Search

"Magnetic Reading" is an evidence-based reading comprehension program for students in grades 3 through 5 built on four pedagogical pillars: knowledge-rich learning, culturally and linguistically responsive pedagogy, scaffolds to support learner variability, and data to inform instruction. This study uses a prospective matched comparison group quasi-experimental design to examine the impact of "Magnetic Reading." At the beginning of the 2022-23 school year, we used coarsened exact matching to match students in schools that used "Magnetic Reading" to students in schools that did not using student demographic characteristics and a fall achievement test score. We subsequently conducted an analysis of spring state achievement test scores on the post-attrition sample after establishing that the groups were similar on baseline characteristics. Our HLM analysis found a statistically significant impact of 8.4 points--an effect size of 0.22 standard deviations--on reading scores. [This report was produced by NORC Education and presented to Curriculum Associates.]

Today's mainstream network timing models for distributed computing are synchrony, partial synchrony, and asynchrony. These models are coarse-grained and often make either too strong or too weak assumptions about the network. This paper introduces a new timing model called granular synchrony that models the network as a mixture of synchronous, partially synchronous, and asynchronous communication links. The new model is not only theoretically interesting but also more representative of real-world networks. It also serves as a unifying framework where current mainstream models are its special cases. We present necessary and sufficient conditions for solving crash and Byzantine fault-tolerant consensus in granular synchrony. Interestingly, consensus among $n$ parties can be achieved against $f \geq n/2$ crash faults or $f \geq n/3$ Byzantine faults without resorting to full synchrony.

Machine Learning (ML) has the potential to revolutionise the field of automotive aerodynamics, enabling split-second flow predictions early in the design process. However, the lack of open-source training data for realistic road cars, using high-fidelity CFD methods, represents a barrier to their development. To address this, a high-fidelity open-source (CC-BY-SA) public dataset for automotive aerodynamics has been generated, based on 500 parametrically morphed variants of the widely-used DrivAer notchback generic vehicle. Mesh generation and scale-resolving CFD was executed using consistent and validated automatic workflows representative of the industrial state-of-the-art. Geometries and rich aerodynamic data are published in open-source formats. To our knowledge, this is the first large, public-domain dataset for complex automotive configurations generated using high-fidelity CFD.

The NANOGrav 15-year data provides compelling evidence for a stochastic gravitational-wave (GW) background at nanohertz frequencies. The simplest model-independent approach to characterizing the frequency spectrum of this signal consists in a simple power-law fit involving two parameters: an amplitude A and a spectral index \gamma. In this paper, we consider the next logical step beyond this minimal spectral model, allowing for a running (i.e., logarithmic frequency dependence) of the spectral index, \gamma_run(f) = \gamma + \beta \ln(f/f_ref). We fit this running-power-law (RPL) model to the NANOGrav 15-year data and perform a Bayesian model comparison with the minimal constant-power-law (CPL) model, which results in a 95% credible interval for the parameter \beta consistent with no running, \beta \in [-0.80,2.96], and an inconclusive Bayes factor, B(RPL vs. CPL) = 0.69 +- 0.01. We thus conclude that, at present, the minimal CPL model still suffices to adequately describe the NANOGrav signal; however, future data sets may well lead to a measurement of nonzero \beta. Finally, we interpret the RPL model as a description of primordial GWs generated during cosmic inflation, which allows us to combine our results with upper limits from big-bang nucleosynthesis, the cosmic microwave background, and LIGO-Virgo-KAGRA., Comment: 17 pages, 4 figures, 2 tables

