Alexey Koshevoy
@alexeykoshevoy.bsky.social
Postdoctoral researcher | LSCP, ENS-PSL | Interested in how communication and cultural evolution jointly shape languages https://alexeykosh.github.io
Not to brag, but I've been working with the author of #9 on this list (the fantastic @oliviermorin.bsky.social ) for years! This is exactly the kind of piece I've been missing since edge.org died out.. www.nytimes.com/2026/10/05/s...
What Are the Biggest Questions in Science Today? 100 Scientists Weigh In.
nytimes.com
A much needed thread!
moin syed@syeducation.bsky.social · last wk.Gave a talk in my dept. on the OSF transition. I wrote up a rough script of the session, which could be of interest to others. I synthesize existing resources and provide my own perspectives. Comments and corrections welcome! zenodo.org/records/2306...
We used virtual mates and genetic algorithms to ask swordtail fish to design their ideal mate. Males all liked larger females but females did not share a clear preference for male appearance. Led by @kathrynbullough.bsky.social @grahancock.bsky.social in @royalsocietypublishing.org shorturl.at/rCeR2
Our latest Spotify playlist: 13 of our episodes that explore different aspects of language! open.spotify.com/playlist/1Qx...
Many Minds explores language
Playlist · kensycoop · 13 items
open.spotify.com
2023: 8,039 proposals 2024: 10,360 proposals 2025: 17,066 proposals 2026: 21,627 proposals Something's gotta give
Iva Vukušić@vukusiciva.bsky.social · 3w agoThis ridiculous number of applications is, in my view, the result of 2 nightmare developments happening concurrently 1.the precariousness of the job market for young PhD graduates which then fuels further: 2.the AI tools enabling overproduction of proposals. #Research #ERC #ResearchFunding #Academia
I am reading a lot of modelling papers lately, and I have one advice to everyone creating figures with large amounts of observations. Please, rasterize your images, it takes a while to load 1000+ vectorized scatter points in a pdf!
Ever feel like it's too hard to keep track of what LLMs cannot do as well as humans? We're making your life easier over at: what-llms-can-not-do.github.io We're compiling a list of papers testing the abilities of LLMs against humans. Check it out! And you can help contribute too!
What LLMs Can(not) Do
A living survey of benchmarks that compare large language models with humans.
what-llms-can-not-do.github.io
Out today! www.nature.com/articles/s41...
tal boger@talboger.bsky.social · 5mo agoComplexity is ubiquitous in the world around us. But different stimuli are complex for different reasons. Does the mind nevertheless represent a ‘unified’ notion of complexity across domains? In @nathumbehav.nature.com, @chazfirestone.bsky.social & I show the answer is: yes! osf.io/preprints/ps...
My lab has been really interested in topological perception the past two years. In a new paper, we ask: How does a topological perspective (re-)shape our understanding of visual complexity? Work led by one of our many wonderful undergraduate students, Ashna Shah! link.springer.com/article/10.3...
Topological structure and the creation of visual complexity - Psychonomic Bulletin & Review
What does it mean for something to be complex? Fundamentally, complexity is a cue that something is less easily represented by our minds. With this fact in view, complexity becomes a tool: By understa...
link.springer.com
Researchers investigated language change by studying Slovotvir, a Ukrainian website where users propose and rank alternatives for foreign-origin words. The site shows that people prefer shorter words, but not necessarily more common words. In PNAS Nexus: https://ow.ly/KWGE50ZNzaj
How does the architecture of a city shape the way we think, talk, and draw space? And how do spatial conventions travel across cultural generations? New paper in Open Mind, spearheaded by Z Alinam (w J Nölle, C Vesper, K Tylén & me): doi.org/10.1162/OPMI.a.380 🧵 1/
When the City Speaks: How Urban Form Shapes the Cultural Transmission of Spatial Conceptualization
Abstract. How do urban environments shape the way people conceptualize and communicate space? Using a cultural transmission experiment in virtual reality, we asked participants to navigate one of two ...
doi.org
After years of analyzing thousands of drawings from hundreds of adults and children, I am excited to share my first ever paper (w/ @samiyousif.bsky.social )! Using a simple drawing task, we show how topological structure is represented in the mind and its importance in spatial memory.
Topological structure in human spatial representation revealed through drawing
elifesciences.org
Millennium math problem solved by OpenAI may have "regurgitated" competitors unpublished work... also they spent more than $22 million in computing power in one week to scoop them. techcrunch.com/2026/09/08/o...
OpenAI fought dirty on career-making math problem, says NYU mathematician | TechCrunch
There is a $1 million bounty for the first person providing a solution to the Navier-Stokes existence and smoothness problem.
techcrunch.com
📢 We're looking for a Master's student to join our team in sunny Aix-en-Provence 🇫🇷 for a *paid* 5-6 month research internship, anytime between now and June 2027, to help replicate an influential behavioural study on inner speech! Details: tinyurl.com/bdcthbp8 #psychjobs #neurojobs #innerspeech
tinyurl.com
Are you interested in lexical competition, agent-based modelling, strong inference and Ukrainian? Then this paper, just published in @pnasnexus.org, is for you. With @oliviermorin.bsky.social and @sobchuk.bsky.social, we asked a simple question: why do some words become more common than others?
Hear! Hear!
Mason Youngblood@masonyoungblood.bsky.social · last mo.New tutorial preprint about simulation-based inference! Shows how to fit a cultural evo model to baby name data using approx Bayesian computation, random forests, and deep learning. Wrote for @eslr.bsky.social but wanted to share with others in @culturalevolsoc.bsky.social! :) osf.io/preprints/so...
