Digital Commons Network^™

Supercharging Simulation-Based Inference For Bayesian Optimal Experimental Design, Samuel Klein, Willie Neiswanger, Daniel Ratner, Michael Kagan, Sean Gasiorowski

Physics Faculty Publications

Bayesian optimal experimental design (BOED) seeks to maximize the expected information gain (EIG) of experiments. This requires a likelihood estimate, which in many settings is intractable. Simulation-based inference (SBI) provides powerful tools for this regime. However, existing work explicitly connecting SBI and BOED is restricted to a single contrastive EIG bound. We show that the EIG admits multiple formulations which can directly leverage modern SBI density estimators, encompassing neural posterior, likelihood, and ratio estimation. Building on this perspective, we define a novel EIG estimator using neural likelihood estimation. Further, we identify optimization as a key bottleneck of gradient based EIG …

Exploring Causes Of Beam Loss At Cebaf, C. Matthews, J. Benesch, R. M. Bodenstein, B. Freeman, D. Moser, D. Turner, K. Price, J. Samari, B. Terzić

Physics Faculty Publications

At Jefferson Lab, the Continuous Electron Beam Accelerator (CEBAF) features a unique design with two linear accelerators and two arc sections allowing for multiple turns of the electron beam, as well as four experimental end stations. This topology leads to increased beam losses, especially in the spreader and recombiner regions connecting the arcs to the LINACs and in the extraction regions connecting the experimental end stations to the accelerator. These losses result in equipment activation and operational interruptions. Recent upgrades to the facility’s diagnostic systems, including the addition of xenon ion chambers, have provided higher-resolution data regarding these loss events. …

Real-Space Imaging Of The Electron-Pair Density Hole In Molecular Auger-Meitner Decay, Mats Simmermacher, Nathan Goff, Andres Moreno Carrascosa, Elke Fasshauer, Thomas Northey, Lingyu Ma, Haiwang Yong, Brian Stankus, Asami Odate, Xuan Xu, Wenpeng Du, Kyle Acheson, Joseph C. Cooper, Daniel Ratner, Mengning Liang, Ruaridh Forbes, Michael P. Minitti, Adam Kirrander, Peter M. Weber

Physics Faculty Publications

Electrons in matter can rearrange extremely quickly under external perturbations, underpinning subsequent structural and chemical transformations. Coulomb interactions between neighbouring electrons often shape this response, giving rise to correlated motion and strongly affecting the distribution of electrons in the system. Here we show that non-resonant hard X-ray scattering can directly access changes in the radial electron-pair density during the rapid rearrangement of core and valence electrons. We do this by studying sulfur hexafluoride molecules undergoing Auger–Meitner decay. We exploit a second-order interaction between the X-ray photons and the molecules to trigger and probe the decay dynamics with a single pulse, …

Reaching The Intrinsic Performance Limits Of Superconducting Nanowire Single-Photon Detectors Up To 0.1 Mm Wide, Kristen M. Parzuchowski, Eli Mueller, Bakhrom G. Oripov, Benedikt Hampel, Ravin A. Chowdhury, Sahil R. Patel, Daniel Kuznesof, Emma K. Batson, Ryan Morgenstern, Robert H. Hadfield, Varun B. Verma, Matthew D. Shaw, Jason P. Allmaras, Martina J. Stevens, Alex Gurevich, Adam N. Mccaughan

Physics Faculty Publications

Superconducting nanowire single-photon detectors combine high detection efficiency, low noise, and excellent timing resolution, making them a leading platform for photon-counting applications. However, despite decades of materials and fabrication research, detector performance has never been shown to match theoretical performance expectations. Here, we demonstrate in situ tuning of a detector from its typical, suboptimal operation, to a regime limited only by material quality, allowing the device to reach its intrinsic performance limit. Our approach is based on current-biased superconducting “rails” placed on either side of the detector that redistribute current across its width to achieve peak performance. This technique reduces …

