Yoda: may the fourth be with you ;) McCoy Tyner: you mean may the fourths be with you *plays quartal chords* Star Wars + Jazz nerds: whoa you’re both huge nerds

What do you do when your area chair recommendation contradicts most reviewers, does not engage with the paper, reviews, or rebuttals, shows lack of basic knowledge in the field, and gives a meta-review that is generally incoherent? Asking for a friend, and yes, it’s really that bad #ICML2026

📣 The difference between replicable and not replicable is not itself scientifically replicable. 📣 New work with Erkan Buzbas, showing that verdicts such as "X% of results replicated" are based on an inferential machinery that doesn't work. arxiv.org/abs/2604.26268

Me: what if it’s not enough tiramisu?

My slightly sobering take on this timeless (and important) question in cognitive/brain science is that full observability, perturbability, etc can be useful (though often overkill) but they are overrated (or far too prioritized compared to other things we might care about to make progress)

Due to idiosyncratic logistics of moving countries I have to be away from my books for 2 months. This is a huge departure from my rule that my book collection should be readily available and at close proximity at all times

New paper 🌟out now in Nat. Neurosci.: www.nature.com/articles/s41.... With advisors @davidpoeppel.bsky.social and @Nai! We show that, while LLMs are optimized to predict the next word, the human brain modulates prediction efficiency by strategically grouping words into constituents. 1/n More 👇

📣 Introducing Easy Peasy, an EC workshop focused on simple, elegant, teachable results! Do you have a result (even old, or a special case of a longer paper) that is so self-contained and easy to communicate that you could explain all the details in under 15 minutes? Then this is the venue for you!

I’m tempted to put as many m-dashes as possible in my next paper

Sometimes academics are invited to give a talk to share their thoughts on a specific topic (e.g., in the context of a targeted symposium or meeting) and instead they give a generic lab promo talk. Why?? Is this not as frowned upon as I’m assuming it should be?

“Even if any given terminology is a reflection of reality, by its very nature it must be a selection of reality, and to this extent it must function as a deflection of reality” (Burke, 1966; possibly paraphrased).

Given we had Velociraptor Awareness Day. Did you know, many species of famous theropods were named after Polish women in palaeontology. Velociraptor osmolskae & Citipati osmolskae are named after Halszka Osmólska, a Mongolian dinosaur expert, responsible for describing Deinocheirus and Gallimimus.

Today in ~research that is cooler than mine~

What function did ~100k-year-old engravings from Blombos Cave & Diepkloof serve? Decoration, identity marking, proto-writing? osf.io/preprints/ps... uses transmission chains + cognitive experiments to find out & help answering one of the hardest questions in cognitive archaeology. long thread! 1/

Me, defending my proposal in front of a grant panel: My name is Fede Adolfi. I am not one of them. I wish to save the world.

I will express my opinions on The Discourse when somebody invents a social media mechanism that prevents them from instantly becoming part of The Discourse. Until then, I’m sorry or you’re welcome

Brief fun survey from Jessica, Andrew & myself: If you are a faculty member, research scientist, postdoc, or senior Ph.D. student in any area of science, please take five minutes and fill it out. We’ll share the results widely along with some reflections. docs.google.com/forms/d/e/1F...

The first few articles in this collection are now published, including ours. I look forward to reading all. Our piece is not open access because of the ridiculous fees but the preprint linked at the end of the quoted thread (will link again below) is the accepted version of our commentary.

We (@edbaggs.bsky.social, @segundo-ortin.bsky.social, @guicogsci.bsky.social, & I) have organised the conference "Dimensions of Radical Embodiment" for the last few years [Stay tuned for the next one!]. And we just edited a special issue around it. Here's the intro: link.springer.com/article/10.1...

Crater Nutella

The death of bluesky is greatly exaggerated Or, I’m getting my shitposts out while I can

Call for a PhD position in Cognitively inspired natural language processing with Lisa Beinborn and me. Part of the new SPP "Robust Assessment & Safe Applicability of Language Modelling: Foundations for a New Field of Language Science & Technology” (LaSTing). huds.uni-goettingen.de/assets/Call_...

There’s a kind of brain rot that can come, it seems, from spending too much time at prestigious/elite academic institutions. It can make otherwise sensible people go “you know [unambiguously bad thing for everyone]? Well hear me out because maybe if we [cognitive acrobatics] then it’s a good thing?”

Me deriving predictions from theories in the cognitive and brain sciences: I said 3 to 6, or 6 to 8. It could be 2. Or it could be 17. It could be any particular number. It’s anything I say it is.

The existence of humble-brag implies the existence of bash-praise and guess what that’s definitely a thing that exists

How do you identify which problems are interesting and valuable? When people don’t work on problems that matter, why do you think that is?

Our ICLR paper 👆🏻 on the parametrized complexity landscape of circuit discovery is cited in this thought-provoking piece 👇🏻 by @gershbrain.bsky.social bsky.app/profile/thet...

Frontiers in Nature Human Behavior Comms Advances

I guess we’re doing this

I guess we’re doing this

The disappearing link to the preprint is here.

A lot of AI/ML relies on sampling from large datasets. These samples need to be relatively unbiased in order to get good performance. Animals learn continually from highly correlated — and therefore biased — behavioral episodes. Is there work exploring how brains circumvent the issue of bias? 1/2

Me when I get a citation alert email

Excited to share my latest piece in Science “AITA for treating graduate students as paper producing meat machines?”

Me when I get a citation alert email

Excited to share my latest piece in Science “AITA for treating graduate students as paper producing meat machines?”

TIL that deer will eat small birds if they happen to find dead ones lying around Two faced macabre little devils what’s wrong with you

Looking forward to what promises to be an interesting week of discussions on explanation in neuroscience. Starting tomorrow!

Reviewed 6 papers and ended up with a uniform split over bad / meh / good *taps earbud* Did I enjoy the process though? Uh no, Janet, I did not. This is Reviewer 4, reporting from the trenches. #icml2026

Call for position papers: you have one job — state your position in the title Authors: The Adventures of Neural Networks #icml2026

Call for position papers: you have one job — state your position in the title Authors: The Adventures of Neural Networks #icml2026

After dislodging the meteorite fragments from the roofs, people sat the fragments at the kitchen table and gave them a speech about being late to deliver signs of the apocalypse

I went to a really cool jazz concert a couple of days ago but no photos or recording of any kind were allowed during the performance so here you go, social media overlords #jazzsky

Upcoming talk in the Theory of Interpretable AI seminar: Federico Adolfi frames interpretability in terms of computational complexity theory.

Me: I’m elated to announce my new suite of agentic LLMs fine-tuned to exhibit subtle behaviors that maximally elicit complex vicarious emotions in humans such as shame and embarrassment Billionaire: I can help you grow with an initial investment of— Me: I call it CrinGPT B: get out

To take this to the limit, we note that for every ε-ssay there exists a δ-ebate.

In this ε-ssay, I develop a kernel of an idea so diminishingly small that you’ve just read it You’re welcome

Me: alright then, time to work on the abstract for my talk and ONLY the abstract for my talk until it’s done Also me: *spends an hour working on a schematic for a single slide* Nailed it

In this household we are generally self deprecating and also we celebrate rejections with dangerously delicious cookies

I don’t know who needs to hear this but *uncanny mouth noises*