Grants

NSF

The ability to debate on a topic; that is, the process of generating and evaluating claims, is a core competency for both academic success and civic life. Students traditionally learn these skills by participating in classroom discussions, during which they share, justify, and challenge each other's ideas. However, such discussions may be rare in classrooms and challenging for teachers. Without proper facilitation the discussion can get off track and even become heated. Unfortunately, teachers lack opportunities to improve these important facilitation skills. Coaching is an effective strategy for helping teachers learn facilitation, but it is hard to deliver at scale due to the lack of qualified coaches, time constraints, and teacher discomfort with being observed by a human coach. This project will address the challenges of teacher professional development by designing a simulated classroom environment in which elementary school teachers can practice facilitating discussions with AI-driven student avatars. The system will offer multiple low-cost and low-stakes opportunities for teachers to practice facilitation and receive expert-informed feedback from an automated coach. This project will provide a much-needed solution to one of the most pressing issues confronting education today: scaling up effective pedagogy that fosters students' skills. To this end, this project will integrate research in artificial intelligence, argumentation, and teacher learning to create a novel automated professional development system for teachers. A key technological innovation of this system is its hybrid AI architecture: a transparent, rule-based symbolic inference engine will structure simulated classroom dialogue and provide meaningful feedback to the teacher, while large language models will generate realistic responses from AI-driven student avatars. The system will formalize and integrate two established frameworks to support the analysis of teacher facilitation and student argumentation: the Argumentation Rating Tool and the Rational Force Model. Using design-based research, the researchers will develop effective design principles for AI-supported teacher learning. A randomized controlled trial will assess whether the system improves teachers' facilitation skills and confidence in facilitating discussions. In addition to contributing to scalable, high-quality professional development, the project will inform the design of ethical and transparent AI systems for education, with potential applications in intelligent tutoring and instructional coaching. This project is funded by the Research on Innovative Technologies for Enhanced Learning (RITEL) program that supports early-stage exploratory research in emerging technologies for teaching and learning. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

When a strong shockwave passes through the interface between a solid and a gas, the solid can melt and form liquid sheets that are released into the gas. Typically, the solid surface will have roughness features that can cause ripples on the interface, leading to the formation of the liquid sheets. This phenomenon is called ejecta and occurs in several engineering applications and natural phenomena such as semiconductor manufacturing, nuclear fusion and weapons, ballistic projectile impact on an armored vehicle, asteroid impacts and supernovae explosions. Once the liquid sheets form, they can undergo further destabilization and break up into ligaments and eventually droplets. The process leading to breakup of the sheets is poorly understood and will be the focus of this study. The research will address this problem through detailed numerical simulations and detonation tube experiments. The findings from the simulations and experiments will be used to develop a ‘lifecycle’ model that can describe all stages of shock-driven liquid sheet evolution from initial formation, to growth, to fragmentation into ligaments and droplets. The project will support training of students in areas vital to national security. The overall goal of this project is to improve our understanding of the hydrodynamic instability mechanisms that govern the breakup of ejecta sheets in a gas, and the effect of such mechanisms on the size and velocity distributions of ejecta particles. Thus, the proposed work will help us understand the complex physics governing the breakup of ejecta into droplets. This is accomplished by investigating the problem at lower strain rates, so that every stage of the flow can be resolved by our detonation tube experiments and numerical simulations. Insights from the experiments and simulations will feed into the development of a ‘lifecycle’ model that will tie breakup dynamics to early-stage, linear instability physics. Having such an end-to-end model will enable control of liquid sheets in several industrial applications, by controlling the surface corrugations on the solid, so that a desired droplet distribution may be achieved. The breakup model developed will find broad usage as a sub grid model in simulations of impulsively driven sheets and jets. The project will help support the defense and security industries and will provide students the training to fill a vital workforce need in areas of national security. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

When a strong shockwave passes through the interface between a solid and a gas, the solid can melt and form liquid sheets that are released into the gas. Typically, the solid surface will have roughness features that can cause ripples on the interface, leading to the formation of the liquid sheets. This phenomenon is called ejecta and occurs in several engineering applications and natural phenomena such as semiconductor manufacturing, nuclear fusion and weapons, ballistic projectile impact on an armored vehicle, asteroid impacts and supernovae explosions. Once the liquid sheets form, they can undergo further destabilization and break up into ligaments and eventually droplets. The process leading to breakup of the sheets is poorly understood and will be the focus of this study. The research will address this problem through detailed numerical simulations and detonation tube experiments. The findings from the simulations and experiments will be used to develop a ‘lifecycle’ model that can describe all stages of shock-driven liquid sheet evolution from initial formation, to growth, to fragmentation into ligaments and droplets. The project will support training of students in areas vital to national security. The overall goal of this project is to improve our understanding of the hydrodynamic instability mechanisms that govern the breakup of ejecta sheets in a gas, and the effect of such mechanisms on the size and velocity distributions of ejecta particles. Thus, the proposed work will help us understand the complex physics governing the breakup of ejecta into droplets. This is accomplished by investigating the problem at lower strain rates, so that every stage of the flow can be resolved by our detonation tube experiments and numerical simulations. Insights from the experiments and simulations will feed into the development of a ‘lifecycle’ model that will tie breakup dynamics to early-stage, linear instability physics. Having such an end-to-end model will enable control of liquid sheets in several industrial applications, by controlling the surface corrugations on the solid, so that a desired droplet distribution may be achieved. The breakup model developed will find broad usage as a sub grid model in simulations of impulsively driven sheets and jets. The project will help support the defense and security industries and will provide students the training to fill a vital workforce need in areas of national security. