Pinned Tweet
To a robot, a glass wall looks like an open path. UMD researchers are using light polarization to help autonomous systems spot invisible barriers like glass—a project that earned 2 of 5 workshop honors captured by UMD’s Perception and Robotics Group @prgumd at IROS, a flagship conference.
Congrats to computer science Ph.D. students Naitri Rajyaguru @naitri_r, Zhi “Leo” Wang @TX_Leo_Wang and Amir-Hossein Shahidzadeh, along with their collaborators, including group co-directors Yiannis Aloimonos and Cornelia Fermüller. Read more: go.umd.edu/2g93
Congrats to computer science Ph.D. students Naitri Rajyaguru @naitri_r, Zhi “Leo” Wang @TX_Leo_Wang and Amir-Hossein Shahidzadeh, along with their collaborators, including group co-directors Yiannis Aloimonos and Cornelia Fermüller. Read more: go.umd.edu/2g93
0
0
9
803
UMIACS retweeted
Some milestones are especially meaningful, both personally and professionally.
I'm incredibly proud to share that @priyathamkat has been promoted to CTO of RELAI! 🚀
My journey with Priyatham goes back well before RELAI. He was my PhD student at the University of Maryland, and after graduating, he joined RELAI in its earliest days, helping build the company from the ground up.
Since then, I've had the privilege of watching him grow from an outstanding researcher into an exceptional technical leader. He has been instrumental in shaping RELAI's technology and engineering team, combining deep technical expertise with remarkable dedication, ownership, and humility.
Congratulations, Priyatham! I'm excited for what we'll accomplish together in this next chapter! 🎉
I'm incredibly proud to share that @priyathamkat has been promoted to CTO of RELAI! 🚀
My journey with Priyatham goes back well before RELAI. He was my PhD student at the University of Maryland, and after graduating, he joined RELAI in its earliest days, helping build the company from the ground up.
Since then, I've had the privilege of watching him grow from an outstanding researcher into an exceptional technical leader. He has been instrumental in shaping RELAI's technology and engineering team, combining deep technical expertise with remarkable dedication, ownership, and humility.
Congratulations, Priyatham! I'm excited for what we'll accomplish together in this next chapter! 🎉
Excited to share that I've stepped into the role of CTO at @ReliableAI!
Grateful to my team and everyone who's helped along the way. Building reliable AI agents is what we're here for, and we're just getting started.
Special thanks to @FeiziSoheil and @py_laurent for their invaluable guidance and support.
1
2
22
3,235
UMIACS retweeted
So happy I could dance! 💃 Our paper "🪩 DiscoTrace: Representing and Comparing Answering Strategies of Humans and LLMs in Information-Seeking Question Answering" has been awarded Best Paper at #EMNLP2026!
aclanthology.org/2026.emnlp-mai…
aclanthology.org/2026.emnlp-mai…
10
20
171
11,049
A biofilm can create big problems for a data center — or it can help remove pollutants from water. Birthe Kjellerup, a @UofMaryland Professor of Civil and Environmental Engineering and co-director of the Center of Excellence in Microbiome Sciences, studies how to control the nd put their capabilities to work.
0
0
0
81
UMIACS retweeted
Stephen @stepscheng is presenting "What drives representation steering?" tomorrow (Sunday) #EMNLP2026
➡️ oral session 2 "Interp & Analysis of Models for NLP 1", room F8
➡️ Sunday 11am-12:30pm Budapest time 🇭🇺
📄 arxiv.org/abs/2604.08524
high-res poster: sarahwie.github.io/assets/poster_…
➡️ oral session 2 "Interp & Analysis of Models for NLP 1", room F8
➡️ Sunday 11am-12:30pm Budapest time 🇭🇺
📄 arxiv.org/abs/2604.08524
high-res poster: sarahwie.github.io/assets/poster_…
Ever wondered what pathways steering vectors work through to affect model outputs, or whether different steering methods are functionally similar? We also find that steering vectors can be sparsified *a lot*.
Check out our recent preprint, led by Stephen: arxiv.org/abs/2604.08524
1
8
51
2,400
UMIACS retweeted
Checkout this work that I'm really excited about, a massive effort by @liusida2007 & @feijianghan!
Turns out the ICA baseline in the original SAE paper had a bug. If you fix it, ICA becomes extremely competitive at finding interesting and interpretable LLM activation directions!
