MIT Laboratory for Information and Decision Systems (LIDS)

Remembering Professor Emeritus Dimitri Bertsekas LIDS PI Dimitri Bertsekas, a pioneering researcher, influential educator, and prolific author whose work transformed fields ranging from optimization and control to reinforcement learning and artificial intelligence, has passed away at the age of 83. Throughout his remarkable career, Bertsekas advanced the foundations of large-scale computation, optimization, and AI while inspiring generations of students through his teaching and scholarship. He authored or co-authored more than 20 influential books, monographs, and textbooks that continue to shape research and education around the world. Beyond academia, he served as a consultant to industry, founded the publishing company Athena Scientific, and was chief scientific advisor to Bayforest Technologies. Yet his greatest legacy lives on through the countless students, collaborators, and researchers influenced by his work. Read a reflection on his extraordinary life, career, and lasting impact: https://bit.ly/3Rg1fip MIT EECS MIT Schwarzman College of Computing MIT School of Engineering

LIDS PI Bailey Flanigan has built an interdisciplinary career driven by a simple idea: follow the questions wherever they lead. A new profile traces her journey from medicine and bioengineering to public health, economics, computer science, and political science. Along the way, Flanigan has conducted research at the University of Wisconsin, the National Institutes of Health, Google, Carnegie Mellon, Drexel, Harvard, Princeton, and Stanford before joining MIT in 2025 as shared faculty member between the MIT Schwarzman College of Computing and the departments of MIT Political Science and Electrical Engineering and Computer Science (MIT EECS). Today, her research combines computational and mathematical tools to design new ways for people to participate meaningfully in democratic decision-making—helping ensure that democratic institutions can better reflect the voices they serve. Read the profile at MIT News: https://bit.ly/452IlyR

A faster way to sample from AI diffusion models MIT LIDS grad student Fan Chen, LIDS PI Sasha Rakhlin, with Sinho Chewi and CSAIL MIT's Constantinos Daskalakis receive #ICML2026 Outstanding Paper for their paper proposing a new algorithm that reaches the same level of accuracy in exponentially fewer steps, suggesting a path to much more efficient image generation & other diffusion-based AI systems. Congrats to the researchers! Read the paper: https://lnkd.in/g4P95izP

Helping #AI models meet the real world Through his research at MIT and his entrepreneurial work, LIDS PI Devavrat Shah is developing AI methods designed to support continuous decision-making while operating under real-world computational constraints. Those ideas led to the founding of Ikigai Labs, where Shah and his team built a foundation model for tabular and time-series data. Following Ikigai's acquisition by Celonis, Shah hopes the technology will help organizations integrate AI with their own data and business processes to improve forecasting, planning, and decision-making. Read a profile in MIT News to learn more about the research, the startup journey, and the future of AI for enterprise decision-making. More: https://bit.ly/3Rnx60p MIT EECS MIT Schwarzman College of Computing MIT School of Engineering

Keeping kids safer from illegal AI-generated content As generative AI becomes more powerful, it's critical to ensure these systems can't be easily adapted to produce harmful or illegal content. Researchers from MIT LIDS, Boston University, and child safety nonprofit Thorn have developed a new evaluation method that can identify potentially dangerous capabilities in generative AI models—without generating illegal outputs. This approach could help auditors assess open-source models for child safety risks while avoiding the creation of harmful material during testing. The team includes LIDS graduate student Vinith Suriyakumar, Ayush Sekhari (Boston University), LIDS postdoc Lena Stempfle, Robertson Wang (Thorn), Michael Simpson (Thorn), Rebecca Portnoff (Thorn), LIDS PIs Marzyeh Ghassemi and Ashia Wilson. Learn more at MIT News: https://bit.ly/4vAqCto

In game theory, generalists sometimes win out over specialists. New research from MIT LIDS and collaborators provides a more balanced framework for evaluating the algorithms that train AI agents to compete in imperfect-information games. The team, including LIDS graduate student Sobhan Mohammadpour and PI Gabriele Farina, found that an often-overlooked class of algorithms known as policy gradient methods can significantly outperform specialized game-theoretic approaches in certain types of games. The findings challenge conventional assumptions about which algorithms are best suited for strategic decision-making under uncertainty and could influence the design of future AI systems. The research was presented at the 2026 International Conference on Learning Representations (ICLR) in Rio de Janeiro. Learn more at MIT News: https://bit.ly/4faViv1 MIT EECS MIT Schwarzman College of Computing MIT School of Engineering