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Many aquatic organisms use a technique called metachronal rowing to swim. They use their paddle-like appendages to row in a coordinated rhythm, starting from the rear and moving toward the front of the organism. Metachronal rowing is observed in organisms that range in size from single cells to large crustaceans such as shrimp and krill. This project will use experiments and computational modeling on live animals and metachronal rowing vehicles to explain why this swimming technique works regardless of organism size. By examining how animals of different sizes optimize their swimming appendages, this research will help design underwater vehicles that can efficiently operate over broad ranges of sizes and speeds. Undergraduate and graduate students will receive cross-disciplinary training in fluid mechanics, robotics, and scientific computing. New summer camps on bio-inspired engineering will be developed for high school students. This project will elucidate the fluid dynamic principles that enable thrust and lift generation by metachronal rowing across an astounding seven orders of magnitude in paddle-scale Reynolds number from strongly viscous (0.01) to inertially dominated (10,000) flow regimes. Previous studies of metachronal rowing have considered tethered paddling without body motion for a limited Reynolds number range. Flow visualization and force measurements will be performed on state-of-the-art dynamically similar remotely operated vehicles to examine the flow physics, swimming performance and scalability of metachronal rowing across the biologically relevant range of Reynolds numbers. Computational simulations will be conducted to examine fluid-structure interaction in untethered metachronal swimming and evaluate the cost of transport for varying Reynolds number, paddle geometry and kinematics. The integrated experimental and computational approach will be used to test whether differences in the mechanical design and paddling kinematics of natural metachronal swimmers can facilitate efficient locomotion in their specific flow regime. The findings can guide the development of new bio-inspired autonomous underwater vehicles that provide efficient propulsive performance across broad ranges of sizes and speeds. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

The formation and sinking of organic particles into the deep ocean are important processes in global element cycles. These particles (also called ‘particulate organic matter’ or POM) carry carbon into the deep ocean where it is stored for hundreds of years. The storage of carbon in the deep ocean contributes to the drawdown of atmospheric carbon dioxide as well as loss of nutrients from the surface ocean. Understanding the mechanisms and processes that govern the fluxes of POM is thus directly relevant to the global carbon cycle, marine ecology, and fisheries. This project uses new tools to measure the isotopes of three different elements (hydrogen, carbon, nitrogen). The approach shows promise for addressing a long-standing question: to what extent do bacteria replace the organic matter from algae (the original source of the POM) as particles sink. Results from this study will be important for predicting the amount and dietary quality of POM in a rapidly changing ocean. The project will support Ph.D. students at both Caltech and University of Miami and educational activities for K-12 and college students. The project takes advantage of a recent analytical development, the measurement of hydrogen isotopes in amino acids as a new tracer. Preliminary measurements suggest that there is a large (up to 20%) shift in the hydrogen isotope ratio of amino acids within POM over the upper 300 m of the ocean, consistent with the depth at which POM degradation is most intense. The investigators hypothesize that the shift in hydrogen isotopes reflects the replacement of phytoplankton biomass with bacterial biomass. The project will test this hypothesis through 1) the analysis of archived POM samples from four different localities in the Atlantic and Pacific oceans, 2) measurements of diverse phytoplankton grown in culture, and 3) degradation experiments in which algal biomass is fed to bacteria and zooplankton. In addition to cutting-edge hydrogen isotope measurements, the investigators will employ a suite of additional measurements to characterize the microbial communities and POM being studied. Collectively, this work aims to develop hydrogen isotopes in amino acids as a novel proxy for the turnover of organic matter within marine particles, which could be applied to studies of marine POM throughout the world’s oceans. The project will also develop outreach activities to introduce K-12 and collegiate students to marine science. These will be implemented through the GO-Outdoors program in Pasadena, the Exploring Marine Science Day in Miami, and undergraduate student cruises in Miami. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

The Earth’s mantle melts as it rises beneath mid-ocean ridges and hotspot volcanoes. This process generates ocean crust and is responsible for most of the volcanic eruptions on Earth. The presence of even small amounts of water in the mantle is known to have large effects on the amount of magma produced. Likewise, small changes in chemical composition of the mantle rocks will also change the amount of magma produced. This project will be the first to study both of these factors together. The project involves both laboratory experiments that replicate the conditions in the mantle and computational models that predict melting. Two graduate students will conduct each aspect of the project, respectively, but also will be cross-trained under the guidance of principle investigators. Several first-year undergraduates will also be involved in the research, leveraging an existing program at the University of Utah. The anticipated outcome of this research will be the first models that accurately predict the combined effect of water and composition variations on mantle melting. The project will also support making video tutorials to help with the adoption of the software developed in this project. The presence of even small amounts of water and other volatiles can significantly alter the way that mantle rocks melt and how much magma is produced to form new crust and create volcanoes. Understanding the mantle melting processes from the composition of basalts produced is a very difficult problem, so laboratory experiments and computational modeling provide the best ways to address these questions. This project integrates high pressure-high temperature laboratory experiments and geochemical numerical models to investigate the impacts of variable water content on melting of a realistic mantle composition containing both peridotites and pyroxenites as they coexist in a decompressing, rising mantle plume. The project will develop the first model parameterization that can accurately predict the effects of water on melting a lithologically realistic mantle containing pyroxenites, and make publicly available open-source software to enable other researchers to run these models. The central hypothesis is that increasing both the abundance of pyroxenite and the concentration of water in the decompressing mantle will generate measurably deeper melting than either factor alone. Experimental data generated during the project will add significantly to the public Library of Experimental Phase Relations and the EarthChem national data repository. This project will enable two graduate students to continue their studies and will provide first-hand research experiences for freshmen students at University of Utah. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