Turns out the ICA baseline in the original SAE paper had a bug. If you fix it, ICA becomes extremely competitive at finding interesting and interpretable LLM activation directions!
check out this intuitive video on why interpretable directions in llms look non-gaussian, how ica finds them in minutes per layer, and the evaluation bug that made ica look weak for 3 years
3
10
69
2,834
UMIACS retweeted
A University of Maryland researcher has joined a new research initiative that seeks to overcome one of artificial intelligence's biggest challenges: combining the adaptability of machine learning with the reliability of logical reasoning. The $7.5 million Department of Defense uld lead to more dependable autonomous systems and help AI tools make new scientific discoveries by drawing more reliable conclusions from complex data.
The five-year Multidisciplinary University Research Initiative (MURI) award includes up to $1.4 million for UMD to support research led by Soheil Feizi, an associate professor of computer science with an appointment in the University of Maryland Institute for Advanced Computer Studies (@umiacs).
The goal is to build a mathematical foundation for AI systems that can both learn from data and reason with structured knowledge, said Feizi, who is also a core faculty member in the UMD Center for Machine Learning.
today.umd.edu/umd-researcher…
The five-year Multidisciplinary University Research Initiative (MURI) award includes up to $1.4 million for UMD to support research led by Soheil Feizi, an associate professor of computer science with an appointment in the University of Maryland Institute for Advanced Computer Studies (@umiacs).
The goal is to build a mathematical foundation for AI systems that can both learn from data and reason with structured knowledge, said Feizi, who is also a core faculty member in the UMD Center for Machine Learning.
today.umd.edu/umd-researcher…
0
2
3
237
UMIACS retweeted
Real2Sim2Real, finally done right with neuro-symbolic representations!
Introducing Dex-One2Many: from one human video in one setting to a robot that generalizes to configurations beyond the demonstration—different object positions, goal poses, and grasps.
The key to the remarkable generalization is the abstraction. Dex-One2Many extracts the task’s relational structure into “scene graphs”, then uses them to generate diverse training states and rewards for RL in simulation. The graphs specify what must hold at each stage (constraints, our inductive bias); RL discovers how to achieve it across different configurations.
One demonstration opens up many ways to accomplish the same task—and the learned policy transfers zero-shot to a real dexterous hand.
Project: dex-one2many.github.io
Introducing Dex-One2Many: from one human video in one setting to a robot that generalizes to configurations beyond the demonstration—different object positions, goal poses, and grasps.
The key to the remarkable generalization is the abstraction. Dex-One2Many extracts the task’s relational structure into “scene graphs”, then uses them to generate diverse training states and rewards for RL in simulation. The graphs specify what must hold at each stage (constraints, our inductive bias); RL discovers how to achieve it across different configurations.
One demonstration opens up many ways to accomplish the same task—and the learned policy transfers zero-shot to a real dexterous hand.
Project: dex-one2many.github.io
Can a robot hand learn a task from ONE human video, and still do it when the objects are somewhere else?
Meet Dex-One2Many 🖐️
Trained only in simulation from one video, it runs on a real robot hand, even from setups the video never showed 🧵
0
2
47
4,199
UMIACS retweeted
Congrats to @NIST & @UofMaryland physicist Michael Gullans on winning the Arthur S. Flemming Award! The prestigious honor recognizes his breakthrough work in taming noisy qubits to make practical quantum computers a reality. Learn more: go.umd.edu/2frg
1
6
27
2,532
UMIACS retweeted
In this remarkable UMD-led project, our @umdscience and @umiacs researchers are using the power of AI to generate accurate, localized water forecasts faster to help decision makers prepare for emergencies. #AIMaryland go.umd.edu/2eji
0
1
7
400
UMIACS retweeted
Hard work pays off! 🏆 Our μ0 world model won Best Paper at the IROS 2026 RoboWoMo workshop! Next stop: CoRL.
So proud of our team: Seungjae (Jay) Lee, Yoonkyo Jung & Jia-Bin Huang!
Go UMD! Go physical intelligence! Go neuro-symbolic world models!
Our paper: “μ0: A Scalable 3D Inteel.”mu0-wm.github.io
For a broader look at world models and where they’re headed, here’s my blog: World Models: A Multiverse We Can Act On furong-huang.com/blog/world-mod…
So proud of our team: Seungjae (Jay) Lee, Yoonkyo Jung & Jia-Bin Huang!
Go UMD! Go physical intelligence! Go neuro-symbolic world models!