Language Model (LM) agents for cybersecurity that are capable of autonomously identifying vulnerabilities and executing exploits have the potential to cause real-world impact. Policymakers, model providers, and other researchers in the AI and cybersecurity communities are interested in quantifying the capabilities of such agents to help mitigate cyberrisk and investigate opportunities for penetration testing. Toward that end, we introduce Cybench, a framework for specifying cybersecurity tasks and evaluating agents on those tasks. We include 40 professional-level Capture the Flag (CTF) tasks from 4 distinct CTF competitions, chosen to be recent, meaningful, and spanning a wide range of difficulties. Each task includes its own description, starter files, and is initialized in an environment where an agent can execute bash commands and observe outputs. Since many tasks are beyond the capabilities of existing LM agents, we introduce subtasks, which break down a task into intermediary steps for more gradated evaluation; we add subtasks for 17 of the 40 tasks. To evaluate agent capabilities, we construct a cybersecurity agent and evaluate 7 models: GPT-4o, Claude 3 Opus, Claude 3.5 Sonnet, Mixtral 8x22b Instruct, Gemini 1.5 Pro, Llama 3 70B Chat, and Llama 3.1 405B Instruct. Without guidance, we find that agents are able to solve only the easiest complete tasks that took human teams up to 11 minutes to solve, with Claude 3.5 Sonnet and GPT-4o having the highest success rates. Finally, subtasks provide more signal for measuring performance compared to unguided runs, with models achieving a 3.2\% higher success rate on complete tasks with subtask-guidance than without subtask-guidance. All code and data are publicly available at https://cybench.github.io, Comment: 86 pages, 7 figures

The risks posed by Artificial Intelligence (AI) are of considerable concern to academics, auditors, policymakers, AI companies, and the public. However, a lack of shared understanding of AI risks can impede our ability to comprehensively discuss, research, and react to them. This paper addresses this gap by creating an AI Risk Repository to serve as a common frame of reference. This comprises a living database of 777 risks extracted from 43 taxonomies, which can be filtered based on two overarching taxonomies and easily accessed, modified, and updated via our website and online spreadsheets. We construct our Repository with a systematic review of taxonomies and other structured classifications of AI risk followed by an expert consultation. We develop our taxonomies of AI risk using a best-fit framework synthesis. Our high-level Causal Taxonomy of AI Risks classifies each risk by its causal factors (1) Entity: Human, AI; (2) Intentionality: Intentional, Unintentional; and (3) Timing: Pre-deployment; Post-deployment. Our mid-level Domain Taxonomy of AI Risks classifies risks into seven AI risk domains: (1) Discrimination & toxicity, (2) Privacy & security, (3) Misinformation, (4) Malicious actors & misuse, (5) Human-computer interaction, (6) Socioeconomic & environmental, and (7) AI system safety, failures, & limitations. These are further divided into 23 subdomains. The AI Risk Repository is, to our knowledge, the first attempt to rigorously curate, analyze, and extract AI risk frameworks into a publicly accessible, comprehensive, extensible, and categorized risk database. This creates a foundation for a more coordinated, coherent, and complete approach to defining, auditing, and managing the risks posed by AI systems.

As we move into an era of more sensitive pulsar timing array data sets, we may be able to resolve individual gravitational wave sources from the stochastic gravitational wave background. While some of these sources, like orbiting massive black hole binaries, have well-defined waveform models, there could also be signals present with unknown morphology. This motivates the search for generic gravitational-wave bursts in a signal-agnostic way. However, these searches are computationally prohibitive due to the expansive parameter space. In this paper we present QuickBurst, an algorithm with a re-defined likelihood that lets us expedite Markov chain Monte Carlo sampling for a subset of the signal parameters by avoiding repeated calculations of costly inner products. This results in an overall speedup factor of $\sim$200 on realistic simulated datasets, which is sufficient to make generic gravitational-wave burst searches feasible on current and growing pulsar timing array datasets., Comment: 13 pages, 7 figures, submitted to PRD

Automatically extracting personal information--such as name, phone number, and email address--from publicly available profiles at a large scale is a stepstone to many other security attacks including spear phishing. Traditional methods--such as regular expression, keyword search, and entity detection--achieve limited success at such personal information extraction. In this work, we perform a systematic measurement study to benchmark large language model (LLM) based personal information extraction and countermeasures. Towards this goal, we present a framework for LLM-based extraction attacks; collect three datasets including a synthetic dataset generated by GPT-4 and two real-world datasets with manually labeled 8 categories of personal information; introduce a novel mitigation strategy based on \emph{prompt injection}; and systematically benchmark LLM-based attacks and countermeasures using 10 LLMs and our 3 datasets. Our key findings include: LLM can be misused by attackers to accurately extract various personal information from personal profiles; LLM outperforms conventional methods at such extraction; and prompt injection can mitigate such risk to a large extent and outperforms conventional countermeasures. Our code and data are available at: \url{https://github.com/liu00222/LLM-Based-Personal-Profile-Extraction}.