This project aims to break the low latency performance barrier in today’s fifth generation (5G) networks that hinders progress and adoption of remote driving industry (the “vertical” application). It advances an innovative “vertical-aware” framework to optimize both 5G networks and the vertical application. Despite tremendous progress, today’s “self-driving” cars may encounter many situations where they cannot drive themselves safely. Examples include construction zones and traffic accidents on the road. By ensuring low latency needed for remote driving, the developed solutions will allow a human teleoperator to remotely steer a “connected and autonomous” vehicle (CAV) through complex situations as if sitting in the driver seat. Technological advances enabled by this project will help (re-)establish the U.S.’s leadership in next-generation (NextG) wireless telecommunications and major vertical industries such as automotive and robotic automation. This project also provides a unique educational platform to train students and expand the STEM (Science, Technology, Engineering & Mathematics) workforce. Two major hurdles in ensuring low latency over 5G networks are i) high mobility of vehicles leads to poor radio channel conditions, causing data delivery errors; ii) frequent handovers among radio base stations further prolong data delivery. The project will develop a novel Open Radio Access Network (O-RAN) enabled, vertical-driven framework with mobility-aware, proactive mechanisms to reduce impacts of high mobility and handovers on the tail latency performance of the target vertical application. This is achieved by enabling 5G networks to utilize information (e.g., vehicle trajectory and speed) provided by remote driving applications to make intelligent decisions to speed up the delivery of sensor and command-and-control data that are critical to remote driving, whereas CAVs can also take advantage of vertical-aware predictions made by 5G networks to decide when and how to transmit data. Additional innovations include incorporation of integrated 5G and cellular vehicle-to-everything (C-V2X) technologies for cooperative situation awareness to further ensure safe remote driving operations. The phased approach to developing the proposed solutions and demonstrating their capabilities will ensure a high chance of successful execution, truly moving the needle with transformative impacts on relevant industrial sectors. The project represents close collaboration across three academic institutions and two industry leaders in key relevant sectors providing an accelerated pathway to technology transition. By demonstrating the value of vertical-aware advanced 5G/NextG networks in support of remote and cooperative driving and other industrial use cases, this project will help create new opportunities and business models for both mobile network operators and network equipment vendors for sustained investments in network innovations. It will also help accelerate adoption of autonomous driving with teleoperation capabilities. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

This project aims to break the low latency performance barrier in today’s fifth generation (5G) networks that hinders progress and adoption of remote driving industry (the “vertical” application). It advances an innovative “vertical-aware” framework to optimize both 5G networks and the vertical application. Despite tremendous progress, today’s “self-driving” cars may encounter many situations where they cannot drive themselves safely. Examples include construction zones and traffic accidents on the road. By ensuring low latency needed for remote driving, the developed solutions will allow a human teleoperator to remotely steer a “connected and autonomous” vehicle (CAV) through complex situations as if sitting in the driver seat. Technological advances enabled by this project will help (re-)establish the U.S.’s leadership in next-generation (NextG) wireless telecommunications and major vertical industries such as automotive and robotic automation. This project also provides a unique educational platform to train students and expand the STEM (Science, Technology, Engineering & Mathematics) workforce. Two major hurdles in ensuring low latency over 5G networks are i) high mobility of vehicles leads to poor radio channel conditions, causing data delivery errors; ii) frequent handovers among radio base stations further prolong data delivery. The project will develop a novel Open Radio Access Network (O-RAN) enabled, vertical-driven framework with mobility-aware, proactive mechanisms to reduce impacts of high mobility and handovers on the tail latency performance of the target vertical application. This is achieved by enabling 5G networks to utilize information (e.g., vehicle trajectory and speed) provided by remote driving applications to make intelligent decisions to speed up the delivery of sensor and command-and-control data that are critical to remote driving, whereas CAVs can also take advantage of vertical-aware predictions made by 5G networks to decide when and how to transmit data. Additional innovations include incorporation of integrated 5G and cellular vehicle-to-everything (C-V2X) technologies for cooperative situation awareness to further ensure safe remote driving operations. The phased approach to developing the proposed solutions and demonstrating their capabilities will ensure a high chance of successful execution, truly moving the needle with transformative impacts on relevant industrial sectors. The project represents close collaboration across three academic institutions and two industry leaders in key relevant sectors providing an accelerated pathway to technology transition. By demonstrating the value of vertical-aware advanced 5G/NextG networks in support of remote and cooperative driving and other industrial use cases, this project will help create new opportunities and business models for both mobile network operators and network equipment vendors for sustained investments in network innovations. It will also help accelerate adoption of autonomous driving with teleoperation capabilities. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Extended reality (XR) technologies have shown significant promise in increasing user engagement and skill acquisition in a variety of domains. The goal of this project is to accelerate adoption of innovative XR applications for rehabilitation, for example, to enhance user experiences in treatments to improve motor function for diseases with motor disabilities. This project develops immersive XR exercise environments that enable users to move and interact in 3D space with each other and with virtual elements. A key goal is to make the experience of rehabilitation more enjoyable and effective by incorporating social interactions between remote users that can boost engagement and skill acquisition. The technology also allows clinicians to guide and interact with their patients remotely. To create a virtual environment that users experience as fast and seamless, this project develops novel approaches to the underlying networking infrastructure needed to run the application. Collaboration with industry partners will support technology adoption for XR-enabled rehabilitation technology and for the XR industry. Realizing multi-user, geo-distributed XR technology is challenging due to stringent motion-to-photon latency requirements for good user experience. Current wide-area Internet routing and cellular wireless management are one-size-fits-all across applications, hurting latency. A key insight of this project is that not all types of XR traffic require uniformly low latency; instead, the project enables prioritization of delivery of important XR traffic while intelligently managing lower-priority XR traffic. The project’s core technical contributions are an adaptive XR application with new delivery mechanisms, leveraging Internet path selection on the testbed, and programmable wireless resources using Open Radio Access Networks (Open RAN) and xApps. Performance of the XR environment will be assessed via user experience and key network performance indicators. The expected outcome is achieving key performance targets currently deemed impossible on today’s Internet by demonstrations of network-supported, multi-user XR technology for rehabilitation at different sites. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