Our paper: “μ0: A Scalable 3D Inteel.”mu0-wm.github.io
For a broader look at world models and where they’re headed, here’s my blog: World Models: A Multiverse We Can Act On furong-huang.com/blog/world-mod…
Hot take: robots should not dream in pixels.
Pixels are too low-level.
Latents are too opaque.
μ₀ predicts a third thing:
3D motion traces.
On real robots, it beats π₀.₅ — with ~1/100 the data scale and no action labels for world-model pretraining. 🧵
w
(this video features voiceover narration)
6
8
133
9,599
UMIACS retweeted
.@umdcs' Mihai Pop uses computational tools to map the microbiome—the trillions of microbes in our bodies and environment.
Submit a question to Mihai and CS postdoc Marcus Fedarko, and they will answer it during tomorrow's #RedditAMA: redd.it/1wmp6cs
Submit a question to Mihai and CS postdoc Marcus Fedarko, and they will answer it during tomorrow's #RedditAMA: redd.it/1wmp6cs
0
3
8
420
UMIACS retweeted
As the field of AI continues to grow, so too does the need for safe, ethical and accountable approaches. This seed funding from @trails_ai supports projects that keep responsible development at the center of AI innovation. #AIMaryland go.umd.edu/2dng
0
1
9
361
UMIACS retweeted
Not all "magic" is enough to make a quantum computer outperform a classical one.
A study published in Physical Review Letters shows that some quantum states previously thought to provide the "magic" needed for quantum computation can actually be simulated efficiently by e findings redefine scientists’ understanding of the boundary between classical and quantum computing, and identify a more precise resource required for achieving a genuine quantum advantage.
Nicole Yunger Halpern (@nicoleyh11), a Fellow in the University of Maryland's Joint Center for Quantum Information and Computer Science (@JointQuICS), a theoretical physicist at the National Institute of Standards and Technology (NIST) and an adjunct faculty member in the University of Maryland Institute for Advanced Computer Studies (@umiacs ), collaborated with researchers at the University of Cambridge and other institutions on the study.
umiacs.umd.edu/news-events/ne…
A study published in Physical Review Letters shows that some quantum states previously thought to provide the "magic" needed for quantum computation can actually be simulated efficiently by e findings redefine scientists’ understanding of the boundary between classical and quantum computing, and identify a more precise resource required for achieving a genuine quantum advantage.
Nicole Yunger Halpern (@nicoleyh11), a Fellow in the University of Maryland's Joint Center for Quantum Information and Computer Science (@JointQuICS), a theoretical physicist at the National Institute of Standards and Technology (NIST) and an adjunct faculty member in the University of Maryland Institute for Advanced Computer Studies (@umiacs ), collaborated with researchers at the University of Cambridge and other institutions on the study.
umiacs.umd.edu/news-events/ne…
0
8
13
846
UMIACS retweeted
How can quantum particle statistics help—or hinder—Maxwell’s demon, and what are the latest experimental updates about this prediction? I enjoyed talking with science writer Lakshmi Chandrasekharan about quantum thermodynamics, creative writing and more. scieye.wordpress.com/2026/08/20/a-c…
1
5
13
585
UMIACS retweeted
.@NSF is investing more than $290 million in eight Quantum Leap Challenge Institutes to tackle the hardest scientific challenges in quantum computing, sensing, and communication.
Three new institutes join five renewed ones—spanning 19 states, 36 universities, national labs, and more than 30 U.S. companies.
nsf.gov/news/eight-nsf…
Three new institutes join five renewed ones—spanning 19 states, 36 universities, national labs, and more than 30 U.S. companies.
nsf.gov/news/eight-nsf…
5
21
63
3,989
UMIACS retweeted
The Institute for Robust Quantum Simulation is thrilled to announce that @NSF has renewed its funding, totaling $37.5 million. @Harvard is joining the RQS NSF consortium, expanding the current collaboration between @UofMaryland, @DukeU, @Princeton & @Yale. go.umd.edu/2dr4
0
2
7
2,849
A new study @PhysRevLett shows that some quantum states previously thought to provide the "magic" needed for quantum computation can actually be simulated efficiently by classical computers. The findings redefine scientists’ understanding of the boundary between classical and identify a more precise resource required for achieving a genuine quantum advantage. The study was coauthored by @JointQuICS Fellow @nicoleyh11, and @ParticulateDave and J.J. Thio from the University of Cambridge's @DeptofPhysics. Learn more: go.umd.edu/2dm8
0
5
2
1,380