Lightnews — Scholar-powered news

From Queries to Criteria: Understanding How Astronomers Evaluate LLMs

JHU Computer Science @jhucompsci.bsky.social · 2d

and @jwuphysics.bsky.social, @jegpeek.bsky.social, @ziangxiao.bsky.social, @anjalief.bsky.social, and more seek to improve evaluation procedures by building an understanding of how users evaluate LLMs in “From Queries to Criteria: Understanding How Astronomers Evaluate LLMs”: (3/3)

There is growing interest in leveraging LLMs to aid in astronomy and other scientific research, but benchmarks for LLM evaluation in general have not kept pace with the increasingly diverse ways that ...

arxiv.org

Rank1: Test-Time Compute for Reranking in Information Retrieval

JHU Computer Science @jhucompsci.bsky.social · 2d

In “Rank1: Test-Time Compute for Reranking in Information Retrieval,” @orionweller.bsky.social, Kathryn Ricci, Eugene Yang, Andrew Yates, Dawn Lawrie, and Benjamin Van Durme introduce the first reranking model trained to take advantage of test-time 🕰️ compute: (2/3)

We introduce Rank1, the first reranking model trained to take advantage of test-time compute. Rank1 demonstrates the applicability within retrieval of using a reasoning language model (i.e. OpenAI's o...

arxiv.org

JHU Computer Science @jhucompsci.bsky.social · 2d

Happy #COLM2025! Check out some of the research our computer scientists are presenting at this week’s @colmweb.org conference this thread. 🧵 (1/3)

Wiser counsel for public defenders

JHU Computer Science @jhucompsci.bsky.social · 2d

CounselAI allows attorneys to create AI-powered case folders to which they can upload raw discovery materials. Built on @amazonwebservices.bsky.social, the secure platform automatically identifies and organizes documents, summarizes key details, and enables users to pinpoint critical evidence.

An AI-powered digital discovery tool could help public defenders manage overwhelming caseloads.

CS student wins Best Poster Award at IPMI 2025

JHU Computer Science @jhucompsci.bsky.social · 5d

Congratulations, @sremedios.bsky.social!

Information Processing in Medical Imaging is a conference held every two years focused on the fields of applied mathematics, computer science, image processing, and image analysis.

Notre Dame and Johns Hopkins team up to bring digital privacy and AI literacy education to the Smithsonian - Lucy Family Institute for Data & Society

JHU Computer Science @jhucompsci.bsky.social · 6d

Join us for our first Gerald M. Masson Distinguished Lecture of the semester next week! Learn more about the upcoming talk here: www.cs.jhu.edu/event/gerald...

Gerald M. Masson Distinguished Lecture Series: Internet for All—High-Speed Broadband for Every Home in the United States. October 9, 2025, B-17 Hackerman Hall. Refreshments available 10:30 a.m. Seminar begins 10:45 a.m. Henning Schulzrinne, Columbia University.

Reposted by JHU Computer Science

The Lucy Family Institute for Data & Society @lucyinstitute-nd.bsky.social · Aug 29

📲 Researchers from one of the Institute's newest labs, the Human-Centered Responsible AI (HRAI) Lab, recently teamed up with @jhucompsci.bsky.social sky.social to showcase a user-friendly tool aimed at transforming how users understand online privacy.

Visitors to the Smithsonian National Museum of Natural History in Washington, D.C., were recently invited to step into the shoes of synthetic

lucyinstitute.nd.edu

1 1 1

JHU Computer Science @jhucompsci.bsky.social · 7d

A @hopkinsengineer.bsky.social senior with law school aspirations, Iris Gupta is the founder and CEO of CounselAI, a digital discovery tool she developed with fellow computer scientist Atharva Barve, the company’s chief technology officer.

Wiser Counsel for Public Defenders - Johns Hopkins Engineering Magazine

An AI-powered digital discovery tool could help public defenders manage overwhelming caseloads.

Don’t Scan that Scam! - Johns Hopkins Whiting School of Engineering

Reposted by JHU Computer Science

Johns Hopkins Engineering @hopkinsengineer.bsky.social · Jul 17

QR code scams are on the rise. Security informatics grad Luis Rivas '24 (MS) combined 4 machine learning models to detect and block malicious QR codes—from fake charity phishing emails to bogus traffic fine texts—before users get fooled. engineering.jhu.edu/news/dont-sc...

QR code fraud is rising, so a Johns Hopkins team built a system to sniff out fakes

Johns Hopkins researchers have created a new kind of voice assistant to help older adults manage their health.

JHU Computer Science @jhucompsci.bsky.social · 9d

Johns Hopkins computer scientists (including @victor-antony.bsky.social) have developed a new personal voice assistant for older adults to help them take their health into their own hands.

Aging gracefully with AI

1 2

JHU Computer Science @jhucompsci.bsky.social · 12d

🥳
And @jhu.edu has been ranked #7 by @usnews.com’s Best Colleges rankings: hub.jhu.edu/2025/09/23/u...

#16 in Computer Science. 2025 U.S. News & World Report. Johns Hopkins computer scientists in a lab with a torso phantom.