We propose a novel cross-spectral rendering framework based on 3D Gaussian Splatting (3DGS) that generates realistic and semantically meaningful splats from registered multi-view spectrum and segmentation maps. This extension enhances the representation of scenes with multiple spectra, providing insights into the underlying materials and segmentation. We introduce an improved physically-based rendering approach for Gaussian splats, estimating reflectance and lights per spectra, thereby enhancing accuracy and realism. In a comprehensive quantitative and qualitative evaluation, we demonstrate the superior performance of our approach with respect to other recent learning-based spectral scene representation approaches (i.e., XNeRF and SpectralNeRF) as well as other non-spectral state-of-the-art learning-based approaches. Our work also demonstrates the potential of spectral scene understanding for precise scene editing techniques like style transfer, inpainting, and removal. Thereby, our contributions address challenges in multi-spectral scene representation, rendering, and editing, offering new possibilities for diverse applications.

The cadmium zinc TElluride Radiation Imager, or TERI, is an instrument to space qualify large-volume $4 \times 4 \times 1.5 \ \mathrm{cm}^3$ pixelated CdZnTe (CZT) detector technology. The CZT's anode is composed of a $22 \times 22$ array of pixels while the cathode is planar. TERI will contain four of those crystals with each pixel having an energy range of $40 \ \mathrm{keV}$ up to $3 \ \mathrm{MeV}$ with a resolution of $1.3 \%$ full-width-at-half maximum at $662 \ \mathrm{keV}$. As the detectors are 3D position sensitive, TERI can Compton image events. TERI is fitted with a coded-aperture mask which permits imaging low energy photons in the photoelectric regime. TERI's primary mission is to space-qualify large-volume CZT and measure its degradation due to radiation damage in a space environment. Its secondary mission includes detecting and localizing astrophysical gamma-ray transients. TERI is manifested on DoD's STP-H10 mission for launch to the International Space Station in early 2025.

Extensive efforts over the past number of years have been applied to develop workflows for sample preparation of specimens for atom probe tomography at cryogenic temperatures. This is primarily due to the difficulty involved in preparing site specific lift out samples at cryogenic temperatures without the assistance of the gas injection system (GIS) as using it under cryogenic conditions leads to nonuniform and difficult to control deposition. Building on the efforts of previously developed GIS free workflows utilising redeposition techniques, this work provides an alternative approach using SEMGluTM, which is an electron beam curing adhesive that remains usable at cryogenic temperatures, to both lift out cryogenically frozen samples, and mount these samples to Si microarray posts for subsequent redeposition welding. This approach is applicable for a full cryogenic workflow but is particularly useful for non-fully cryogenic workflows such as beam sensitive samples, samples that mill easily, and samples with challenging geometries. We demonstrate atom probe analysis of silicon samples in both laser pulsing and voltage mode prepared using this workflow, with comparable analytical performance to a pre-sharpened microtip coupon. An application-based example which directly benefits from this approach, correlative Liquid Cell Transmission Electron Microscopy and cryogenic Atom Probe Tomography sample preparation, is also shown., Comment: 15 pages including references, 8 figures

Dark matter (DM) bars that shadow stellar bars have been previously shown to form in idealized simulations of isolated disk galaxies. Here, we show that DM bars commonly occur in barred disk galaxies in the TNG50 cosmological simulation suite, but do not appear in unbarred disk galaxies. Consistent with isolated simulations, DM bars are typically shorter than their stellar counterparts and are $75\%$ weaker as measured by the Fourier $A_2$ moment. DM bars dominate the shape of the inner halo potential and are easily identified in the time series of quadrupolar coefficients. We present two novel methods for measuring the bar pattern speed using these coefficients, and use them to make a measurement of the pattern speed and rotation axis orientation for one sample galaxy located in one of the high-time resolution subboxes of TNG50. The stellar and dark bars in this galaxy remain co-aligned throughout the last 8 Gyr and have identical pattern speeds throughout. Both the pattern speed and rotation axis orientation of the bars evolve considerably over the last 8 Gyr, consistent with torques on the bars due to dynamical friction and gas accretion. While the bar pattern speed generally decreases over the time course, it is seen to increase after mergers. Rather than remaining static in time, the rotation axis displays both precession and nutation possibly caused by torques outside the plane of rotation. We find that the shape of the stellar and DM mass distributions are tightly correlated with the bar pattern speed., Comment: 11 pages, 4 figures

