Now out in Journal of Cognitive Neuroscience! Distributed Sensitivity to Syntax and Semantics throughout the Language Network. direct.mit.edu/jocn/article-a…. A 🧵...

Yay Greta Tuckute and Ev (like in 'evidence', not Eve) Fedorenko 🇺🇦 what a privilege to work with you both on this piece!

GPT4's acceptability judgments on wide range of wh-questions correlate with its semantic judgments of backgroundedness (Nora Graves P '26) and with humans

Talk titles and abstracts are up now for the May 13–14 #NSF-hosted workshop New Horizons in Language Science: Large Language Models, Language Structure, and the Cognitive and Neural Basis of Language! Learn more and register for the Zoom webinar at newhorizonsinlanguagescience.github.io 1/

Have LLMs developed an emergent reasoning capacity, or are they merely mimicking their training data? In this new preprint, we reply to recent commentaries that employed counterfactual analogy tasks, finding that GPT-4 *can* generalize to these new tasks. arxiv.org/abs/2404.13070

thank you, Jason, you will be missed! And yes! we need someone to teach grad stats including mixed models, Bayesian modeling using R and/or Python; stay tuned

This is shaping up to be one awesome Linguistic Society of America Summer Institute in Eugene, Oregon in Summer 2025! Course proposals are due in a good month. There are preliminary commitments from an amazing list of faculty: Come hang out it Eugene, teach a course, and

Prof. Karina Tachihara (UIUC) shows that statistical preemption impacts adult learners of Spanish. Now out in Language Learning! open access: onlinelibrary.wiley.com/doi/10.1111/la…

Princeton PSY seeks Lecturer for grad level stats. roughly 95K, w annual increases. 1 course in F & 1 in Sp w/ 7-13 students Organize guest lecture series and consult Apply here! princeton.edu/acad-positions… Great dept & benefits!

Wow, this is not in theory. On Tuesday we sat with Palestinian moms and heard how hard it is for them and their children, living in a place where settler terrorists make their lives miserable every day. Two days later settlers burned the very house we sat in. Every day terrorism.

The workshop on LLMs and CogSci was one of the most exciting workshops I've ever attended--thanks to Roger and NSF for making everything available!

Ditto! 'Usage-based constructionist approaches and Large Language Models' is out: adele.scholar.princeton.edu/sites/g/files/…

Stanford Linguistics is hiring!!! We have an open area, open rank faculty position! Apply here: facultypositions.stanford.edu/en-us/job/4947…

Thanks Matt!

Linguistics @ Princeton is hiring 2 open-rank faculty: one in syntax/pragmatics/semantics and another in computational work. Modern empirical linguists PLEASE apply! puwebp.princeton.edu/AcadHire/apply… . .

We've now got the full range of data comparing people w GPT4 on semantic judgments of backgroundedness and acceptability judgments of long-distance dependencies: r > .80 Few if any stimuli in training data, most are unacceptable to varying degrees

Congrats to TTIC! Having Kanishka at UT was wonderful for his innovative, super-meticulous work but also his social, joyful, positive energy. He had a HUGE impact in our group and dept. Students: note that [Research [Assistant Professors]] can mentor! And hiring committees:👀

To CoT or not to CoT?🤔 300+ experiments with 14 LLMs & systematic meta-analysis of 100+ recent papers 🤯Direct answering is as good as CoT except for math and symbolic reasoning 🤯You don’t need CoT for 95% of MMLU! CoT mainly helps LLMs track and execute symbolic computation

amazingly clever and insightful work!