Search

Controlling wave transmission is crucial for various applications. In this work, we apply the concept of unitary control to manipulate multiport wave transmission. Unitary control aims to control the behaviors of a set of orthogonal waves simultaneously. The approach fully harnesses the capability of wavefront shaping techniques, with promising applications in communication, imaging, and photonic integrated circuits. Here we present a detailed theory of unitary control of wave transmission, focusing on two key characteristics: total (power) transmittance and direct (field) transmission. The total transmittance for an input port represents the fraction of total transmitted power with respect to the input power for wave incident from an input port. The direct transmission for an input port denotes the complex transmission amplitude from that input port to its corresponding output port. We address two main questions: (i) the achievable total transmittance and direct transmission for each port, and (ii) the configuration of unitary control to attain desired transmission values for each port. Our theory illustrates that unitary control enables uniform total transmittance and direct transmission across any medium. Furthermore, we show that reciprocity and energy conservation enforce direct symmetry constraints on wave transmission in both forward and backward directions under unitary control., Comment: 29 pages, 3 figures

We investigate the majorization order for comparing wave coherence and reveal its fundamental consequences in transport measurements, including power distribution, absorption, transmission, and reflection. We prove that all these measurements preserve the majorization order under unitary control, enabling direct experimental characterization of the majorization order. Specifically, waves with lower coherence in the majorization order exhibit more restricted ranges of achievable measurement values. Our results deepen the understanding of coherence in transport phenomena., Comment: 11 pages, 3 figures

Entanglement is a unique feature of quantum mechanics. In coupled systems of light and matter, entanglement manifests itself in the linear superposition of multipartite quantum states (e.g., parametrized by the multiple spatial, spectral, or temporal degrees of freedom of a light field). In bipartite systems, the Schmidt decomposition provides a modal decomposition of the entanglement structure over independent, separable states. Although ubiquitous as a tool to describe and measure entanglement, there exists no general method to decompose a bipartite quantum state onto its Schmidt modes. Here, we propose a method that relies on bipartite self-configuring optics that automatically "learns" the Schmidt decomposition of an arbitrary pure quantum state. Our method is agnostic to the degrees of freedom over which quantum entanglement is distributed and can reconstruct the Schmidt modes and values by variational optimization of the network's output powers or coincidences. We illustrate our method with numerical examples of spectral entanglement analysis for biphotons generated via spontaneous parametric down conversion and provide experimental guidelines for its realization, including the influence of losses and impurities. Our method provides a versatile and scalable way of analyzing entanglement in bipartite integrated quantum photonic systems.

Backdoor data poisoning, inserted within instruction examples used to fine-tune a foundation Large Language Model (LLM) for downstream tasks (\textit{e.g.,} sentiment prediction), is a serious security concern due to the evasive nature of such attacks. The poisoning is usually in the form of a (seemingly innocuous) trigger word or phrase inserted into a very small fraction of the fine-tuning samples from a target class. Such backdoor attacks can: alter response sentiment, violate censorship, over-refuse (invoke censorship for legitimate queries), inject false content, or trigger nonsense responses (hallucinations). In this work we investigate the efficacy of instruction fine-tuning backdoor attacks as attack "hyperparameters" are varied under a variety of scenarios, considering: the trigger location in the poisoned examples; robustness to change in the trigger location, partial triggers, and synonym substitutions at test time; attack transfer from one (fine-tuning) domain to a related test domain; and clean-label vs. dirty-label poisoning. Based on our observations, we propose and evaluate two defenses against these attacks: i) a \textit{during-fine-tuning defense} based on word-frequency counts that assumes the (possibly poisoned) fine-tuning dataset is available and identifies the backdoor trigger tokens; and ii) a \textit{post-fine-tuning defense} based on downstream clean fine-tuning of the backdoored LLM with a small defense dataset. Finally, we provide a brief survey of related work on backdoor attacks and defenses., Comment: Under review