Congratulations to the 2026 L4DC Best Paper Award winners! 🎉 LIDS graduate students Andrea Goertzen and Sunbochen Tang, together with PI Navid Azizan, have received the Best Paper Award at the 2026 Learning for Dynamics and Control (L4DC) Conference for their paper, "ECO: Energy-Constrained Operator Learning for Chaotic Dynamics with Boundedness Guarantees." Goertzen and Tang are equal-contributing student co-authors of the award-winning paper, which introduces a new framework for modeling chaotic systems that draw on control theory and mathematical optimization to ensure that neural network models satisfy energy-dissipation constraints, enabling more reliable and physically meaningful long-horizon forecasts. The award was presented at the 8th Annual L4DC Conference, held at the University of Southern California on June 19, 2026. Learn more and read the paper: https://bit.ly/3RfP0Ch MIT Department of Mechanical Engineering (MechE) MIT AeroAstro MIT Chemical Engineering (ChemE) MIT School of Engineering MIT Schwarzman College of Computing Image: (L-R) Andrea Goertzen, Navid Azizan, and Sunbochen Tang

Could AI tell you where you left your keys? MIT researchers have developed a new long-term memory framework that enables robots to remember what they see—and answer questions about it using natural language. Created by LIDS researchers Nicolas Gorlo, Lukas Rosenberger Schmid, and PI Luca Carlone, the system—Describe Anything, Anywhere, Anytime, at Any Moment (DAAAM)—combines advanced map representations with rich semantic descriptions of an environment. This allows robots to efficiently store and retrieve information about objects they encounter while exploring. The framework answers questions more accurately than state-of-the-art methods while running efficiently enough for real-time use on a mobile robot. "If we want robots to work side-by-side with humans, they must speak the same language. The robot must be able to reason about time and space the same way humans do," says MIT AeroAstro Professor, LIDS PI, and MIT SPARK Lab Director Luca Carlone. "Our method turns a traditional map into a language-based map that is easier for the robot to reason about and access using language." Learn more and read the paper at MIT News: https://bit.ly/4vAv780 MIT AeroAstro MIT Schwarzman College of Computing MIT School of Engineering

Helping tiny robots navigate complex environments A new chip developed by researchers at MIT LIDS and MIT RLE could enable tiny, low-power robots to safely navigate tight, challenging spaces—such as industrial HVAC systems where they can inspect equipment and detect gas leaks. The team combined an efficient mapping algorithm with specialized hardware to rapidly generate 3D maps for navigation while using minimal memory and power. The result is a system that helps small robots avoid obstacles and maneuver around corners in environments where larger robots can't operate. The research team includes co-lead authors and MIT graduate students Zih-Sing Fu and Peter Zhi Xuan Li, LIDS Director Sertac Karaman, and senior author RLE's Vivienne Sze. “Real-time 3D mapping has been the missing piece for small autonomous systems. A drone inspecting a pipeline or a pair of AR glasses navigating a room both need to understand the space around them — instantly, continuously, and at almost no power cost. Gleanmer makes that possible for the first time in a chip you can hold between your fingers,” says Karaman. The work was recently presented at the IEEE Very Large-Scale Integrated Circuits (VLSI) Symposium. Learn more at MIT News: https://bit.ly/3QZVEfO MIT EECS MIT AeroAstro MIT Schwarzman College of Computing MIT School of Engineering Research Laboratory of Electronics at MIT

Creating distinguishable states for quantum systems A new study from MIT and the University of Ferrara could help advance the next generation of quantum technologies. In a paper just published in Physical Review A, LIDS PI Moe Z. Win, LIDS affiliate Peter Falb, and collaborators Andrea Giani and Andrea Conti introduce a new framework for improving the distinguishability of quantum states—a fundamental challenge in quantum information science. By making quantum states easier to distinguish, the framework can support the development of more effective quantum systems for sensing, communication, computing, and control. As quantum technologies continue to evolve, this work provides new theoretical tools that could help guide the design of future quantum devices. More at MIT News: https://bit.ly/4gtGYjo MIT AeroAstro MIT Schwarzman College of Computing MIT School of Engineering