B in 2023 USD) $2.5B in 2023 alone. arxiv.org/abs/2407.16551 #metasci

Lorenzo Posani reposted

Today in @thenation.com: a harrowing piece written from inside the Dilley ICE detention center by Maryam Tahmasebi, a green card holder who's been locked up along with her husband and son for months despite none of them committing any crime. www.thenation.com/article/soci...

thenation.comOne Woman’s Message From the Hell of ICE DetentionMy husband, son, and I have been locked up for months—despite being permanent residents with no criminal record. It is an unendurable, unending nightmare.

Lorenzo Posani reposted

Lorenzo Posani reposted

For a neural circuit to be "specialized", does it need specialized neurons? In our new work - now out in @NatureNeuro - we address this question in the BLA, using representational geometry to propose a new population-centric way of defining circuit specialization. 🧵👇 1/

View full thread

12/ In short, specialized readouts can be supported by a neural population even with mixed selectivity, as long as the geometry is right. We find that BLA simultaneously encodes multiple variables within an emotional experience, with the right geometry for specialized readouts.

For a neural circuit to be "specialized", does it need specialized neurons? In our new work - now out in @NatureNeuro - we address this question in the BLA, using representational geometry to propose a new population-centric way of defining circuit specialization. 🧵👇 1/

View full thread

11/ The geometry then changes after ingressing, and the mouse no longer represents valence. Instead, only the exposed, aversive condition stands out from the bulk. This is compatible with the coding of a new variable, "safety", which has the highest decoding and CCGP.

For a neural circuit to be "specialized", does it need specialized neurons? In our new work - now out in @NatureNeuro - we address this question in the BLA, using representational geometry to propose a new population-centric way of defining circuit specialization. 🧵👇 1/

View full thread

8/ The key is that specialized neurons are not necessary to satisfy these requirements! We can have the same properties (high CCGP, low CVI) as long as the representational geometry of the encoded variables is disentangled / compositional (left panel):

9/ This is indeed what we observe in the BLA: it represents behavioral and valence variables as "abstract", with a disentangled compositional geometry, characterized by a high CCGP (generalization) and low CVI (interference):

For a neural circuit to be "specialized", does it need specialized neurons? In our new work - now out in @NatureNeuro - we address this question in the BLA, using representational geometry to propose a new population-centric way of defining circuit specialization. 🧵👇 1/

View full thread

6/ A circuit made of specialized sub-populations, such as "valence neurons" and "tremble neurons", clearly satisfies these requirements: sub-populations encode variables independently, and each trivially generalizes across contexts because it cares about only one thing.

7/ But we can also characterize these properties (high generalization, low interference) on a population level. For generalization, we can use cross-condition generalization performance (CCGP). For interference, we developed a new measure: cross-variable interference (CVI):

For a neural circuit to be "specialized", does it need specialized neurons? In our new work - now out in @NatureNeuro - we address this question in the BLA, using representational geometry to propose a new population-centric way of defining circuit specialization. 🧵👇 1/

View full thread

4/ What about the neural code? All variables were highly decodable, but representing the world is not enough to act on it. What matters is "how" this information is organized!

5/ For example, to make quick, precise decisions when faced with threats, the brain needs to read out information precisely (low interference) and flexibly (high generalization) - this is a "specialized" circuit. How can we design one?

For a neural circuit to be "specialized", does it need specialized neurons? In our new work - now out in @NatureNeuro - we address this question in the BLA, using representational geometry to propose a new population-centric way of defining circuit specialization. 🧵👇 1/

2/ The BLA is known to have specialized circuits for valence (+/-), but what about other features of emotional experiences, such as behavior? In a virtual burrow assay, we observed that mice both tremble (~freeze) and ingress (~flee) in response to aversive (shock-paired) odors.

3/ Both trembling and freezing are more frequent in response to aversive odors (CS+), but far from exclusive. Mice typically tremble and then ingress when CS+ is presented, in accordance with the predator-imminence theory (freeze -> flee).

Lorenzo Posani reposted

Lorenzo Posani reposted

Lorenzo Posani reposted

This recent post by @earlkmiller.bsky.social got my attention. It shows how abstraction could emerge in the primate PFC. The paper is hard, but the main idea can be simplified to low-dims, so I made fun a little animation. A🧵with my toy model and notes: #neuroskyence #compneuro #NeuroAI

Lorenzo Posani reposted

Lorenzo Posani reposted

Lorenzo Posani reposted

When reality surpasses fiction! Top: fictional forecast for August 2050, broadcast by French TV in 2014 to warn about the consequences of global warming. Bottom: Real French forecast for yesterday, June 22, 2026.

Lorenzo Posani reposted

In my utopian vision of the future, instead of offloading all the pointless, annoying work tasks (like reports that none one will read and needlessly opaque forms) to robots, we restructure work and society so that the pointless tasks don't happen at all.

Lorenzo Posani reposted

Somebody should invent a socioeconomic system which is good at allocating resources

Lorenzo Posani reposted

Postdoc position in Paris: come help develop new generation human brain computer interfaces ⚡🧠💻 Interested? Contact me if you have experience with machine learning (e.g. simulation-based inference, RL, generative/diffusion models) or dynamical systems. See below for + details and retweet 🙏

Lorenzo Posani reposted

Reminder: All it would take to end the murder of American citizens by an untrained government goon squad is 16 Republicans in Congress voting with Dems to defund ICE (or 23 to impeach and remove Trump — 3 in House & 20 in Senate). That’s it. 23 Americans can vote for the public and end all of this.

Lorenzo Posani reposted

Europe's MSCA postdoctoral fellowships are now attracting many researchers who would have gone to the US. So the payline has dropped below 10% (~1,600 fellowships to 17k applicants). Super tough situation. One more reason to double the next Horizon budget, as proposed by the EU Commission.

Tom Emery@drtomemery.bsky.social

Applications for MCSA are up 64%! 50% now come from outside the EU. Trump effect? The European Research market just got a hell of a lot more crowded. EU should turn on the tap marie-sklodowska-curie-actions.ec.europa.eu/news/msca-po...