Lightnews — Scholar-powered news

LightNews

Aryaman Arora

@aryaman.io

2.8K followers 480 following 24 posts

member of technical staff @stanfordnlp.bsky.social

Posts Media Videos Starter Packs

Reposted by Aryaman Arora

juliekallini.bsky.social @juliekallini.bsky.social · Jan 14

"Mission: Impossible" was featured in Quanta Magazine! Big thank you to @benbenbrubaker.bsky.social for the wonderful article covering our work on impossible languages. Ben was so thoughtful and thorough in all our conversations, and it really shows in his writing!

Quanta Magazine @quantamagazine.bsky.social · Jan 13

Large language models may not be so omnipotent after all. New research shows that LLMs, like humans, prefer to learn some linguistic patterns over others. @benbenbrubaker.bsky.social reports: www.quantamagazine.org/can-ai-model...

Can AI Models Show Us How People Learn? Impossible Languages Point a Way. | Quanta Magazine

Certain grammatical rules never appear in any known language. By constructing artificial languages that have these rules, linguists can use neural networks to explore how people learn.

www.quantamagazine.org

Reposted by Aryaman Arora

Chris Potts @cgpotts.bsky.social · Jan 13

I've posted the practice run of my LSA keynote. My core claim is that LLMs can be useful tools for doing close linguistic analysis. I illustrate with a detailed case study, drawing on corpus evidence, targeted syntactic evaluations, and causal intervention-based analyses: youtu.be/DBorepHuKDM

Finding linguistic structure in large language models

YouTube video by Chris Potts

Aryaman Arora @aryaman.io · Nov 26

What are the broad open problems in your view?

Aryaman Arora @aryaman.io · Nov 25

i am now going to write a massive reply that will have no effect on this score you have given me

Aryaman Arora @aryaman.io · Nov 25

hmm bluesky feed is 80% reviewing complaints. twitter is slightly better in this regard

Aryaman Arora @aryaman.io · Nov 25

I was reading Tim Bodt's new book on Proto-Western Kho-Bwa while waiting for code to run

www.ling.sinica.edu.tw/item/en?act=...

very nice work

LANGUAGE AND LINGUISTICS >

www.ling.sinica.edu.tw

Aryaman Arora @aryaman.io · Nov 25

e.g. Proto-Western Kho-Bwa *n̥a-jʷa-kʰa "chin" (> Khoina Sartang nyjukʰu) seems to be composed of

- PWKB *n̥a- "lower face" < Proto-Sino-Tibetan *s-na ~ s-naːr
- ?PST *g(j/w)ar "cheek, chin, jaw"
- PST *m/s-k(w)a-j "mouth, opening"

very interesting derivational morphology strat

Aryaman Arora @aryaman.io · Nov 25

it's pretty interesting how (some?) Sino-Tibetan languages, when they historically underwent too much degradation through sound change, just decided to make compounds with synonymous elements or stick semantic prefixes on everything

Aryaman Arora @aryaman.io · Nov 24

The cat is the mech interp researcher right

Aryaman Arora @aryaman.io · Nov 23

@avzaagzonunaada.bsky.social

Aryaman Arora @aryaman.io · Nov 22

added 🫡

Aryaman Arora @aryaman.io · Nov 22

added 🫡

Aryaman Arora @aryaman.io · Nov 22

made this thing, reply to be added
go.bsky.app/AKGJ82V

Aryaman Arora @aryaman.io · Nov 21

sus

Aryaman Arora @aryaman.io · Nov 20

caleb is so cool pls follow

Caleb Ziems @calebziems.com · Nov 19

@butanium.bsky.social I nominate @aryaman.io

Reposted by Aryaman Arora

Juan Diego Rodriguez (@ COLM 2025) @juand-r.bsky.social · Nov 8

How do language models organize concepts and their properties? Do they use taxonomies to infer new properties, or infer based on concept similarities? Apparently, both!

🌟 New paper with my fantastic collaborators @amuuueller.bsky.social and @kanishka.bsky.social

Title: "Characterizing the Role of Similarity in the Property Inferences of Language Models"
Authors: Juan Diego Rodriguez, Aaron Mueller, Kanishka Misra

Left figure: "Given that dogs are daxable, is it true that corgis are daxable?" A language model could answer this either using taxonomic relations, illustrated by a taxonomy dog-corgi, dog-mutt, canine-wolf, etc., or by similarity relations (dogs are more similar to corgis than cats, wolves or shar peis).

Right figure: illustration of the causal model (and an example intervention) for distributed alignment search (DAS), which we used to find a subspace in the network responsible for property inheritance behavior. The bottom nodes are "property", "premise concept (A)" and "conclusion concept (B)" , the middle nodes are "A has property P", "B is a kind of A", and the top node is "B has property P".

Aryaman Arora @aryaman.io · Nov 14

so true

Aryaman Arora @aryaman.io · Nov 14

the fact that you are posting here (and not there) has significantly increased my desire to use this platform

Aryaman Arora @aryaman.io · Nov 14

i do like my username here

Aryaman Arora @aryaman.io · Nov 14

wow i have so many followers here somehow, is it time to start posting here too

Reposted by Aryaman Arora

Naomi Saphra @nsaphra.bsky.social · Jan 5

I decided what we need to make blueskAI happen is a feed. Reply here to get added to the whitelist! Whitelisted users can post to the feed by adding the following keywords to a post:

🤖
bskAI
blueskAI

Aryaman Arora @aryaman.io · Jan 5

😮

Aryaman Arora @aryaman.io · Jan 5

Hey @noviscl.bsky.social should we start shitposting here too

Aryaman Arora @aryaman.io · Nov 27

lol I think we talked about it the first time I reviewed (SIGMORPHON)

Aryaman Arora @aryaman.io · Jul 26

There will never be a random Burushaski speaker from Pakistan in my mentions here