What more exciting is Nikhil will start his PhD at Hopkins and work together with me! Many more to come 🥳 Congrats to the team Nikhil Sharma Vera Liao

Happy to share that we received the Amazon Research Award to study active learning and model calibration under subpopulation shifts!

🎉Exciting News! 🎉 Delighted to receive the Best Paper award at #CHI2024 📝🏆 for our work on Generative Echo Chambers, where we show the effects of LLM-powered conversational search on opinion polarization!

Congratulations, and thanks to Vera Liao Ziang Xiao
#LLMs #NLProc #HCI

Ziang Xiao Nikhil Sharma Vera Liao Congrats to all of you!!

Congrats to the 2024 Singhal Family Entrepreneurship Award winners CounselAI and Bamboo Resumes! In collaboration with the Pava Marie LaPere Center for Entrepreneurship, the awards help launch student-run software ventures w/ resources to translate innovative concepts into viable products. cs.jhu.edu/news/new-stude…

A multidisciplinary student team (feat. Himanshi Sharma) designed an AI-powered chatbot to answer cancer patients’ questions accurately and empathetically. See it in action at #DesignDay2024 !
cs.jhu.edu/news/an-ai-tha…

I'm typically impressed by the Community Notes I see on the platform but was particularly surprised to see their accuracy mirrored (97.5% agreement) ID docs and docs in training when discussing vaccination. Just look at these examples. I want more notes on X and every social…

In late 2022, X (formerly Twitter) introduced Community Notes, a crowdsourced misinformation countermeasure. Study evaluated the topics, accuracy, and credibility of these Community Notes addressing COVID-19 vaccination. ja.ma/4aOce78

'what is much more important is why we spend ~120k GPU hours on the H100 cluster to prepare/share a ... dataset?'

The impact of data quality is huge —— I wonder if we have a fixed number of GPU hours, how much should we spend on filtering data and how much on model training?

What frameworks are people using to train SOTA dense retrievers? Tevatron and SentenceTransformers seem dated these days… has no one open-sourced anything new?

Leo Boytsov search founder Daniel Campos Rui Meng ̷̨̨̛̛̲͎̣̞̘͈̯͚͂̈́̂̄̽̎́̽̔͑̄̏̽̏͒̾́̅̐̈́̾̎̆͆̽́͌̽̀̚̕̚̕͠͝͝ Bo

See how Hopkins engineers are using #AI to ensure equity in health care, featuring John C. Malone Associate Professor of Electrical and Computer Engineering Bisi Bell and Center affiliate Axel Krieger. Watch here: youtube.com/watch?v=eZR0Rs…

We're heading to the Johns Hopkins University Bloomberg Center on Wed. May 8 to celebrate 10 years of the DATA Act. Written by GovEx Executive Director Amy Holmes, it's the country's first open data law.

Register today to join us for reflections on the last decade:
hub.jhu.edu/events/2024/05…

See how Hopkins engineers are using #AI to extend human capabilities, ensure equity in health care, enable a new age of space exploration, and more. Featuring CS’ Anjalie Field! 🥳 Watch here: youtube.com/watch?v=eZR0Rs…

#ArtificialIntelligence seminar with Sarah Cen on Thursday! Learn more here: cs.jhu.edu/event/cs-semin…

The IWSLT 2024 submission portal is open!
We welcome your paper submissions on all aspects of speech translation.

We look forward to seeing you in Bangkok!

🗓️ Deadline: Mon Apr 29
🔗 Softconf: softconf.com/n/iwslt2024/
🛈 More info: iwslt.org/2024/

Hyp-OC: Hyperbolic One Class Classification for Face Anti-Spoofing, has been accepted by FG2024
w/ Vishal Patel

Paper: arxiv.org/abs/2404.14406
Code: github.com/Kartik-3004/hy…
Project Website: kartik-3004.github.io/hyp-oc/

#FG2024

Yinzhi Cao is part of a team selected by DARPA to develop cyber reasoning systems using #ArtificialInteligence technology. Learn more: cs.jhu.edu/news/yinzhi-ca…

Check out our new work MaxFusion where we

✅ Unlock scaling diffusion models to new modalities without training!
✅ Easily bring in effects from multiple conditions that otherwise need descriptive captions.
✅ Tackle the need of paired training data for multi-task conditioning.