Kyle Mahowald
@kmahowald
UT Austin linguist https://t.co/1GaRxR8rOu. cognition, psycholinguistics, data, NLP, crosswords. He/him.
ID:22515678
http://mahowak.github.io 02-03-2009 18:28:50
511 Tweets
1,6K Followers
722 Following
Congrats to Venkat for successfully defending his thesis today! David Beaver and I are so proud of his excellent work.
Venkat will join Ithaca College as a faculty member this fall 🎉
Huge thanks to friends, collaborators & committee members Kyle Mahowald and Malihe Alikhani
Have you ever wondered about what's the impact of having so many near duplicate subwords in your model's vocabulary (e.g., _book vs book vs books vs Book 📚) on its performance?! If yes, checkout Anton Schäfer's new paper on it! :)
'Conversations with AI,' by Bennie Mols, looks at the promises and pitfalls of using #LLMs as tools in the study of #languages . bit.ly/3Vy2zgh
Do LLMs really need to be so L?
That's a rejected title for a new paper w/ Andrey Gromov, Kushal Tirumala, Hassan Shapourian, Paolo Glorioso on pruning open-weight LLMs: we can remove up to *half* the layers of Llama-2 70B w/ essentially no impact on performance on QA benchmarks.
1/
Dissociating language and thought in large language models
Feature Review by Kyle Mahowald (Kyle Mahowald), Anna Ivanova (@neuranna), Idan Blank (@IbanDlank), Nancy Kanwisher (@Nancy_Kanwisher), Joshua Tenenbaum, & Evelina Fedorenko (@ev_fedorenko)
doi.org/10.1016/j.tics…