Search for: All records

Robotic systems for long-term monitoring in dense natural environments require a minimal physical footprint with the ability to maneuver through confined spaces, while being robust to harsh conditions. This work presents the design, modeling, and implementation of a compact, tendon driven continuum robot capable of controlled extension and spatial bending within a self-contained, protective actuation package. The system is comprised of a meter-long, single section, three-tendon “tendril” that deploys from a central reel mechanism. Integrated load cells provide tension sensing, and a tip camera supports control feedback and data collection. A Denavit-Hartenberg-based kinematic model is adapted to enable motion planning for the variable-length continuum section. To compensate for modeling inaccuracies arising from the inherent compliance of the flexible tendril, a closed-form tension distribution model is introduced to support feedback based refinement of the robot pose. The robot prototype demonstrated navigating cluttered natural environments and was deployed outdoors over a total of five months, 

Sanitizing sensitive text data typically involves removing personally identifiable information (PII) or generating synthetic data under the assumption that these methods adequately protect privacy; however, their effectiveness is often only assessed by measuring the leakage of explicit identifiers but ignoring nuanced textual markers that can lead to re-identification. We challenge the above illusion of privacy by proposing a new framework that evaluates re-identification attacks to quantify individual privacy risks upon data release. Our approach shows that seemingly innocuous auxiliary information -- such as routine social activities -- can be used to infer sensitive attributes like age or substance use history from sanitized data. For instance, we demonstrate that Azure's commercial PII removal tool fails to protect 74\% of information in the MedQA dataset. Although differential privacy mitigates these risks to some extent, it significantly reduces the utility of the sanitized text for downstream tasks. Our findings indicate that current sanitization techniques offer a \textit{false sense of privacy}, highlighting the need for more robust methods that protect against semantic-level information leakage. 

We report features of the design, construction, installation, and performance of the BigBite Calorimeter (BBCal), a lead-glass electromagnetic calorimeter constructed as part of the BigBite Spectrometer (BBS), which served as the electron arm for the Super Bigbite Spectrometer (SBS) program of high-precision neutron electromagnetic form factor measurements in Hall A at Jefferson Lab. As a total-absorption calorimeter, BBCal provided the primary electron trigger for BBS, detecting quasi-elastically scattered electrons in the 1–4 GeV energy range with an energy resolution of approximately 6.2%, position resolution of 1.2 cm, and timing resolution of 0.5 ns. 

Prediction-Powered Inference (PPI) is a powerful framework for enhancing statistical estimates by combining limited gold-standard data with machine learning (ML) predictions. While prior work has demonstrated PPI's benefits for individual statistical problems, modern applications require answering numerous parallel statistical questions. We introduce Prediction-Powered Adaptive Shrinkage (PAS), a method that bridges PPI with empirical Bayes shrinkage to improve the estimation of multiple means. PAS debiases noisy ML predictions within each task and then borrows strength across tasks by using those same predictions as a reference point for shrinkage. The amount of shrinkage is determined by minimizing an unbiased estimate of risk, and we prove that this tuning strategy is asymptotically optimal. Experiments on both synthetic and real-world datasets show that PAS adapts to the reliability of the ML predictions and outperforms traditional and modern baselines in large-scale applications. 

Icosahedral capsids are ubiquitous among spherical viruses, yet their assem- bly pathways and governing interactions remain elusive. We present a molecular dynamics model that incorporates essential physical and biological interactions, including protein diffusion, genome flexibility, and a conformational switch that mimics allostery and activates the elastic properties of proteins upon binding. This switch makes the simulations computationally feasible and enables the assembly of icosahedral capsids around a flexible genome—overcoming long-standing lim- itations in previous models. Using this framework, we successfully reproduce the self-assembly of subunits around a flexible genome into icosahedral shells with numbers greater than one – most notably 3, the most common structure in na- ture – a feat that rigid-body models have so far failed to achieve. We systematically explore the range of morphologies formed with different genome architectures, in line with in vitro experiments using cowpea chlorotic mottle virus capsid proteins: viral RNAs with more complex structure form more complete and stable capsids than linear ones. These results provide a predictive framework for genome-guided assembly and capsid design.