The circular restricted three-body problem (CR3BP) with solar radiation pressure (SRP) has often been analyzed with assumptions made on a spacecraft's attitude, such that the problem remains Hamiltonian. These assumptions are unsatisfactorily limiting for a starshade mission since the starshade's attitude will inherently vary from the configuration that corresponds to Hamiltonian dynamics. This paper presents the derivation of the equations of motion for CR3BP with SRP that permit the application of the Lindstedt-Poincare method, such that approximate solutions are produced, which may serve as invaluable trajectory design tools. Examples of periodic orbits and manifolds corresponding to three sets of attitude angles are shown and the accuracy of their seventh-order approximations is considered.

A variety of defenses have been proposed against backdoors attacks on deep neural network (DNN) classifiers. Universal methods seek to reliably detect and/or mitigate backdoors irrespective of the incorporation mechanism used by the attacker, while reverse-engineering methods often explicitly assume one. In this paper, we describe a new detector that: relies on internal feature map of the defended DNN to detect and reverse-engineer the backdoor and identify its target class; can operate post-training (without access to the training dataset); is highly effective for various incorporation mechanisms (i.e., is universal); and which has low computational overhead and so is scalable. Our detection approach is evaluated for different attacks on benchmark CIFAR-10 and CIFAR-100 image classifiers.

Optical phenomena always display some degree of partial coherence between their respective degrees of freedom. Partial coherence is of particular interest in multimodal systems, where classical and quantum correlations between spatial, polarization, and spectral degrees of freedom can lead to fascinating phenomena (e.g., entanglement) and be leveraged for advanced imaging and sensing modalities (e.g., in hyperspectral, polarization, and ghost imaging). Here, we present a universal method to analyze, process, and generate spatially partially coherent light in multimode systems by using self-configuring optical networks. Our method relies on cascaded self-configuring layers whose average power outputs are sequentially optimized. Once optimized, the network separates the input light into its mutually incoherent components, which is formally equivalent to a diagonalization of the input density matrix. We illustrate our method with arrays of Mach-Zehnder interferometers and show how this method can be used to perform partially coherent environmental light sensing, generation of multimode partially coherent light with arbitrary coherency matrices, and unscrambling of quantum optical mixtures. We provide guidelines for the experimental realization of this method, paving the way for self-configuring photonic devices that can automatically learn optimal modal representations of partially coherent light fields.

We study an effective theory of flavour in which the $SU(2)_L$ interaction is `flavour-deconstructed' near the TeV scale. This arises, for example, in UV models that unify all three generations of left-handed fermions via an $Sp(6)_L$ symmetry. Flavour-universality of the electroweak force emerges accidentally (but naturally) from breaking the $\prod_{i=1}^3 SU(2)_{L,i}$ gauge group to its diagonal subgroup, delivering hierarchical fermion masses and left-handed mixing angles in the process. The heavy gauge bosons transform as two $SU(2)_L$ triplets that mediate new flavour non-universal forces. The lighter of these couples universally to the light generations, allowing consistency with flavour bounds even for a TeV scale mass. Constraints from flavour, high mass LHC searches, and electroweak precision are then highly complementary, excluding masses below 9 TeV. The heavier triplet must instead be hundreds of TeV to be consistent with meson mixing constraints. Because only the lighter triplet couples to the Higgs, we find radiative Higgs mass corrections of a few hundred GeV, meaning this model of flavour is arguably natural. The natural region will, however, be almost completely covered by the planned electroweak programme at FCC-ee. On shorter timescales, significant parameter space will be explored by the High-Luminosity LHC measurements at high-$p_T$, and upcoming lepton flavour violation experiments, principally Mu3e., Comment: 50 pages. v2: version accepted in JHEP. Changed plot ranges to remove physically equivalent regions; added more details on some electroweak observables; fixed typos