This project will advance next generation wireless technology to substantially reduce communication delay, known as latency, such that low latency communication service can become economically and broadly available. The project develops this technology for remote medical collaboration applications, to make medical specialists and clinicians available when they are urgently needed. Extended reality (XR) technology, including virtual reality and augmented reality, could allow remote specialists to assist on-site personnel, but enabling such real-time shared experiences critically depends on the availability of low latency wireless networking which is not generally deployed. This project will create an innovative system and platform that delivers low-latency wireless XR for remote medical collaboration, or LXM. LXM will solve technology barriers that ease deployment of low latency service for a wide range of applications even beyond medicine. The project is thus advancing U.S. global leadership in next-generation wireless technology, enabling timely patient medical response, and developing the U.S. workforce in wireless networking and XR-based medical collaboration. The project will develop both XR-based medical applications and the underlying networking technology stack to enable practically deployable low latency wireless service. It will also develop surgical planning and medical training applications that support remote collaboration and cross-layer performance requirement signaling. These will leverage new capabilities in underlying systems to provide (1) fine-grained heterogeneous network service, to enable using low latency service judiciously which results in an improved tradeoff between network latency and network bandwidth; (2) dynamic adaptivity to deal with variable network conditions; and (3) cross-layer performance intent signaling, so the technology stack can make good context-sensitive, application-aware decisions. Achieving these properties will involve new innovations in the XR runtime, transport protocol, 5G radio access network (RAN) and user equipment (UE), and will leverage advanced low-latency WiFi. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Melting of ice shelves by warm ocean waters destabilizes glaciers, enhancing ice flow into the ocean and contributing to global sea level rise. Measurements of ocean properties under ice shelves are needed to improve global predictions of sea level rise and to anticipate its societal consequences. Such measurements, however, are challenging to obtain. In this Ideas Lab: Engineering Technologies to Advance Underwater Sciences (ETAUS) project, a proof-of-concept mothership-and-passenger system will be developed to permit the future deployment of a highly capable, autonomous underwater vehicle (the mothership), programmed to travel as far as safely possible under ice shelves to release a swarm of novel, low-cost passenger robots that will coordinate to explore further into the ice cavity. The hardware prototypes, networked communication systems and protocols, and coordination algorithms developed as part of this project’s mothership-and-passenger system will help advance the field of underwater exploration in confined and hard-to-reach environments. The project will also foster the training of future scientists and engineers by engaging youths from small fishing communities in Oregon through presentations at Oregon’s MATE ROV Regional Competition, by employing high-school interns through the Apprenticeships in Science and Engineering (ASE) Summer Academy Program, and by training multiple undergraduate and graduate students at participating institutions. The goal of this project is to develop a mothership-and-passenger sampling system to reach difficult-to-access glacier grounding zones via the open ocean to measure the extent of ice cavities and surrounding water properties. The project will innovate along three main areas of inquiry: 1) passenger robot design, 2) acoustic communication protocols and hardware, and 3) mothership-and-passenger coordination algorithms. Novel, low-cost passenger robots will be conceived that can switch through different operation modes to optimize maneuverability, power consumption, or a combination of both, as needed for various tasks throughout a deployment. Acoustic communication protocols and hardware will be developed to prioritize robust communication between passenger robots over throughput and permit swarm self-localization by utilizing time-of-flight and angle-of-arrival between passengers to estimate relative positions. Swarm coordination algorithms will be designed to estimate flow direction and strength from the passenger robots’ relative positions to optimize navigation and power consumption. The performance of the network will be tested in increasingly challenging environments, i.e., tests will be conducted in a pool, an unfrozen lake, and finally in a frozen lake, while network capabilities with a larger swarm will be modeled to ensure the scalability of the system to ocean deployments. Finally, the software developed for the mothership-and-passenger communication and self-localization protocols will be generalizable and made available open-source to allow other research teams to adapt the system to their own needs. The mothership-and-passenger sampling system will not only advance under-ice-observation capabilities but also have wider oceanographic applications such as detection and monitoring of underwater harmful algal blooms or anoxic events threatening fisheries. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

