My list on what to read about theory-driven cognitive science has now been updated to add a few more thoughts and references about the role of AI as 'in silico' models of human cognition. Your comments welcome - please add to the googledoc. docs.google.com/document/d/1...

🥳I'm excited to introduce everyone to my group at Leipzig University, the LION (Linguistically-Oriented NLP) Lab! 🦁 Looking forward to exciting work on computational linguistics, interpretability, and multilinguality! Find out more on our website lionlabnlp.github.io

I’ll be giving two talks at @conll-conf.bsky.social, so make sure to stop by and hear me talking about a) child-directed speech and b) child-directed speech! 🙌

Tip on reviewing ARR papers in 2026: - Check the references FIRST. I wasted 40 minutes reviewing a paper before realizing the references were LLM-generated.

I mean, I think the term as it is used today has been voided of anything meaningful, and the labs have so perverted public discourse about AI that each term that was once used to mean a specific (set of) thing(s) no longer reliably mean any of those things. bsky.app/profile/vcar...

📢 ARR May 2026 cycle update We have published a blog post about the increased review assignments in the May cycle, what happened, and what comes next. Please read the full post here: aclrollingreview.org/may26-cycle-... #ARR #ACL #NLProc

> the reality is that there appears to be too many good-faith, properly formatted research papers being submitted for the system to handle in certain cycles. It is a “luxury problem”: the community is simply too productive by present research standards

We took a few LLMs, had them play 800 games against themselves in different instruction languages while keeping the board and actions the same, balanced across player roles, and measured win-loss margins If skills were language-invariant every matchup should be near 0.0 wl margin But they aren't!

New work from my lab! @teagrjohnson.bsky.social built a 12-dimensional narrative framework, annotated Dolma (no small feat given its extreme diversity), and analyzed narrative features across pretraining subsections. Highlight: pretraining data space displays strong narrative organization!

I’m collecting visions of the university at least ten years out (i.e, stories about equilibrium conditions, not short term fixes). What interests me is that the ones I’ve seen so far are compatible. Molly Worthen’s proposal for a technology fast has a radical title, but when you dig deeper +

Meaning emerges from context: by mapping how words share linguistic environments, algorithms can mathematically capture the nuanced relationships… Distributional Semantics by Gemma Boleda https://oecs.mit.edu/pub/gbaucew9 #CognitiveScience

More evidence, from a large-scale study in China, that using AI hurts learning if it undermines mental effort. When homework time drops due to AI use, so do test scores. Across studies, there is a clear theme: AI tutoring in support of classes is good, using AI to "help" with homework is bad.

Don’t let people tell you AI killed the academic essay. We ran a structured essay process in class this year, with successive problem finding, literature search, multiple drafting, and peer review sessions. It worked, and the vibe in class was the best I’ve seen in a long time. Repeat next year.

Announcing the first industrial application of SYNTH: we trained a 600m reasoning model for one of the largest infrastructure in the world, the subway of Paris. pleias.ai/blog/sillon-...

Neat

@author-calnewport.bsky.social does a fantastic job writing about AI, and this one is his best yet. Gift link in quoted post.

How uninformed you have to be to suggest this prompt... “Create a research grant proposal focusing on [topic]. Align it with [funding agency] guidelines (linked) and ensure the proposal includes a robust literature review, detailed methods section [...]” academy.openai.com/public/clubs...

Last week, the 46th edition of 𝗧𝗮𝗯𝘂𝗗𝗔𝗚 took place, the long-standing linguistics conference organized by the Center for Language and Cognition (CLCG) at the @rug.nl. This year, I had the pleasure of being part of the organizing committee!

you're telling me a desk rejected this paper

new post: how I develop recently using local models. the tooling is now good enough to do agentic workflows and everyone should give them a try! vickiboykis.com/2026/06/15/r...

It's been a long time since I reconnected with my interpretability past; here's our new analysis of the phenomenon of "detokenization": models reconstruct full word representations ("error") at the last state of the subword-tokenized location ["err", "or"].

Doing the most responsible thing an European AI labs can do after this weekend: shipping a blogpost. Why the EU can't into AI, how it's not about compute, but actual skill issue and failing for years to build an actual training ecosystem. pleias.ai/blog/fable-eu

word is there’s going to be a new mistral model

I am back to San Francisco after three weeks of driving around the western US. I drove 4,000 miles and learned absolutely nothing about B2B SaaS. You should consider setting your Claudes down and get the fuck out of the Bay sometimes. There is much more to life.