Driven by the desire to find positions that satisfy keepout constraints for a space-based telescope mission, this work develops a process for tracing a point in space in the regime of the restricted three-body problem to a halo orbit, characterized by its out-of-plane amplitude, and its position on that halo orbit, denoted by the halo orbit time. This process utilizes third-order solutions from the Lindstedt-Poincare method, which have been partially inverted to expect a point in space as an input. Three different methodologies that use these partially inverted expressions are presented. Results are produced for 1,000 randomly selected points using all three methods and are compared to truth. Ultimately, the method that employed two distinct accuracy metrics yielded the most accurate results for the dataset., Comment: Submission for the 33rd Space Flight Mechanics Meeting

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of the Gemini family in cross-modal reasoning and language understanding will enable a wide variety of use cases. We discuss our approach toward post-training and deploying Gemini models responsibly to users through services including Gemini, Gemini Advanced, Google AI Studio, and Cloud Vertex AI.

We solve a long-standing set of problems in optics and waves: why does a volume have only so many useful orthogonal wave channels in or out of it, why do coupling strengths fall off dramatically past this number, and, indeed, just what precisely defines that number? Increasingly in applications in communications, information processing, and sensing, in optics, acoustics, and electromagnetic waves generally, we need to understand this number. We can numerically find such channels for many problems, but more fundamentally, these questions have arguably never had a clear answer or physical explanation. We have found a simple and general result and intuition that lets us understand and bound this behavior for any volume. This is based on a tunneling that has been somewhat hidden in the mathematics of spherical waves: beyond a certain complexity of the wave, it must tunnel to escape the volume. By counting the number of waves that do not have to tunnel, we get a simple and precise number or bound for well-coupled channels, even for arbitrary volumes. The necessary tunneling for other waves explains the rapid fall-off in their coupling, and shows all such waves do escape to propagation to some degree after tunneling. This approach connects multipole expansions in electromagnetic antennas and nanophotonics smoothly to apparently evanescent waves in large optics. It works over all size scales, from nanophotonics, small radio-frequency antennas, or acoustic microphones and loudspeakers up to imaging optics with millions of channels, and gives a precise diffraction limit for any volume.

We present first results from a dark photon dark matter search in the mass range from 44 to 52 $\mu{\rm eV}$ ($10.7 - 12.5\,{\rm GHz}$) using a room-temperature dish antenna setup called GigaBREAD. Dark photon dark matter converts to ordinary photons on a cylindrical metallic emission surface with area $0.5\,{\rm m}^2$ and is focused by a novel parabolic reflector onto a horn antenna. Signals are read out with a low-noise receiver system. A first data taking run with 24 days of data does not show evidence for dark photon dark matter in this mass range, excluding dark photon - photon mixing parameters $\chi \gtrsim 10^{-12}$ in this range at 90% confidence level. This surpasses existing constraints by about two orders of magnitude and is the most stringent bound on dark photons in this range below 49 $\mu$eV., Comment: 6 pages, 4 figures, matches published version

We searched the Gaia DR3 database for ultramassive white dwarfs with kinematics consistent with having escaped the nearby Hyades open cluster, identifying three such candidates. Two of these candidates have masses estimated from Gaia photometry of approximately 1.1 solar masses; their status as products of single stellar evolution that have escaped the cluster was deemed too questionable for immediate follow-up analysis. The remaining candidate has an expected mass >1.3 solar masses, significantly reducing the probability of it being an interloper. Analysis of follow-up Gemini GMOS spectroscopy for this source reveals a non-magnetized hydrogen atmosphere white dwarf with a mass and age consistent with having formed from a single star. Assuming a single-stellar evolution formation channel, we estimate a 97.8% chance that the candidate is a true escapee from the Hyades. With a determined mass of 1.317 solar masses, this is potentially the most massive known single-evolution white dwarf and is by far the most massive with a strong association with an open cluster., Comment: 13 pages, 4 figures, 1 table. Accepted for publication in the Astrophysical Journal Letters