As our planet continues to warm, we are facing a future climate unlike anything humanity has ever experienced. To better understand and prepare for what might happen, this project examines periods in Earth’s distant past when global temperatures were much higher than today. While previous studies have focused on single warming events, this research will analyze more than ten consecutive events, providing more accurate and dependable insights into how the Earth responds to extreme warming. Using new, cutting-edge methods, the team will measure changes in rainfall intensity, flooding, and other climate shifts in ways that were not possible before. The knowledge gained will improve our ability to anticipate and manage risks such as flooding and extreme rainfall, strengthening our nation’s resilience and protecting public welfare. Beyond advancing the frontiers of climate science, the project will help to build a strong, competitive STEM workforce by integrating research findings into K–12 outreach and university teaching materials, thereby directly contributing to national priorities. This project aims to advance our fundamental understanding of how rainfall patterns and river systems respond to a changing climate by collecting new and unique data on extremely warm climates of the past. The team expects to generate novel insights, as this will be the first terrestrial dataset to (1) quantify both climate changes and Earth system responses across more than ten consecutive global warming events, and (2) apply innovative reconstruction and analytical methods to measure rainfall intensity and intermittency—key information that is difficult to obtain using traditional approaches. Combining these new methods with the traditional state-of-the-art stable isotopic and geochemical methods will allow the team to collect new and original data on precipitation changes and extremes, and determine whether these changes occurred gradually or involved sudden "tipping points" that lead to dramatic shifts. To reduce uncertainty typically associated with sedimentary record-based climate reconstructions, the study will focus on a single paleo-river system. Together, these efforts promise to deliver transformative insights into Earth’s past, providing critical context for understanding our planet’s future. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

As our planet continues to warm, we are facing a future climate unlike anything humanity has ever experienced. To better understand and prepare for what might happen, this project examines periods in Earth’s distant past when global temperatures were much higher than today. While previous studies have focused on single warming events, this research will analyze more than ten consecutive events, providing more accurate and dependable insights into how the Earth responds to extreme warming. Using new, cutting-edge methods, the team will measure changes in rainfall intensity, flooding, and other climate shifts in ways that were not possible before. The knowledge gained will improve our ability to anticipate and manage risks such as flooding and extreme rainfall, strengthening our nation’s resilience and protecting public welfare. Beyond advancing the frontiers of climate science, the project will help to build a strong, competitive STEM workforce by integrating research findings into K–12 outreach and university teaching materials, thereby directly contributing to national priorities. This project aims to advance our fundamental understanding of how rainfall patterns and river systems respond to a changing climate by collecting new and unique data on extremely warm climates of the past. The team expects to generate novel insights, as this will be the first terrestrial dataset to (1) quantify both climate changes and Earth system responses across more than ten consecutive global warming events, and (2) apply innovative reconstruction and analytical methods to measure rainfall intensity and intermittency—key information that is difficult to obtain using traditional approaches. Combining these new methods with the traditional state-of-the-art stable isotopic and geochemical methods will allow the team to collect new and original data on precipitation changes and extremes, and determine whether these changes occurred gradually or involved sudden "tipping points" that lead to dramatic shifts. To reduce uncertainty typically associated with sedimentary record-based climate reconstructions, the study will focus on a single paleo-river system. Together, these efforts promise to deliver transformative insights into Earth’s past, providing critical context for understanding our planet’s future. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Discovering planets orbiting other stars can teach us about how they form, and capabilities now exist to find gas giants orbiting at 5-30AU from their stars (an AU is the distance between the Earth and the Sun), like Jupiter and Saturn do in the Solar System. Unlike commonly employed indirect methods to sense a planet, the technique of infrared direct imaging requires that the planet still glows from its heat of formation. Hence stars that have newly formed, which may be accompanied by newly-formed planets, need to be identified and targeted. Imaged planets can be followed up with spectroscopy to characterize their atmospheres and determine their fundamental properties, and about a dozen have been found to date. Sometimes the youth of nearby stars can be inferred from how they move in groups with other stars, however a very limited number of these stars are currently known. In this work, isolated stars that are less than 1 billion years old will be identified, and their ages determined. Some of those stars will be targeted to search for planets, and the full catalog of stars will be published for follow-on imaging by others. Mentoring programs for undergraduate students in summer research will be done with combination of a traditional in-person mentoring at NMSU and an innovative online experience for a larger number of students at ASU. To produce an all-sky catalog of nearby (<50 parsec), young (<1 gigayear) stars, a combination of multiple spectroscopic and photometric age indicators across a range of spectral types will be interpreted with a robust statistical framework. The team will employ X-ray flux, Gaia-measured space velocity, and TESS-derived rotational periods, then confirm their youth with high resolution optical spectra for lithium abundance, H-alpha emission, and calcium emission. They will also screen the targets for binarity with snapshot Adaptive-Optics observations and radial-velocity monitoring. As the youngest nearby stars are identified, the team will begin a direct imaging planet search with available high contrast AO imagers, including Gemini-North/GPI, LBT/LMIRcam+SHARK, and Subaru/SCEx-AO. Undergraduate researchers will participate in observing runs (in-person or virtually), gain experience in data analysis, and present their results at conferences or in virtual reports/posters at team meetings. Graduate students at both institutions will participate in mentoring. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Large language models (LLMs), such as GPT-4, also known as foundation models, represent a groundbreaking advancement in artificial intelligence, powering diverse applications such as chatbots, information retrieval engines, and scientific discovery tools. The success of LLMs is primarily attributed to their vast internal knowledge learned from massive unstructured texts. However, LLMs' internal knowledge is inherently constrained by static snapshots of closed-world texts used for training. This limitation presents two key challenges when deploying such a closed-world LLM in an open-world environment where knowledge (including unstructured texts and structured data) is rapidly evolving. First, closed-world LLMs have a limited and static view of open-world knowledge, often leading to unfaithful yet overconfident hallucinations. Second, since they primarily process unstructured text, they struggle with tasks that require reasoning over structured data, such as databases or scientific records. This project addresses key challenges in artificial intelligence by developing new algorithms, theorems, and systems to ensure the reliability of advanced foundation models. At its core, the research focuses on an open-world foundation model (OWFM), a powerful AI model designed to interact with ever-changing real-world information. This model is built on an open-world knowledge network (OKN), a flexible and expandable data structure that organizes semi-structured information from diverse sources. Moving beyond unstructured knowledge and closed-world LLMs, this project establishes a new paradigm of knowledge organization and foundation models in an open-world environment through three key thrusts. The first thrust creates a highly expandable OKN across domains and builds an OWFM with plug-and-play modular components. The second thrust develops adaptation methods that enable OWFMs to accurately answer questions on rapidly evolving topics. The third thrust deploys the OWFM in two knowledge-intensive, high-stakes applications: medical diagnostic reasoning and financial portfolio generation. Beyond advancing scientific research, this project also integrates its findings into educational activities, ensuring broad dissemination and real-world impact. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Large language models (LLMs), such as GPT-4, also known as foundation models, represent a groundbreaking advancement in artificial intelligence, powering diverse applications such as chatbots, information retrieval engines, and scientific discovery tools. The success of LLMs is primarily attributed to their vast internal knowledge learned from massive unstructured texts. However, LLMs' internal knowledge is inherently constrained by static snapshots of closed-world texts used for training. This limitation presents two key challenges when deploying such a closed-world LLM in an open-world environment where knowledge (including unstructured texts and structured data) is rapidly evolving. First, closed-world LLMs have a limited and static view of open-world knowledge, often leading to unfaithful yet overconfident hallucinations. Second, since they primarily process unstructured text, they struggle with tasks that require reasoning over structured data, such as databases or scientific records. This project addresses key challenges in artificial intelligence by developing new algorithms, theorems, and systems to ensure the reliability of advanced foundation models. At its core, the research focuses on an open-world foundation model (OWFM), a powerful AI model designed to interact with ever-changing real-world information. This model is built on an open-world knowledge network (OKN), a flexible and expandable data structure that organizes semi-structured information from diverse sources. Moving beyond unstructured knowledge and closed-world LLMs, this project establishes a new paradigm of knowledge organization and foundation models in an open-world environment through three key thrusts. The first thrust creates a highly expandable OKN across domains and builds an OWFM with plug-and-play modular components. The second thrust develops adaptation methods that enable OWFMs to accurately answer questions on rapidly evolving topics. The third thrust deploys the OWFM in two knowledge-intensive, high-stakes applications: medical diagnostic reasoning and financial portfolio generation. Beyond advancing scientific research, this project also integrates its findings into educational activities, ensuring broad dissemination and real-world impact. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Machine learning has grown to increase prominence over the past years, finding applications in various domains from image and speech processing to disease diagnosis. Despite the great success of machine learning techniques, massive amounts of data are collected and used to train the machine learning models. The privacy of sensitive data has become a big concern. Existing efforts are still preliminary, and enormous challenges remain to be resolved. Crucially, stronger privacy protection guarantees often sacrifice important properties of machine learning models, such as predictive utility and fairness, which can be undesirable or completely unacceptable. This project develops a consolidated privacy protection framework for machine learning systems that comprehensively considers the optimal trade-offs between computational privacy and several critical properties of machine learning, including utility, fairness, and distributed learning. The project will provide a comprehensive set of tools to protect data privacy for real-world machine learning applications under different circumstances. The privacy-preserving techniques will have a transformative impact on machine learning systems used by various sectors, allowing companies and hospitals to enjoy the advantages of machine learning techniques on big data while protecting data privacy under corresponding regulations. The research project thoroughly examines and discusses the real-world complicacy or restrictions when applying differential privacy, from privacy-utility trade-off, privacy-fairness relation, privacy in distributed learning, to post-learning privacy protection. The framework developed by the project takes deep root in rigorous optimization frameworks, often accompanied by theoretical guarantees and aided by cutting-edge algorithmic tools such as meta-learning, adversarial learning, and federated learning. Besides, the framework carries the following methodological innovations: differential privacy tailored to learning problems; customized privacy addressing heterogeneity in collaborative learning; privacy-protection of learned models through unlearning; consolidated privacy and fairness in learning. Those efforts will significantly augment the practicality and scalability of differential privacy. The project will be systematically evaluated on various real-world medical applications, and the tools will be readily used to tackle critical challenges in medical research. The outcomes will be incorporated into multiple courses at both undergraduate and graduate levels. The research outcomes will be disseminated broadly and comprehensively through open-source software releases and workshops, the involvement of undergraduate research, and outreach to K-12 education, focusing on minorities and under-representative groups in STEM education. Students at different levels and disciplines, STEM and liberal arts, will be participating in the research on privacy and machine learning. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Large language models (LLMs) are increasingly integrated into daily life, powering applications in education, healthcare, code generation, and more. However, these models can inadvertently memorize and reproduce sensitive or harmful content, including private user data, copyrighted material, and unsafe instructions. Retraining LLMs from scratch to remove such content is often impractical due to high cost and complexity. This project charts a new course toward more controllable, debuggable, and secure artificial intelligence (AI) through LLM unlearning, a paradigm that enables the targeted removal of harmful data influences and behaviors from pretrained models without compromising their overall performance. The research advances national priorities by promoting trustworthy AI, strengthening data privacy, ensuring safe deployment across sectors such as cybersecurity, healthcare, and education, and enabling contextually-adaptive systems aligned with a wide range of social norms. The project also offers strong educational and outreach opportunities, including curriculum development, research dissemination through workshops, tutorials, publications, and open-source software, as well as the creation of inclusive mentoring programs. This project aims to establish a comprehensive foundation