LLMs are put into a metaphorical isolation chamber, given a dataset, and are instructed to make predictions about the next token over that dataset. And this raises interesting questions about learnability, but it seems to undermine what makes UBL so interesting in the first place - the interaction.

Fable is being taken down? IMMEDIATELY?? www.anthropic.com/news/fable-m...

Oh my lord

New International Construction Grammar Association (ICGA)! icgaonline.wpcomstaging.com announced at ICCG14 thanks to @adelegoldberg.bsky.social Sign up forms.gle/gXYDKC6xavkU... @alexbergs.bsky.social H Boas M Fried, @stefanhartmann.bsky.social Th Hoffmann F Masini B Nir @remivantrijp.bsky.social

Now published in open access! Your one-stop shop for the philosophy of language models. It's the spiritual descendant of our two-part preprint from 2024, fully updated. This should be particularly useful for anyone looking for an entry point into this rapidly growing field.

Academic writing should be addressed neither to the public nor to other academics. It should be addressed to a single interlocutor with whom you have a petty rivalry and whose work you are trying to discredit

“Dimicillin” isn’t real. We made it up. Yet many LLMs still call it an antibiotic. Across 9 models and 653 drugs, we find that drug-name affixes alone can drive pharmacological reasoning. Models often rely on morphology over facts. We trace this shortcut from behavior to mechanism. 🧵

In film, "we'll fix it in post" is what you say when something went wrong on set and you don't want to redo it. AI research has made it our entire methodology: train the model, then patch whatever comes out. Our new ICML oral argues this can't be the basis of a science of AI. 🧵

Post-doctoral Research Associate in Computational Linguistics. King's College London. Salary: £45,031 to £47,379 www.jobs.ac.uk/job/DRU560/p...

We acknowledge that manual coders and engineers, including C, Haskell, and Java speaking peoples, have stewarded through generations the compilers and datasets of what is now Claude Code. We honor and respect the enduring relationship that exists between these peoples and nations and this LLM.

Open call: has anyone managed to make significant use of chatbot LLMs or agents for open-ended research & analysis in a domain that is *not* SWE/coding-heavy or math, while also intending to make the final work public? (Interpret "significant use" as you wish - this is anecdotal, no wrong answers)

I so agree that usage phenomena explorable thru LLMs go beyond frequency. In my ICCG plenary on verbing nouns, I showed LLMs can model both figurativity, via ‘source-noun alignment,’ & the rel btwn figuration & syntactic complexity—the latter a bell curve peaking at metaphorical tokens!

There's also a less technical but more wide ranging version here, if either are of interest. vincentcarchidi.substack.com/p/do-languag...

In any event, I have a forthcoming piece targeting exactly that cross-talk, and I argue that no computational models that actually exist, including LLMs, can test for the issues that actually divide generative and UBLists (UG, domain-generality vs. specificity, etc). philarchive.org/rec/CARTFO-47

Since last week's (totally awesome btw) ICCG14 in Princeton, I can't stop thinking about LMs and linguistics. I know this topic has been discussed to death, so pls bear with me (long thread incoming). 1/n

At #ICCG14 in Princeton (huge thanks to @adelegoldberg.bsky.social & team for a brilliant conference!) we announced our plans for the International #ConstructionGrammar Association (ICGA), to be officially founded at #ICCG15 in Kiel in 2028. See here for more information: construction-grammar.org

One final pre-conference coffee before ICCG starts ☕️

Excited to present at ICCG14, check out the program for a lot of cool works by a lot of cool people: iccg14.oa-event.com/program

Requesting that everyone call me Dr. Parameter-Efficient Korean Character-Level Language Modeling for now on. I'm honestly so disappointed in spammers these days. No excuse in the age of LLMs.

Hi language science bluesky! If you have 20 minutes, PhD student Dhanya Charan (dhanya.humspace.ucla.edu) and I are recruiting volunteer participants for a 20-minute speech production study! Link here: www.labvanced.com/player.html?... Please re-skeet / share widely! Thank you!

Attending ICCG14? See the poster “Constructions in German child-directed and child-available language” ― presented by @bbunzeck.bsky.social We argue that previous studies have not sufficiently acknowledged the breadth and variation in child-directed and child-available input. #CRC1646 #LINCC

@edithkaan.bsky.social @iwinther.bsky.social and I are editing a special issue "Bridging psycholinguistic and computational approaches to bilingualism". Have a look at mmmm2026.github.io/LAB/ if you're interested!

Soon, herds of feral cleaning robots roaming the streets of Canterbury. They have no natural predators