Well-known (non-malicious) sources of overfitting in deep neural net (DNN) classifiers include: i) large class imbalances; ii) insufficient training-set diversity; and iii) over-training. In recent work, it was shown that backdoor data-poisoning also induces overfitting, with unusually large classification margins to the attacker's target class, mediated particularly by (unbounded) ReLU activations that allow large signals to propagate in the DNN. Thus, an effective post-training (with no knowledge of the training set or training process) mitigation approach against backdoors was proposed, leveraging a small clean dataset, based on bounding neural activations. Improving upon that work, we threshold activations specifically to limit maximum margins (MMs), which yields performance gains in backdoor mitigation. We also provide some analytical support for this mitigation approach. Most importantly, we show that post-training MM-based regularization substantially mitigates non-malicious overfitting due to class imbalances and overtraining. Thus, unlike adversarial training, which provides some resilience against attacks but which harms clean (attack-free) generalization, we demonstrate an approach originating from adversarial learning that helps clean generalization accuracy. Experiments on CIFAR-10 and CIFAR-100, in comparison with peer methods, demonstrate strong performance of our methods.

Backdoor (Trojan) attacks are an important type of adversarial exploit against deep neural networks (DNNs), wherein a test instance is (mis)classified to the attacker's target class whenever the attacker's backdoor trigger is present. In this paper, we reveal and analyze an important property of backdoor attacks: a successful attack causes an alteration in the distribution of internal layer activations for backdoor-trigger instances, compared to that for clean instances. Even more importantly, we find that instances with the backdoor trigger will be correctly classified to their original source classes if this distribution alteration is corrected. Based on our observations, we propose an efficient and effective method that achieves post-training backdoor mitigation by correcting the distribution alteration using reverse-engineered triggers. Notably, our method does not change any trainable parameters of the DNN, but achieves generally better mitigation performance than existing methods that do require intensive DNN parameter tuning. It also efficiently detects test instances with the trigger, which may help to catch adversarial entities in the act of exploiting the backdoor.

Deep neural networks are vulnerable to backdoor attacks (Trojans), where an attacker poisons the training set with backdoor triggers so that the neural network learns to classify test-time triggers to the attacker's designated target class. Recent work shows that backdoor poisoning induces over-fitting (abnormally large activations) in the attacked model, which motivates a general, post-training clipping method for backdoor mitigation, i.e., with bounds on internal-layer activations learned using a small set of clean samples. We devise a new such approach, choosing the activation bounds to explicitly limit classification margins. This method gives superior performance against peer methods for CIFAR-10 image classification. We also show that this method has strong robustness against adaptive attacks, X2X attacks, and on different datasets. Finally, we demonstrate a method extension for test-time detection and correction based on the output differences between the original and activation-bounded networks. The code of our method is online available.

PELICAN is a novel permutation equivariant and Lorentz invariant or covariant aggregator network designed to overcome common limitations found in architectures applied to particle physics problems. Compared to many approaches that use non-specialized architectures that neglect underlying physics principles and require very large numbers of parameters, PELICAN employs a fundamentally symmetry group-based architecture that demonstrates benefits in terms of reduced complexity, increased interpretability, and raw performance. We present a comprehensive study of the PELICAN algorithm architecture in the context of both tagging (classification) and reconstructing (regression) Lorentz-boosted top quarks, including the difficult task of specifically identifying and measuring the $W$-boson inside the dense environment of the Lorentz-boosted top-quark hadronic final state. We also extend the application of PELICAN to the tasks of identifying quark-initiated vs.~gluon-initiated jets, and a multi-class identification across five separate target categories of jets. When tested on the standard task of Lorentz-boosted top-quark tagging, PELICAN outperforms existing competitors with much lower model complexity and high sample efficiency. On the less common and more complex task of 4-momentum regression, PELICAN also outperforms hand-crafted, non-machine learning algorithms. We discuss the implications of symmetry-restricted architectures for the wider field of machine learning for physics., Comment: 52 pages, 34 figures, 12 tables