for LLM unlearning by addressing challenges across four interconnected areas: optimization, model, data, and application. On the optimization front, it develops new algorithmic frameworks to enhance the effectiveness, robustness, and efficiency of LLM unlearning. At the model level, it investigates how internal components of LLMs contribute to memorization, introducing interpretability-driven approaches to identify and adjust influential weights without compromising essential capabilities such as LLMs' "emergent" abilities. On the data side, the project examines the role of watermarking and coreset selection in shaping unlearning outcomes, advancing methods for handling imperfect or proxy forget sets. These innovations are applied to privacy-sensitive scenarios such as conversational risk assessment in online dating, enabling LLMs to evaluate behavioral risks while erasing personal information. Conducted by a multidisciplinary team with a strong track record in trustworthy machine learning, the project is expected to deliver principled algorithms, practical tools, and rigorous benchmarks that advance responsible, adaptable, and secure AI systems. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Graph Neural Networks (GNNs) are a powerful class of artificial intelligence models that help analyze complex relationships within data, from understanding how our brains function to predicting molecular interactions or identifying financial anomalies. While these models have shown remarkable promise, their widespread application in the real world faces significant hurdles: they often struggle to process extremely large datasets, adapt to unseen data, and maintain reliability when faced with intentional disruptions or faulty information. This project aims to overcome these limitations by focusing directly on the data itself, rather than solely on refining the GNN models. By making graph data more compact, cleaner, and better aligned with learning objectives, this project will enable more efficient, accurate, and robust AI systems across critical domains such as healthcare, finance, and national security. The project will address core challenges of GNNs related to data scale, distribution, and quality through three research tasks. The first task will tackle scalability by identifying key structural properties necessary for effective learning and developing graph condensation methods that significantly reduce data size while automatically preserving critical information. The second task will investigate how distribution shifts relate to graph properties and will introduce new data augmentation and test-time adaptation strategies to enhance generalization under out-of-distribution conditions. The third task will focus on data quality by creating unsupervised graph purification techniques to remove adversarial perturbations and by designing detection mechanisms to identify and mitigate various types of attacks. This project will include comprehensive evaluations using publicly available datasets and real-world applications, supported by collaborations with academic institutions and industry partners. The project outcomes will complement existing model-centric approaches and promote more efficient, robust, and generalizable GNN solutions across domains such as finance, neuroscience, and cybersecurity. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Graph Neural Networks (GNNs) are a powerful class of artificial intelligence models that help analyze complex relationships within data, from understanding how our brains function to predicting molecular interactions or identifying financial anomalies. While these models have shown remarkable promise, their widespread application in the real world faces significant hurdles: they often struggle to process extremely large datasets, adapt to unseen data, and maintain reliability when faced with intentional disruptions or faulty information. This project aims to overcome these limitations by focusing directly on the data itself, rather than solely on refining the GNN models. By making graph data more compact, cleaner, and better aligned with learning objectives, this project will enable more efficient, accurate, and robust AI systems across critical domains such as healthcare, finance, and national security. The project will address core challenges of GNNs related to data scale, distribution, and quality through three research tasks. The first task will tackle scalability by identifying key structural properties necessary for effective learning and developing graph condensation methods that significantly reduce data size while automatically preserving critical information. The second task will investigate how distribution shifts relate to graph properties and will introduce new data augmentation and test-time adaptation strategies to enhance generalization under out-of-distribution conditions. The third task will focus on data quality by creating unsupervised graph purification techniques to remove adversarial perturbations and by designing detection mechanisms to identify and mitigate various types of attacks. This project will include comprehensive evaluations using publicly available datasets and real-world applications, supported by collaborations with academic institutions and industry partners. The project outcomes will complement existing model-centric approaches and promote more efficient, robust, and generalizable GNN solutions across domains such as finance, neuroscience, and cybersecurity. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Algorithmic and machine learning tools are increasingly central to high-stakes decision-making across domains such as healthcare, finance, and digital services. These tools depend on data that are often incomplete, unbalanced, or poorly aligned with the populations they impact. While existing methods typically treat datasets as externally fixed and data imbalances as externally-determined, this project views data as the outcome of a strategic production process that is shaped by the incentives of individuals, platforms, and institutions. By leveraging these incentives, the project aims to design data ecosystems that yield more representative and reliable datasets, better reflecting the needs of all Americans. The research will develop new theoretical foundations and algorithmic methods that improve data quality through incentives rather than constraints. It will also support education on data economics in the context of AI and foster early-career development through workshops and mentoring activities. The project integrates ideas from computer science, economics, and operations research to analyze incentive-driven data production and exchange. It investigates three key dimensions: (1) how individual data providers can shape outcomes through strategic data sharing; (2) how platforms can build robust data markets by aligning incentives between producers and consumers; and, (3) how large-scale effects such as platform competition and user interactions influence data quality and decision performance. These efforts will advance the scientific understanding of dynamic data environments and inform the design of more effective data-driven systems. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Generating nuanced yet interpretable hypotheses from noisy and high-dimensional observations is a core activity across numerous scientific disciplines. For example, across neuroscience, genomics, biomechanics, and ecology, researchers analyze images and videos to study phenomena ranging from the genetic expression of lab mice to global species distributions of birds. Useful scientific hypotheses are typically instantiated on symbolically interpretable concepts and attributes (such as stride periods and center of mass oscillations when studying kinematics of walking). A key challenge is to extract such symbolically interpretable hypotheses from raw high-dimensional observational data. To address this challenge, this project aims to develop a novel neurosymbolic programming framework, called foundation model programming, for generating symbolically interpretable scientific hypotheses from high dimensional observational data. The main idea is to represent interpretable hypotheses as neurosymbolic programs that use symbolic primitives as well as neural modules, including foundation models. The use of neural and foundation models allows the hypotheses to be full-stack, modeling both the extraction of relevant patterns and motifs from high-dimensional raw data and reasoning over those patterns. The core technical benefit of this approach is that it inherits both the contextual flexibility of modern foundation models given high-dimensional inputs, and the rigorous reasoning abilities offered by neurosymbolic approaches. The proposal is backed by a substantial amount of prior work on both neurosymbolic learning and applying foundational models in scientific domains, including foundation model-enabled hypothesis generation for low-dimensional data, self-supervised symbolic feature extraction from high-dimensional data, data-efficient expert-in-the-loop training approaches, and deep deployment into real scientific workflows. Importantly, the use of foundation models enables building methods that are more autonomous, and require fewer manual annotations by the expert. This project will develop algorithms that can jointly reason over what are the most useful symbolically interpretable concepts or motifs that can be extracted from the raw data, as well as how to compose those concepts into coherent hypotheses. This paradigm mirrors how humans develop hypotheses, by jointly establishing a discrete vocabulary of concepts (from continuous high dimensional descriptions) and reasoning over those concepts, thus enabling interpretability. This project will benchmark the developed methods on several scientific tasks and domains and collaborations. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Graphs are powerful tools for representing relationships in complex systems, from social networks to weather monitoring stations. Graph Neural Networks (GNNs) have emerged as effective methods for analyzing these interconnected systems, but their "black box" nature poses significant challenges in critical applications such as environmental monitoring, healthcare, and finance. This project develops a comprehensive framework for making GNN predictions explainable and trustworthy. The research addresses the urgent need for artificial intelligence systems that can not only make accurate predictions but also explain their reasoning in ways that domain experts can understand and verify. For instance, in South Florida's water management network where monitoring stations form a graph connected by hydrological pathways, emergency managers need to understand which stations and their interconnections are most influential in flood predictions. This capability is essential for building trust in these systems and ensuring their responsible deployment in applications that affect public safety and welfare. The project will train students in interdisciplinary research combining machine learning, information theory, and practical applications, while developing educational materials that bridge theoretical foundations with real-world implementations. This project establishes a unified framework using information theory concepts for explainable GNNs through two complementary research thrusts. The first thrust develops rigorous mathematical foundations for quantifying explainability in graph learning, including necessary and sufficient conditions for classifier explainability, methods to address out-of-distribution challenges, and ways to demonstrate how accurate the finding are. The second thrust translates these theoretical insights into practical architectures and algorithms, including computationally efficient explainers, generative models for robust explainers, and co-design frameworks that balance prediction accuracy with explainability. The research introduces novel concepts such as nonverbal signatures for characterizing explanation patterns and explanation-assisted learning mechanisms that leverage extracted explanations to improve model performance. Extensive evaluations will be conducted on benchmark datasets and specially curated weather forecasting datasets from South Florida's water management systems. The project advances the state-of-the-art by providing both theoretical rigor in quantifying explainability and practical solutions for deploying trustworthy GNN systems in critical applications. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF

Reinforcement Learning (RL) is a machine learning paradigm that strives to make optimal decision-making based on experience acting in an environment. In many cases, the "environment" refers to a simulator in the training stage and refers to the real world in the deployment stage. Training in the simulator brings a lot of advantages: lower cost, more safety, and more flexibility. However, it is almost impossible to design a perfect simulator that is identical to the real world. Thus, a decision-maker trained in the simulator may not function well in the real world. The discrepancy between the simulator and the real world is called the simulation-to-reality (sim-to-real) gap. This project will build new technologies to close the sim-to-real gap in both the training and the deployment stages. The research outcomes will benefit the development of next-generation RL techniques, which can improve the availability, applicability, and generalization of RL, and minimize the gap of RL between common practices and real-world practices. This project proposes to close the sim-to-real gap in reinforcement learning by three mechanisms: randomization, alignment, and derivation. Specifically, 1) the randomization mechanism generates a set of homogeneous simulators by original simulator parameter randomization. The simulator set will cover a wider range of state-action regions than the original simulator, have a larger overlap with the real-world environment, and thereafter result in a smaller sim-to-real gap. This mechanism is especially useful when the sim-to-real gap is large and the simulator is only accessible for training the simulator-optimal policy, but not accessible during the sim-to-real transfer process. 2) The alignment mechanism makes the simulator more like the real world during the transfer process. The alignment mechanism not only closes the sim-to-real gap but also is low-cost and high-efficiency, thus, accelerating the transfer process. This mechanism is especially useful when the sim-to-real gap is relatively small and the simulator is accessible in both simulator-optimal policy training and sim-to-real transfer. 3) The derivation mechanism directly derives an optimal policy from real-world offline data without any simulator. It first estimates state-action values from offline data and then derives the policy by function approximation. This mechanism is especially useful when offline data has been collected, but the real-world dynamics are unknown so it is unlikely to build a faithful simulator. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.