kanishka.bsky.social - Profile | ThreadSky | a Reddit-style client for Bluesky

exciting to see more language acquisition work shifting the lens to peer communication! @clairebergey.bsky.social is organizing a CogSci symposium highlighting related lines of work this summer!

submitted 1 day ago • 0 comments

Fun new study by Ellie Breitfeld!

submitted 4 days ago • 0 comments

Made a new assignment for a class on Computational Psycholinguistics: - I trained a Transformer language model on sentences sampled from a PCFG - The students' task: Given the Transformer, try to infer the PCFG (w/ a leaderboard for who got closest) Would recommend! 1/n

submitted 5 days ago • 1 comment

not at naacl (nanaacl)

submitted 8 days ago • 1 comment

Six months ago, I accepted a postdoc offer in a fabulous lab. Three days ago, the Trump administration terminated the NSF grant supporting the position. Today, my offer had to be rescinded, and my PhD defense is a week away. (1/3)

submitted 9 days ago • 14 comments

If you're at #NAACL2025 and want to hear about similarity effects for property inheritance in LMs, please stop by! I will be presenting this work on Wednesday at the 11-12:30 poster session on Interpretability & analysis for language models (Hall 3). aclanthology.org/2025.naacl-l...

submitted 9 days ago • 0 comments

If you are at #NAACL2025 @naaclmeeting.bsky.social catch @juand-r.bsky.social presenting our poster on the interplay between similarity and category membership in the property inferences of LMs @ Poster Session 1 on Wednesday! Or if you're at home like me, read our paper: arxiv.org/abs/2410.22590

submitted 10 days ago • 0 comments

Our final South by Semantics lecture at UT Austin is happening on Wednesday April 23!

submitted 16 days ago • 2 comments

Always wanted to test this, and because of Sriram's awesome effort we finally have some results! Check out how LMs respond to suspicious coincidences:

submitted 16 days ago • 0 comments

Had nothing but the best time postdoc-ing with Kyle -- consider applying to work with him!

submitted 16 days ago • 0 comments

One of the ways that LLMs can be inconsistent is the "generator-validator gap," where LLMs deem their own answers incorrect. 🎯 We demonstrate that ranking-based discriminator training can significantly reduce this gap, and improvements on one task often generalize to others! 🧵👇

submitted 21 days ago • 1 comment

Yay - so excited Nick is joining us at TTIC! 🥳 🎉

submitted 23 days ago • 0 comments

Super excited to submit a big sabbatical project this year: "Continuous developmental changes in word recognition support language learning across early childhood": osf.io/preprints/ps...

submitted 23 days ago • 1 comment

The Department of Linguistics at Yale University is seeking to hire a one-year lecturer with expertise in Computational Linguistics and/or Syntax for the 2025-2026 school year. See the job ad for more info: linguistlist.org/issues/36/11... Please send to any who might be interested!

submitted 26 days ago • 0 comments

Does anyone know how to get image representations from the Llama-3.2-11B-Vision models? Specifically a pooled representation of the input image

submitted 29 days ago • 0 comments

✨New paper ✨ Introducing 🌍MultiBLiMP 1.0: A Massively Multilingual Benchmark of Minimal Pairs for Subject-Verb Agreement, covering 101 languages! We present over 125,000 minimal pairs and evaluate 17 LLMs, finding that support is still lacking for many languages. 🧵⬇️

submitted 30 days ago • 3 comments

Cognitive scientists who work across levels of analysis should be called Marr's Rovers

submitted 31 days ago • 0 comments

@kmahowald.bsky.social with a beautiful high-tech illustration 🎨 while describing @qyao.bsky.social's latest paper at the HSP online seminar series! Paper: arxiv.org/abs/2503.20850

submitted 33 days ago • 1 comment

another day another minicons update (potentially a significant one for psycholinguists?) "Word" scoring is now a thing! You just have to supply your own splitting function! pip install -U minicons for merriment

submitted 36 days ago • 3 comments

What might LMs learn about datives if no one gives anything to anyone and no one gives anyone anything? Check out this amazing tour de force work by @qyao.bsky.social!

submitted 37 days ago • 2 comments

LMs learn argument-based preferences for dative constructions (preferring recipient first when it’s shorter), consistent with humans. Is this from memorizing preferences in training? New paper w/ @kanishka.bsky.social , @weissweiler.bsky.social , @kmahowald.bsky.social arxiv.org/abs/2503.20850

submitted 37 days ago • 1 comment

[1/10] Is scene understanding solved? Models today can label pixels and detect objects with high accuracy. But does that mean they truly understand scenes? Super excited to share our new paper and a new task in computer vision: Visual Jenga! 📄 arxiv.org/abs/2503.21770 🔗 visualjenga.github.io

submitted 39 days ago • 7 comments

What a win for BU!!

submitted 41 days ago • 0 comments

Life update: I'm starting as faculty at Boston University @bucds.bsky.social in 2026! BU has SCHEMES for LM interpretability & analysis, I couldn't be more pumped to join a burgeoning supergroup w/ @najoung.bsky.social @amuuueller.bsky.social. Looking for my first students, so apply and reach out!

submitted 42 days ago • 35 comments

Looking for emergency reviewers for two ARR papers on linguistic coordination in LLM agents. Reviews needed within 2 days 🙏

submitted 44 days ago • 0 comments

Require a bunch of emergency reviewers for a paper on personality induction in LLMs -- please DM me if you'd like to review it (turnaround maybe in 2-4 days)

submitted 46 days ago • 1 comment

AANN usage in the wild: “an unbelievable 34000 people”

submitted 46 days ago • 0 comments

Ah yes USA - where two months feel like two years!

submitted 47 days ago • 0 comments

I’ve long used FiveThirtyEight’s interactive “Hack Your Way To Scientific Glory” to illustrate the idea of p-hacking when I teach statistics. But ABC/Disney killed the site earlier this month :( So I made my own with #rstats and Observable and #QuartoPub ! stats.andrewheiss.com/hack-your-way/

submitted 48 days ago • 59 comments

I just realized reml is the reverse of lmer (context: R)

submitted 49 days ago • 1 comment