An embedding is a mapping from a set of nodes of a network into a real vector space. Embeddings can have various aims like capturing the underlying graph topology and structure, node-to-node relationship, or other relevant information about the graph, its subgraphs or nodes themselves. A practical challenge with using embeddings is that there are many available variants to choose from. Selecting a small set of most promising embeddings from the long list of possible options for a given task is challenging and often requires domain expertise. Embeddings can be categorized into two main types: classical embeddings and structural embeddings. Classical embeddings focus on learning both local and global proximity of nodes, while structural embeddings learn information specifically about the local structure of nodes' neighbourhood. For classical node embeddings there exists a framework which helps data scientists to identify (in an unsupervised way) a few embeddings that are worth further investigation. Unfortunately, no such framework exists for structural embeddings. In this paper we propose a framework for unsupervised ranking of structural graph embeddings. The proposed framework, apart from assigning an aggregate quality score for a structural embedding, additionally gives a data scientist insights into properties of this embedding. It produces information which predefined node features the embedding learns, how well it learns them, and which dimensions in the embedded space represent the predefined node features. Using this information the user gets a level of explainability to an otherwise complex black-box embedding algorithm., Comment: 21 pages, 9 figures

Optical phase measurement is critical for many applications and traditional approaches often suffer from mechanical instability, temporal latency, and computational complexity. In this paper, we describe compact phase sensor arrays based on integrated photonics, which enable accurate and scalable reference-free phase sensing in a few measurement steps. This is achieved by connecting multiple two-port phase sensors into a graph to measure relative phases between neighboring and distant spatial locations. We propose an efficient post-processing algorithm, as well as circuit design rules to reduce random and biased error accumulations. We demonstrate the effectiveness of our system in both simulations and experiments with photonic integrated circuits. The proposed system measures the optical phase directly without the need for external references or spatial light modulators, thus providing significant benefits for applications including microscope imaging and optical phased arrays.

Nanoelectromechanical (NEMS) resonator networks have drawn increasing interest due to their potential applications in emergent behavior, sensing, phononics, and mechanical information processing. A challenge toward realizing these large-scale networks is the ability to controllably tune and reconfigure the collective, macroscopic properties of the network, which relies directly on the development of methods to characterize the constituent NEMS resonator building blocks and their coupling. In this work, we demonstrate a scalable optical technique to spatially map graphene NEMS networks and read out the fixed-frequency collective response as a single vector. Using the response vectors, we introduce an efficient algebraic approach to quantify the site-specific elasticity, mass, damping, and coupling parameters of network clusters. We apply this technique to accurately characterize single uncoupled resonators and coupled resonator pairs by sampling them at just two frequencies, and without the use of curve fitting or the associated a priori parameter estimates. Our technique may be applied to a range of classical and quantum resonator systems and fills in a vital gap for programming NEMS networks., Comment: Main--15 pages, 3 figures; Supplementary Information--17 pages, 7 figures

Covariate adjustment and methods of incorporating historical data in randomized clinical trials (RCTs) each provide opportunities to increase trial power. We unite these approaches for the analysis of RCTs with binary outcomes based on the Cochran-Mantel-Haenszel (CMH) test for marginal risk ratio (RR). In PROCOVA-CMH, subjects are stratified on a single prognostic covariate reflective of their predicted outcome on the control treatment (e.g. placebo). This prognostic score is generated based on baseline covariates through a model trained on historical data. We propose two closed-form prospective estimators for the asymptotic sampling variance of the log RR that rely only on values obtainable from observed historical outcomes and the prognostic model. Importantly, these estimators can be used to inform sample size during trial planning. PROCOVA-CMH demonstrates type I error control and appropriate asymptotic coverage for valid inference. Like other covariate adjustment methods, PROCOVA-CMH can reduce the variance of the treatment effect estimate when compared to an unadjusted (unstratified) CMH analysis. In addition to statistical methods, simulations and a case study in Alzheimer's Disease are given to demonstrate performance. Results show that PROCOVA-CMH can provide a gain in power, which can be used to conduct smaller trials., Comment: 18 pages, 11 tables, appendix

