CRL Talks
Fall Quarter 2024
CRL Talks are Friday at 11:00 a.m. – 12:00 p.m. (PST, GMT -08:00) in CSB 280 or via Zoom.
November 22
Towards multilingual and linguistically diverse Large Language Models
Ben Bergen
(reporting on work with Tyler Chang, Catherine Arnett, and James Michaelov)
Cognitive Science Department at University of California, San Diego
If Large Language Models (LLMs) are to have their broadest possible scientific and social benefits, they must reflect the world’s linguistic diversity. Yet to date, a small number of languages (particularly English and Mandarin) have enjoyed the most attention and investment in LLM development, and if LLMs are occasionally multilingual, this is usually by accident rather than by design. I will discuss several lines of recent work in my lab that aim to better understand first how multilingual LLMs work and second how to build LLMs for under-resourced languages. We find that multilingual LLMs encode shared multilingual representations for abstract grammatical structures, as well as language-specific ones. We test this by administering a cross-linguistic structural priming task, where LLMs produce similar behavioral effects to human multilinguals. We also find that learning multiple languages influences how models learn each language. For under-resourced languages with relatively little available training data, training LLMs on other languages can produce better outcomes, depending on a variety of factors, including the size of the model and the training sets and the similarity between the languages. Finally, we tackle the finding that LLMs seem to perform better for some types of languages (like fusional languages) than others (like agglutinative languages). We find a surprising explanation for this difference that turns out to have relatively little to do with language typology, and more to do with typography.
CRL Talks Schedule
Oct 11
The unique behavioral profile of high reading efficiency: Evidence from deaf skilled readers
Elizabeth Schotter
University of South Florida
Oct 25
Learning an Artificial Sign Language: Neural Constraints on Cultural Evolution
Seana Coulson & Tania Delgado
Cognitive Science Department at University of California, San Diego
Nov 8
Peekbank: Building a large-scale infant eye-tracking database to understand the development of word recognition
Martin Zettersten
Cognitive Science Department at University of California, San Diego
Nov 16–Nov 17
CAMP7 Meeting
Nov 22
Towards multilingual and linguistically diverse Large Language Models
Ben Bergen
(reporting on work with Tyler Chang, Catherine Arnett, and James Michaelov)
Cognitive Science Department at University of California, San Diego
Nov 29
No Meeting (Thanksgiving Break)
Dec 6
OPEN
CRL Talks are back in person and on Zoom
We look forward to seeing you at Friday at 11 a.m. in CSB 280 or via Zoom. The Zoom link for each talk will be provided in the CRL Talks announcement email. If you are not subscribed to CRL Talks announcements, just sign up to our new Google Group!.
If you still do not want to subscribe to the CRL Talks mailing list, that's okay. Contact the CRL Talks organizer for the details.
Subscribe/Unsubscribe
The CRL Talks is now using Google Groups for posting and distributing announcements. If you were on the old Mailman list, you were automatically moved to the new Google Group. NOTE: You must have a Google account to sign up for the CRL Talks mailing list, but you do not need to sign up with a Gmail address.
All UC San Diego academics, staff, and students:
All UC San Diego academics, staff, and students are automatically provisioned access to Google Groups. Sign into Google with your @ucsd.edu email address and AD password (Duo two-factor authentication required). Navigate to https://groups.google.com/a/ucsd.edu/g/crl-g. Click on the "Ask to join group button" to sign up for the email list.
Sign up for the email list with your personal, work, or school Gmail address:
Log into your personal, work, or school Google account and navigate to https://groups.google.com/a/ucsd.edu/g/crl-g. Click on the "Ask to join group button" to sign up for the email list.
Sign up for the email list without a Gmail address:
Create a Google Account with your non-Gmail address as your username by completing the following: https://accounts.google.com/signupwithoutgmail. Once the new account is confirmed, go to https://groups.google.com/a/ucsd.edu/g/crl-g and click on the "Ask to join group button" to sign up for the email list.
Mailing preferences:
If you want to temporarily disable messages from the CRL Talks list or unsubscribe from the list, access https://groups.google.com/a/ucsd.edu/g/crl-g/membership or see the Google Group help center.