Office Hours: Interview with Patricio Simari

JHU Computer Science @jhucompsci.bsky.social · 13d

Learn more about Senior Lecturer Patricio Simari, from his academic history to teaching here at Hopkins.

Second-year undergraduate student Johnathan Njau interviews the senior lecturer.

An industry-academia partnership with Amazon gives doctoral fellows funding to pursue big ideas in AI research.

JHU Computer Science @jhucompsci.bsky.social · 14d

Learn from @drewprinster.bsky.social and more how @jhu.edu’s AI2AI partnership with Amazon advances technologies in machine learning, computer vision, natural language understanding, and speech processing:

Research meets reality

It’s OK to say “IDK” - Johns Hopkins Whiting School of Engineering

Reposted by JHU Computer Science

Johns Hopkins Engineering @hopkinsengineer.bsky.social · Jul 11

Why won't the LLM just admit it doesn't know instead of making stuff up?! @williamjurayj.bsky.social is combining test-time compute with confidence thresholds for better, safer answers—teaching LLMs that it's risky to be wrong. engineering.jhu.edu/news/its-ok-...

Johns Hopkins researchers show how different “odds” can teach AI models to admit when they’re not confident enough in an answer.

For AI tumor detection, a picture isn’t always worth a thousand words

1 2

JHU Computer Science @jhucompsci.bsky.social · 16d

In work appearing at @miccaisociety.bsky.social this week, Johns Hopkins researchers have developed a new AI-driven tumor segmentation method that takes advantage of something radiologists already do every day: writing reports.

Johns Hopkins researchers have developed a new method that uses existing radiology reports to train AI models to locate tumors on CT scans more quickly and accurately.

Research Meets Reality - Johns Hopkins Engineering Magazine

JHU Computer Science @jhucompsci.bsky.social · 19d

It was a pleasure welcoming everyone at our new faculty reception yesterday! 🥰

Kai Presler-Marshall, Anton Dahbura, and Gagan Garg pose together.

Kim Franklin and Laura Henneman pose together, holding wine glasses. Mark Dredze crosses in the background.

A spread of samosas, bacon-wrapped dates, and sliders.

JHU Computer Science @jhucompsci.bsky.social · 20d

Learn about @drewprinster.bsky.social’s work as a doctoral fellow in the @jhu.edu + Amazon Initiative for Interactive #AI. Launched in 2022, it’s a novel industry-academia partnership designed to advance technologies in #ML, computer vision, natural language understanding, and speech processing.

An industry-academia partnership with Amazon gives doctoral fellows funding to pursue big ideas in AI research.

Analysis | Google says U.K. hasn’t demanded ‘back door’ to users’ private data

JHU Computer Science @jhucompsci.bsky.social · 21d

JHU’s @matthewdgreen.bsky.social called Google’s initial response to @wyden.senate.gov “pretty disturbing. Maybe they’re just being careful. But to me that sounds a little bit like a cry for help.” Read the full @washingtonpost.com article here: www.washingtonpost.com/politics/202...

Lawmaker’s letter suggests Apple might not be the only company the British government asked for special access to users’ private data.

www.washingtonpost.com

JHU Computer Science @jhucompsci.bsky.social · 22d

Helping robots become a “part” of our world, @jhu.edu computer scientists release new findings and a specialized dataset for training #robots to recognize specific object parts. 🤖 Learn more here: www.cs.jhu.edu/news/helping...

Multilingual artificial intelligence often reinforces bias

JHU Computer Science @jhucompsci.bsky.social · 29d

Congratulations to CS faculty @mdredze.bsky.social, Jason Eisner, Peter Kazanzides, and @tom-lippincott.bsky.social
on their @jhu.edu Nexus Awards! Learn more about their funded projects here: www.cs.jhu.edu/news/compute...

Headshots of Mark Dredze, Jason Eisner, Peter Kazanzides, and Tom Lippincott.

1 4

JHU Computer Science @jhucompsci.bsky.social · Sep 8

“If we want to shift the power to the people and enable them to make informed decisions, we need #AI systems capable of showing them the whole truth with different perspectives,” says @nikhilsksharma.bsky.social.

A new study from Johns Hopkins finds that multilingual AI privileges dominant languages, deepening divides rather than democratizing access to information.

JHU Computer Science @jhucompsci.bsky.social · Sep 8

Join us for the first CS seminar of the semester with our very own Zongwei Zhou! Learn more here: www.cs.jhu.edu/event/cs-sem...

Computer Science Seminar Series. Early Cancer Detection by Computed Tomography and Artificial Intelligence. September 18, 2025, 228 Malone Hall. Refreshments available 10:30 a.m. Seminar begins 10:45 a.m. Zongwei Zhou, Johns Hopkins University.

Cancer AI Alliance named to TIME100 AI 2025

JHU Computer Science @jhucompsci.bsky.social · Sep 5

CS’ @alexisbattle.bsky.social co-leads @jhu.edu’s participation in the alliance, which has been recognized for its groundbreaking work in #AI and #cancer research:

The partnership includes Johns Hopkins researchers developing AI-powered cancer care.