Mark Dredze(@mdredze) 's Twitter Profileg
Mark Dredze

@mdredze

John C Malone Professor at @JohnsHopkins @JHUCompSci @jhuclsp @jhumceh; Part time @techatbloomberg (tweets my own) Mastodon @[email protected]

ID:757339410

linkhttp://www.dredze.com calendar_today14-08-2012 15:20:23

1,8K Tweets

4,1K Followers

786 Following

Dr. Sue Vazakas(@SuetheLibrarian) 's Twitter Profile Photo

Congratulations to Mark Dredze on a great Inaugural Professorial Lecture! Glad to hear about his values; e.g., diversity, humility, priorities; and his hopes for the future of the DSAI Initiative. engineering.jhu.edu/Datascience-AI/ Johns Hopkins Engineering JHU Computer Science

account_circle
Johns Hopkins Engineering(@HopkinsEngineer) 's Twitter Profile Photo

Today at 4:30 in Shriver Hall: Join us for an Inaugural Professorial Lecture honoring Mark Dredze, John C. Malone Professor, JHU Computer Science. 'What I’ve Learned (So Far) in 15 Years at Johns Hopkins.' You can watch LIVE on YouTube! eventbrite.com/e/inaugural-pr…

account_circle
Suzanna Sia(@suzyahyah) 's Twitter Profile Photo

1. Problem:

Base LLMs can perform zero-shot Translation, but are poorly calibrated and sometimes “fail to translate”; i.e. they continue generating in the source language or produce empty generations.

2. Proposed solution
3. Limitations
4. Additional
5. Concurrent Work

1/5

account_circle
Mark Dredze(@mdredze) 's Twitter Profile Photo

The deadline for abstracts is coming up soon! This is a great event for students in the Mid-Atlantic area.

account_circle
elvis(@omarsar0) 's Twitter Profile Photo

Aligning LLMs to Quote from Pre-Training Data

Interesting approach to enable verifiability in LLMs.

Proposes techniques to align LLMs to leverage memorized information quotes directly from pre-training data.

Looks like the alignment approach is not only able to generate…

Aligning LLMs to Quote from Pre-Training Data Interesting approach to enable verifiability in LLMs. Proposes techniques to align LLMs to leverage memorized information quotes directly from pre-training data. Looks like the alignment approach is not only able to generate…
account_circle