🤖🧠NEW PAPER🧠🤖 (The result of an 8-year project!) LLMs seem very different from symbolic systems. Yet LLMs excel in symbolic domains (e.g., language/code/math). How do they do it? Our finding: LLM representations have implicit symbolic structure! Link in thread ⬇️ 1/n
![Overview of the paper.
Title: The Emergent Symbolic Structure of Artificial Neural Networks
Authors: Tom McCoy, Paul Soulos, Tal Linzen, Paul Smolensky
Left: Neural networks encode information in vectors (there is then an image of a vector), yet they excel at tasks long thought to require symbolic structure (there is then an image of a symbolic representation, specifically a syntax tree). How do LLMs do it?
Right: We find that LLM representations can be closely approximated with symbolic structures. This approximation lets us edit the structure of an LLM’s output by editing the structure of its internal representations, as shown. There is then an image of two edits to LLMs. In the first one, the original input is 3 + 6 * 8, with an answer of 51. But if we swap the positions of the 3 and the 6, the output becomes 30. In the second one, the original input is a Python command repeating the list [Z, U] three times, producing [Z, U, Z, U, Z, U]. But if we edit the input in a way that adds a Q at the end of the input, the output becomes [Z, U, Q, Z, U, Q, Z, U, Q].](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:bcsvlv7ubeagj5virdvatadd/bafkreic3zm2rpc4vtqrz2t3y4vcl3t2owp2sxay42vuh5zm5zkg4jhr64e)
Schrodinger’s language models LLMs randomly choose between possible words, and certain choices can lead to wildly different results Goodfire’s latest work shows that those decisions largely end in roughly the same result. arxiv.org/abs/2608.19611
I am currently recruiting for a 2-year postdoc/research fellow role at the University of Auckland here in New Zealand, focusing on auditory perception, hearing, and music in older adults, including people with Alzheimer's disease or other dementias ~NZ$100k/yr, flexible start date, please apply!
Postdoc in The Music Lab // Expression of interest
Note. The below text is a general overview of the sorts of postdocs we often recruit. We are currently recruiting candidates for a 2-year position focusing on auditory perception in aging populations,...
forms.gle
Interested by a corpus of 19th/early 20th c. French for your research? Here's the French corpus of 11K! novels I tried to curate during my PhD 😅, now live at JOHD, and downloadable on Zenodo! 📄 doi.org/10.5334/johd.597 📦 zenodo.org/records/20233178
Hors d’usage: Curated French Fiction from Gallica, 1800–1950 | Journal of Open Humanities Data
doi.org
A 2025 Science paper reported that bonobos combine calls to create new meanings. We analyzed its stats and then reanalysed the data. --> The claim did not hold up. Our paper (open access): doi.org/10.7717/peer... With @johanlind.bsky.social @annajonand.bsky.social
How many languages have existed over the Holocene—and what does that reveal about the design space of languages and cultures? Now out in @science.org www.science.org/eprint/QDZNY....
The rise and fall of language diversity through the Holocene
Characterizing the factors that have shaped linguistic diversity is fundamental for understanding human history, culture, and cognition. In this study, we combined statistical and social computational...
science.org
📣 I'm hiring! Two positions (PhD student & Postdoc) in my ERC project Macroevolution of European Literature. Let's study the cultural evolution of literature using massive data 📚📈 📍Frankfurt | Apply by August 15 Details: www.aesthetics.mpg.de/en/career/jo... Please repost to help spread the word!
Postdoc position @unimarburg.bsky.social in the project: "Bridging the Gap Between Verbal Psychological Theories & Formal Statistical Modeling with Large Language Models" (funded by @volkswagenstiftung.de) 📅Start: 01.10.2026 |⏳3 years 🔗 Job posting: uni-marburg.de/78NrWT Thanks for sharing!
"CLDF Meta" (meta.clld.org) links to data from over 800 CLDF datasets that have been released on Zenodo (e.g. WALS, APiCS and Grambank). There are links to data on over 9000 languoids, e.g. 98 entries on Ambulas (to take a random language). Great work by my colleague Johannes Englisch!
Scientists are asking if it might be unconscious
The Independent@the-independent.com · 4mo agoClaude down: Anthropic AI not working in major outage
Multiplier par 15 les frais d’inscription universitaires des étudiant·es étrangèr·es non européen·nes. Piétiner l'égal accès aux services publics. Appeler le programme "Bienvenue en France" 🇫🇷🥖 🔴 Signez et faites signer la pétition, ça prend 2 min : petitions.assemblee-nationale.fr/initiatives/...
Pour la suppression du programme Bienvenue en France mettant en place des frais différenciés pour les étudiant·es extra-européen·nes - Pour la suppression du programme Bienvenue en France mettant en p...
La stratégie Bienvenue en France prétend renforcer l'attractivité internationale de la France en multipliant par 15 les frais d’inscription à l’université des étudiant·es non européen·nes. Si les exon...
petitions.assemblee-nationale.fr
With big thanks to Alexey Koshevoy, @alexeykoshevoy.bsky.social Marie Hallo @marie-hl.bsky.social and Rowan Hall Maudslay!
Alexey Koshevoy@alexeykoshevoy.bsky.social · 5mo ago🚨New paper alert!🚨 In our new paper with @oliviermorin.bsky.social, Rowan Hall, and Marie Halo published in the Journal of Language Evolution, we explored the relationship between word length and ambiguity on a large-cross linguistic dataset. 1/6
New post: Tagging my blog with BERTopic and LLMs. I suspect as token costs increase, we'll see more blended traditional ML/LLM systems. And, in general, they work really well! vickiboykis.com/2026/05/18/t...
Tagging my blog posts with BERTopic and LLMs
LLMs mean you still need a human in the loop, but in a different part
vickiboykis.com
