Fantastic research group meeting today at Johns Hopkins University Bloomberg Center. We then went to see the Cherry Blossoms. We also saw Stumpy!

Stumpy inspired meme. Credit to my PhD students for the idea.

📢 MASC-SLL Call for Papers is out! 📢 🎓 Are you a student passionate about Speech, Language, and Learning? 🗣️✨ 🌟 Present your research at the Mid-Atlantic Student Colloquium on Speech, Language, and Learning (MASC-SLL) at Johns Hopkins Johns Hopkins Engineering on May 3rd! 🌟

Who wore it better? Credit to Alexandra DeLucia

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

JHU at MASC-SLL 2024

Help review more ARR submissions if you can! Otherwise, this is the reality we face.

NAACL goes from Mexico to New Mexico!

‼️Can we predict LLM performance on an unseen language if we know its linguistic relationships with a seen language? What makes an LRL difficult for an LLM?‼️ We present a study of LLM performance degradation as a function of linguistic distances... arxiv.org/pdf/2406.13718

If you write a paper review and provide a list of weaknesses, and the author's response addresses those weaknesses, raise your score. Otherwise, you need to better explain your concerns.

At what point in my CS PhD do I magically stop running into the `install package > run script > package not found > install package …` loop? 🙃

I will be at #COLM2024 on Mon and Tue. Find me or my other JHU CLSP colleagues to talk about opportunities at Johns Hopkins Engineering. We have faculty, postdoc and PhD positions. Benjamin Van Durme Daniel Khashabi 🕊️

Me, pushing off everything until "after the ARR deadline" ACLRollingReview

I’m presenting my Microsoft internship work today at #EMNLP2024! I’ll be in Riverfront from 2-3:30. Come say hi and chat about structured document generation and evaluation! TLDR: We unify the generation and evaluation of templatic views of documents. aclanthology.org/2024.findings-…

Congratulations to Carlos Aguirre Carlos Aguirre Pocasangre on successfully defending his PhD thesis: Improving the Fairness in Language Models for Decision-Making! Next stop? Amazon pocaguirre.com

PhDone! 🎉 Thanks to everyone who was there along the way, family, friends, colleagues and advisors! Yes, I am a doctor now. Yes, I got a sword now (thanks friends!) Yes, we got Mark Dredze to knight me. What a nice finishing touch for this era!

MASC is a great opportunity to interact with fellow NLP/Speech/AI researchers, please consider hosting at your university!

🎤Invited Talks at MASC! Roger Beaty Beyond Generation: Language Models as Creativity Evaluators Xiang Lorraine Li - Every Opinion Matters: Distributional and Long-tail Evaluation for LLMs Daphne Ippolito – Troubles with Training Data for Large Language Models #NLP #MASC2025

1/6🤖🩺 Medical answers are context-dependent, hypothetical, and subjective (If pain gets worse, see a doctor). Existing systems break here: they generate invalid claims or omit these sentences. MedScore solves this with context-aware claims and verifies against trusted sources.

I will present MedExpert, a clinician-annotated Factuality and Completeness dataset with Severity levels for Medical Chatbot Evaluation at ML4H in San Diego today. Happy to chat from Dec 1 to 3! See paper at: openreview.net/pdf?id=rkLAzDP… Mark Dredze Alexandra DeLucia JHU CLSP Sonal Joshi