μ₀ receives second-place Best Paper Award at the IROS 2026 RoBoWoMo Workshop
μ₀: A Scalable 3D Interaction-Trace World Model learns from video to predict 3D interaction traces that transfer across embodiments. Congratulations to all coauthors!
Lab notebook
News
Research releases, talks, student milestones, awards, and other signals from Furong Lab.
What to Attend, What to Keep wins Best Paper Award at the IROS 2026 Touch2Action Workshop
The paper introduces SkillFormer, which combines skill-conditioned vision and touch with event memory to track progress in contact-rich robotic manipulation. Congratulations to all coauthors!
From Atoms to Bits: Building the Digital Development Loop for Physical Intelligence
Furong presented at the ECCV 2026 tutorial “Evaluating Visual Foundation and World Models” in Malmö, Sweden. Drawing on IVE, DynaFLIP, μ₀, and Guava, the talk examined evaluation as infrastructure for a reliable digital development loop grounded in physical evidence.
Toward Self-Improving Agentic Systems Across Thinking, Actions, and Workflow
Furong spoke at Math & AI Summer in Seattle at Amazon, covering controlled decoding, agentic critical training, and query-adaptive workflows for self-improving agentic systems.
From CMSC 848N to a Y Combinator-backed startup
UMD CS alumnus Ananth Sankaralingam (B.S. ’26), co-founder of Chromie.dev, reflects on how Furong’s graduate AI course broadened his view of what could be built with large language models.
Reasoning as Control: Toward Self-Improving Agentic Systems
Featured talk at the Agentic AI Summit 2026 at UC Berkeley, with slides on control across thinking, actions, and workflow.
Generative AI Agents at DeepLearn 2026
An advanced summer-school course on agent architectures, reasoning, alignment, safety, and world models.
Building Self-Improving Foundation Models
Auditors, actuators, and amplifiers for trustworthy AI.
Reasoning as Control
Adaptive test-time compute for planning agents.
Seven papers accepted to ICML 2026
Work spanning reasoning, safety, multimodal systems, scientific prediction, and multi-agent learning.
PropensityBench adopted in Meta Muse
The agentic safety benchmark is used in the released Muse Spark MSL model.
TraceGen accepted to CVPR 2026
World modeling in 3D trace space for cross-embodiment robot learning.
Five papers accepted to ICLR 2026
Including MomaGraph, selected for an oral presentation.
AAAI 2026 oral
One paper accepted and selected for an oral presentation.
Four papers accepted to NeurIPS 2025
Including one spotlight presentation.
World-model learning for generalist agents
Keynote at the RLC workshop on Reinforcement Learning Beyond Rewards.
Computer Vision in the Wild
Keynote at the fourth CVPR Workshop on Computer Vision in the Wild.
Test-Time Thinking for Trust
Keynote at the second Texas Colloquium on Distributed Learning on enhancing generative AI agents.
Erasing the Invisible
Keynote at the ICLR Generative AI Watermarking Workshop on stress-testing watermark robustness.
World models for sequential decision-making
Keynote at the ICLR World Models Workshop on learning foundation models for action.
Towards Generative AI Security
Keynote at the New York Academy of Sciences Annual Machine Learning Symposium on stress-testing and alignment.
Promotion to associate professor with tenure
Promotion in the University of Maryland Department of Computer Science, effective July 1, 2024.
Integrity in AI at the U.S. SEC
A talk for the SEC AI Community of Practice on multimodal approaches to misinformation and content authenticity.
AskScience AMA on artificial intelligence
A public Reddit discussion about machine learning, trustworthy AI, and current research questions.
NSF–Amazon Fairness in AI PI Meeting
Chair and organizer of the Fairness in Artificial Intelligence principal-investigator meeting.
Trustworthy Machine Learning in an Ever-Changing World
University of Maryland Department of Computer Science colloquium.
Research releases, invited talks, recognition, career milestones, public engagement, and academic service.