Many current approaches to machine learning in particle physics use generic architectures that require large numbers of parameters and disregard underlying physics principles, limiting their applicability as scientific modeling tools. In this work, we present a machine learning architecture that uses a set of inputs maximally reduced with respect to the full 6-dimensional Lorentz symmetry, and is fully permutation-equivariant throughout. We study the application of this network architecture to the standard task of top quark tagging and show that the resulting network outperforms all existing competitors despite much lower model complexity. In addition, we present a Lorentz-covariant variant of the same network applied to a 4-momentum regression task.

Programmable unitary photonic networks that interfere hundreds of modes are emerging as a key technology in energy-efficient sensing, machine learning, cryptography, and linear optical quantum computing applications. In this work, we establish a theoretical framework to quantify error tolerance and scalability in a more general class of "binary tree cascade'' programmable photonic networks that accept up to tens of thousands of discrete input modes $N$. To justify this scalability claim, we derive error tolerance and configuration time that scale with $\log_2 N$ for balanced trees versus $N$ in unbalanced trees, despite the same number of total components. Specifically, we use second-order perturbation theory to compute phase sensitivity in each waveguide of balanced and unbalanced networks, and we compute the statistics of the sensitivity given random input vectors. We also evaluate such networks after they self-correct, or self-configure, themselves for errors in the circuit due to fabrication error and environmental drift. Our findings have important implications for scaling photonic circuits to much larger circuit sizes; this scaling is particularly critical for applications such as principal component analysis and fast Fourier transforms, which are important algorithms for machine learning and signal processing., Comment: 32 pages, 12 figures

Many real-world systems exhibit noisy evolution; interpreting their finite-time behavior as arising from continuous-time processes (in the It\^o or Stratonovich sense) has led to significant success in modeling and analysis in a variety of fields. Here we argue that a class of differential equations where evolution depends nonlinearly on a random or effectively-random quantity may exhibit finite-time stochastic behavior in line with an equivalent It\^o process, which is of great utility for their numerical simulation and theoretical analysis. We put forward a method for this conversion, develop an equilibrium-moment relation for It\^o attractors, and show that this relation holds for our example system. This work enables the theoretical and numerical examination of a wide class of mathematical models which might otherwise be oversimplified due to a lack of appropriate tools., Comment: 13 pages, 6 figures

We show why and when optics needs thickness as well as width or area. Wave diffraction explains the fundamental need for area or diameter of a lens or aperture to achieve some resolution or number of pixels in microscopes and cameras. Now we show that, if we know what the optics is to do, even before design, we can also deduce minimum required thickness. This limit comes from diffraction together with a novel concept called "overlapping non-locality" C that can be deduced rigorously just from the mathematical description of what the device is to do. C expresses how much the input regions for different output regions overlap. This limit applies broadly to optics from cameras to metasurfaces, and to wave systems generally.

Covariate adjustment is desired by both practitioners and regulators of randomized clinical trials because it improves precision for estimating treatment effects. However, covariate adjustment presents a particular challenge in time-to-event analysis. We propose to apply covariate adjusted pseudovalue regression to estimate between-treatment difference in restricted mean survival times (RMST). Our proposed method incorporates a prognostic covariate to increase precision of treatment effect estimate, maintaining strict type I error control without introducing bias. In addition, the amount of increase in precision can be quantified and taken into account in sample size calculation at the study design stage. Consequently, our proposed method provides the ability to design smaller randomized studies at no expense to statistical power.