Rf Design Progress Of The 197 Mhz Crab Cavity For Eic, B. Xiao, J. C. Brutus, A. Blednykh, Z. Conway, J. Cox, J. R. Delayen, S. U. De Silva, R. Fernandes, J. Guo, D. Holmes, N. Huque, G. Li, Z. Li, J. B. Mandel, S. Verdú-Andrés, G. Wang, Q. Wu, W. Xu

Physics Faculty Publications

The interaction region (IR) crab cavity system is a special RF system designed to compensate for the luminosity loss caused by the 25 mrad crossing angle at the interaction point (IP) of the Electron-Ion collider (EIC). The configuration includes six crab cavities in the Hadron (proton or ion) Storage Ring (HSR) - four operating at 197 MHz and two at 394 MHz - installed on each side of the IP, along with one 394 MHz crab cavity on each side of the IP in the Electron Storage Ring (ESR). This paper presents the recent progress in the RF design of …

First Measurement Of Polarized Spin-Density Matrix Elements And Differential Cross Sections Dσ/Dt In Ω Photoproduction Off The Proton For 2.7 < E𝛾 < 5.2 Gev Using Clas At Jefferson Lab, T. Hu, Z. Akbar, V. Crede, J. M. Laget, V. Mathieu, M. J. Amaryan, W. R. Armstrong, H. Atac, N. A. Baltzell, L. Barion, M. Battaglieri, I. Bedlinskiy, B. Benkel, F. Benmokhtar, A. Bianconi, L. Biondo, A. Biselli, M. Bondi, F. Bossu, S. Boiarinov, W. J. Briscoe, S. Bueltmann, D. Bulumulla, V. D. Burkert, R. Capobianco, D. S. Carman, J. C. Carvajal, A. Celentano, T. Chetry, G. Ciullo, G. Clash, P. L. Cole, M. Contalbrigo, G. Costantini, A. D'Angelo, N. Dashyan, R. De Vita, M. Defurne, A. Deur, S. Diehl, C. Djalali, R. Dupre, H. Egiyan, A. El Alaoui, L. El Fassi, L. Elouadrhiri, P. Eugenio, S. Fegan, A. Filippi, G. Gavalian, Y. Ghandilyan, G. P. Gilfoyle, F. X. Girod, A. A. Golubenko, G. Gosta, R. W. Gothe, K. A. Griffioen, M. Guidal, K. Hafidi, H. Hakobyan, M. Hattawy, T. B. Hayward, D. Heddle, A. Hobart, M. Holtrop, Y. Ilieva, D. G. Ireland, E. L. Isupov, D. Jenkins, H. S. Jo, K. Joo, S. Joosten, D. Keller, A. Khanal, M. Khandaker, A. Kim, W. Kim, F. J. Klein, V. Klimenko, A. Kripko, V. Kubarovsky, V. Lagerquist, L. Lanza, M. Leali, P. Lenisa, K. Livingston, I.J.D. Macgregor, D. Marchand, L. Marsicano, V. Mascagna, M. E. Mccracken, B. Mckinnon, Z. E. Meziani, S. Migliorati, R. G. Milner, T. Mineeva, M. Mirazita, V. Mokeev, C. Munoz Camacho, P. Nadel-Turonski, K. Neupane, J. Newton, S. Niccoai, M. Nicol, G. Niculescu, M. Osipenko, A. I. Ostrovidov, P. Pandey, M. Paolone, L. L. Pappalardo, R. Paremuzyan, E. Pasyuk, S. J. Paul, W. Phelps, N. Pilleux, O. Pogorelko, J. W. Price, Y. Prok, B. A. Raue, T. Reed, M. Ripani, J. Ritman, A. Rizzo, G. Rosner, F. Sabatié, C. Salgado, S. Schadmand, A. Schmidt, R. A. Schumacher, E. V. Shirokov, U. Shrestha, P. Simmerling, D. Sokhan, N. Sparveris, S. Stepanyan, I. I. Strakovsky, S. Strauch, R. Tyson, M. Ungaro, L. Venturelli, H. Voskanyan, A. Vossen, E. Voutier, D. P. Watts, X. Wei, M. H. Wood, B. Yale, M. Yurov, N. Zachariou, J. Zhang, The Clas Collaboration At Jefferson Laboratory

