researching language as a cognitive technology Faculty Fellow at NYU Center for Data Science 💬 🧠 💻 🗼 PhD @ MIT Brain&CogSci former HS math and CS teacher
Joined December 2013
- Tweets115
- Following309
- Followers900
- Likes203
Pinned Tweet
If you've noticed a distinct dialect emerging in frontier AI models and struggled to understand what they're talking about, you're not crazy!
We turned to psycholinguistics to investigate what makes the writing style so weird, and why this might be happening
🚨New preprint! 🧵
We demonstrate this with a proof of concept in which we fine-tune a Qwen model with a length penalty. It converges towards a categorical preference for brevity in alternations ("he sent us a photo" > "he sent a photo to us"), which spills over to untrained examples.
Joint work with @kmahowald and Matias Martin at UT. Check out our pre-print: arxiv.org/abs/2610.29162
At #COLM2026 in SF -- would love to see you at my poster "We Hebben Een Serieus Translatie: Modeling Intercomprehension as Probabilistic Inference", Thursday 4:40pm -- work from my PhD with @LanguageMIT and @roger_p_levy. Or let's grab coffee!
Thomas Hikaru Clark retweeted
Thanks so much for the fun and stimulating conversation Jekaterina, @WiAIR_podcast !
Such a pleasure to chat about language, LLMs, and memory--covering some work with @bkhmsi @martin_schrimpf @Mikey_Lepori @KlemenKotar @ev_fedorenko @thomashikaru, among others!
New episode of #WiAIRpodcast: @GretaTuckute (@KempnerInst @Harvard) talks to @J_Novikova_NLP about what brain-LLM alignment does and does not tell us, and why, asked to keep one label out of neuroscience, cognitive science and AI, she drops AI first
Thomas Hikaru Clark retweeted
How do you read a menu in a language you've never studied?
A warm welcome to @thomashikaru, joining CDS this fall as an incoming Faculty Fellow.
He models how people handle typos, language barriers, and other places language breaks down.
nyudatascience.medium.com/me…
Thomas Hikaru Clark retweeted
Words and sentences that are highly dissimilar from anything we’ve seen before are more likely to be remembered accurately, according to research from @ev_fedorenko @GretaTuckute @bj_mdn @thomashikaru @LanguageMIT.
mcgovern.mit.edu/2025/10/01/… @mitbrainandcog @ScienceMIT
What makes some sentences more memorable than others? Our new paper gathers memorability norms for 2500 sentences using a recognition paradigm, building on past work in visual and word memorability.
@GretaTuckute @bj_mdn @ev_fedorenko
We release both aggregate and trial-level data (including memorability measures, reaction times, and a variety of word- and sentence-level predictors): github.com/thomashikaru/sent…
1/7 If you're at CogSci 2025, I'd love to see you at my talk on Friday 1pm PDT in Nob Hill A! I'll be talking about our work towards an implemented computational model of noisy-channel comprehension (with @postylem, @LanguageMIT, and @roger_p_levy).
6/7 We release our model's code on GitHub: github.com/thomashikaru/nois…
7/7 Check out our CogSci proceedings paper for more details, and stay tuned for updates! Thanks to all who provided their feedback :) escholarship.org/uc/item/9kr…