Language Generation Day
A day-long workshop on language generation in the limit.
Registration is free but RSVP is required.
Overview
When is language generation possible from examples? Language generation in the limit studies this question at a foundational level: given samples from an unknown language, can we generate new strings that are valid and have not already been seen in the data?
This is a growing area, which already includes a rich set of extensions (including stronger and more diverse notions of generation and algorithms robust to contamination in data), while building bridges to other parts of theoretical CS, learning theory, and mathematics.
This workshop brings together Stanford students and researchers interested in these foundations and recent developments, with short talks and discussion on emerging directions and open problems.
Program
All times PST. Schedule is subject to minor updates.
| Time | Session |
|---|---|
| 09:30–10:00 | Coffee |
| 10:00–11:00 |
Introduction to Language Generation in the Limit
Jon Kleinberg (50 min talk + 10 min Q&A)
|
| 11:00–12:00 |
Stronger Notions of Generation: Uniform and Non-uniform
Chirag Pabbaraju (25 min + 5 min Q&A)
Mistake-Bounded Language Generation
Charlotte Peale (25 min + 5 min Q&A)
|
| 12:00–13:00 |
Lunch (self-catered)
Suggested option: Greenfish (nearby venue)
|
| 13:00–14:00 |
Generation in Presence of Contamination
Anay Mehrotra (25 min + 5 min Q&A)
Beyond Countable Collections: With and without Feedback
Grigoris Velegkas (25 min + 5 min Q&A)
|
| 14:00–14:45 |
Discussion session
Open problems and future directions
|
| 14:45–15:00 |
Closing comments
|
Speakers
Talk titles and abstracts will be posted here as they are confirmed.
Resources
An actively updated collection of papers related to this topic is available here: languagegeneration.github.io