Physics Faculty Publications

We report on the differential cross sections dσ/dt, the unpolarized spin-density matrix elements ρ⁰₀₀, ρ⁰₁₋₁, Re ρ⁰₁₀, and the first extraction of the polarized elements Im ρ³₁₀, Im ρ³₁₋₁ for the reaction γp → pω using the CLAS spectrometer at Jefferson Laboratory. The ω mesons were detected in their dominant charged decay mode, ω→π+π−π⁰, and all t-dependent results are presented in a fine binning for incident photon energies between 2.73 and 5.16 GeV (corresponding to the center-of-mass energy range W ∈ [ 2.45, 3.25 ] GeV). All matrix elements are first measurements for −t>0.6 GeV². …

Efficient Dynamic And Momentum Aperture Optimization For Lattice Design Using Multipoint Bayesian Algorithm Execution, Z. Zhang, I. Agapov, S. Gasiorowski, T. Hellert, W. Neiswanger, X. Huang, D. Ratner

Physics Faculty Publications

We demonstrate that multipoint Bayesian algorithm execution (multipointBAX) can overcome fundamental computational challenges in storage ring design optimization. Dynamic aperture and momentum aperture optimization is a multipoint, multiobjective design task for storage rings, ultimately informing the flux of x-ray sources and luminosity of colliders. Current production black-box optimization methods require extensive particle-tracking simulations for each trial configuration; the high computational cost restricts the extent of the search to ~ 10³ configurations and therefore limits the quality of the final design. We remove this bottleneck using multipointBAX, which selects, simulates, and models each trial configuration at the single-particle level. We demonstrate …

Interpolation Between Sudakov And Bfkl Rapidity Evolutions For Tmd Factorization At Small X, Ian Balitsky

Physics Faculty Publications

Within the framework of rapidity-only small-x TMD factorization, the evolution with respect to the rapidity cutoff is initially governed by Sudakov double logarithms and subsequently by BFKL/BK single logarithms. The evolution equation proposed in this paper correctly reproduces both the Sudakov and BFKL limits, while providing a consistent interpolation between these two regimes.

Taking Scale Seriously In Technology Law, Mark P. Mckenna, Woodrow Hartzog

Faculty Scholarship

Issues of scale—the relationship between the amount of an activity and its associated costs and benefits—permeate discussions around law and technologies. Indeed, it’s not much of an exaggeration to say that scale is the reason for most technology regulation.

But it’s not always clear how lawmakers and judges conceptualize “scale” when approaching questions around automated technologies. Scale is often used intuitively, just to mean “more.” But scale is not always just about more—scale can introduce new harms and benefits along different dimensions, not simply costs or efficiencies of greater magnitude.

In this Article, we argue for a more sustained interrogation …

Informed Applicants: Anti-Discrimination Enforcement Based On The Use Of Criminal Records, Benjamin David Pyle

Faculty Scholarship

Navigating the labor market with a criminal record is treacherous. Countless job applicants will be turned down immediately after a potential employer learns they have had contact with the criminal legal system. Criminal records reflect and exacerbate long-standing racial and economic inequalities in the United States. Thus, the use of records in employment decisions will often implicate anti-discrimination laws. However, successfully enforcing these protections requires job applicants to, at minimum, understand why they were not hired. Many applicants never learn why they were not hired and thus will never be able to protect their rights. This Article argues that understanding …

Against Ai Half Measures, Woodrow Hartzog, Neil Richards, Ryan Durrie, Jordan Francis

Faculty Scholarship

So far, U.S. consumer protection policy for artificial intelligence (AI) accountability has largely consisted of industry-led approaches such as encouraging transparency, mitigating bias, promoting principles of ethics, and empowering people. These approaches are vital, but they are only half measures. To bring AI within the rule of law, lawmakers must start drawing substantive lines.

In this Article, we identify four AI regulatory approaches to consumer, data, and democratic harms as half measures. First, transparency does not produce accountability on its own. Second, while mitigating bias in AI systems is critical, even unbiased systems are a threat to the vulnerable. Third, …

Obscurity And The Fallacy Of No Privacy In Public Information, Woodrow Hartzog

Faculty Scholarship

The main thesis of this chapter is that because there are so many different possible interpretations of ‘public information’, the concept cannot be used to justify data practices and surveillance without first articulating a more precise meaning that recognises what is at stake. By disposing of the myth that there is an objective and dispassionate concept of public information, judges and lawmakers can clear the way for information rules based on overt value choices. In short, if the concept of ‘public’ is going to shape people’s social and legal obligations, its meaning and neutrality should not be assumed.

My argument …

The Punishment Paradox, Steven Arrigg Koh

Faculty Scholarship

What does criminal punishment do? Across centuries of criminal legal thought, a shared answer persists: punishment incapacitates the wrongdoer. And yet punishment sometimes yields its opposite. Paradoxically, punishment sometimes becomes a rallying point-expanding community identification and solidarity around the defendant. This Article shows this punishment paradox using the 2023-24 prosecutions of President Trump and recent and historical case examples-from Luigi Mangione to Karen Read to Rosa Parks. By arguing that criminal law can generate solidarity rather than stigma, the Article both rethinks foundational theories of punishment and reconsiders the role of criminal law in democratic politics. Such an approach transcends …

The Hypocrisy Of Data Governance, Zubair Shafiq, Olivia Figueira, Athina Markopoulou, Woodrow Hartzog, Michael Lavine

Faculty Scholarship

“Data governance” is an empty term, like a Rorschach inkblot just waiting to be filled with meaning. Tech companies take advantage of this ambiguity to craft narratives about their data-governance capabilities to fit their audience and purpose. On one hand, tech companies brag about their data-governance capabilities when it fits their business model (for example, to advertisers) and public image (for example, to their customers). On the other hand, tech companies claim that meaningful data governance is challenging or impossible when accountability is demanded.

In this Article, we argue that tech companies systematically misrepresent or selectively ignore their data-governance capabilities. …

Small Language Model Jailbreak Defender Framework, Eduardo Alberto Castro Puello

CCAC Theses and Dissertations

Small Language Models (SLMs) are increasingly deployed in enterprise, edge, and domain‑specialized applications due to their lower computational cost and latency compared to Large Language Models (LLMs). However, SLMs typically exhibit weaker safety alignment and are disproportionately vulnerable to jailbreak, prompt‑injection, and fine‑tuning attacks that can erode refusal behavior and enable unsafe outputs. Existing defensive techniques such as rule‑based filtering or heavy safety fine‑tuning often degrade task performance or introduce excessive over‑refusal, limiting the safe adoption of SLMs in regulated or security‑sensitive environments. This dissertation proposes the design and empirical evaluation of a modular security alignment framework for fine‑tuned SLMs …

Comparative Evaluation Of Data Augmentation Strategies For Rare-Class Intrusion Detection: Assessing Conditional Generative Adversarial Networks Across Benchmark Datasets And Classifiers, Brogan Oconnor

CCAC Theses and Dissertations

Intrusion Detection Systems (IDS) are widely used to identify malicious network activity, unauthorized access, and anomalous system behavior. Although machine learning and deep learning classifiers often achieve strong aggregate performance on IDS benchmark datasets, rare attack detection remains difficult because of severe class imbalance. Minority attack categories may contain very few training observations, which can bias classifiers toward normal traffic and more frequent attack classes. This problem is especially important because rare attacks may represent high-impact security events even when they occur infrequently. This quantitative experimental study evaluated whether conditional generative adversarial network (cGAN) augmentation improved rare attack detection in …