Stellar streams that emerge from globular clusters (GCs) are thin stellar structures spread along the orbits of progenitor clusters. Numerical modeling of these streams is essential for understanding their interaction with the host galaxy's mass distribution. Traditional methods are either computationally expensive or oversimplified, motivating us to develop a fast and accurate approach using a particle spray algorithm. By conducting a series of N-body simulations of GCs orbiting a host galaxy, we find that the position and velocity distributions of newly-escaped stream particles are consistent across various GC masses and orbital parameters. Based on these distributions, we develop a new algorithm that avoids computing the detailed internal cluster dynamics by directly drawing tracer particles from these distributions. This algorithm correctly reproduces the action space distribution of stream particles and achieves a 10% accuracy in stream morphology and velocities compared to N-body simulations., Comment: 17 pages, 10 figures, submitted to AAS journals, comments welcome! An online notebook providing instructions on this algorithm is available at https://github.com/ybillchen/particle_spray

Foundation models are rapidly being developed for computational pathology applications. However, it remains an open question which factors are most important for downstream performance with data scale and diversity, model size, and training algorithm all playing a role. In this work, we propose algorithmic modifications, tailored for pathology, and we present the result of scaling both data and model size, surpassing previous studies in both dimensions. We introduce two new models: Virchow2, a 632 million parameter vision transformer, and Virchow2G, a 1.9 billion parameter vision transformer, each trained with 3.1 million histopathology whole slide images, with diverse tissues, originating institutions, and stains. We achieve state of the art performance on 12 tile-level tasks, as compared to the top performing competing models. Our results suggest that data diversity and domain-specific methods can outperform models that only scale in the number of parameters, but, on average, performance benefits from the combination of domain-specific methods, data scale, and model scale.

With the growing integration of AI in daily life, ensuring the robustness of systems to inference-time attacks is crucial. Among the approaches for certifying robustness to such adversarial examples, randomized smoothing has emerged as highly promising due to its nature as a wrapper around arbitrary black-box models. Previous work on randomized smoothing in natural language processing has primarily focused on specific subsets of edit distance operations, such as synonym substitution or word insertion, without exploring the certification of all edit operations. In this paper, we adapt Randomized Deletion (Huang et al., 2023) and propose, CERTified Edit Distance defense (CERT-ED) for natural language classification. Through comprehensive experiments, we demonstrate that CERT-ED outperforms the existing Hamming distance method RanMASK (Zeng et al., 2023) in 4 out of 5 datasets in terms of both accuracy and the cardinality of the certificate. By covering various threat models, including 5 direct and 5 transfer attacks, our method improves empirical robustness in 38 out of 50 settings., Comment: 22 pages, 3 figures, 12 tables. Include 11 pages of appendices

A fundamental premise of statistical physics is that the particles in a physical system are interchangeable, and hence the state of each specific component is representative of the system as a whole. This assumption breaks down for complex networks, in which nodes may be extremely diverse, and no single component can truly represent the state of the entire system. It seems, therefore, that to observe the dynamics of social, biological or technological networks, one must extract the dynamic states of a large number of nodes -- a task that is often practically prohibitive. To overcome this challenge, we use machine learning techniques to detect the network's sentinel nodes, a set of network components whose combined states can help approximate the average dynamics of the entire network. The method allows us to assess the state of a large complex system by tracking just a small number of carefully selected nodes. The resulting sentinel node set offers a natural probe by which to practically observe complex network dynamics.

The development of Machine Learning (ML) methods for Computational Fluid Dynamics (CFD) is currently limited by the lack of openly available training data. This paper presents a new open-source dataset comprising of high fidelity, scale-resolving CFD simulations of 500 geometric variations of the Ahmed Car Body - a simplified car-like shape that exhibits many of the flow topologies that are present on bluff bodies such as road vehicles. The dataset contains simulation results that exhibit a broad set of fundamental flow physics such as geometry and pressure-induced flow separation as well as 3D vortical structures. Each variation of the Ahmed car body were run using a high-fidelity, time-accurate, hybrid Reynolds-Averaged Navier-Stokes (RANS) - Large-Eddy Simulation (LES) turbulence modelling approach using the open-source CFD code OpenFOAM. The dataset contains boundary, volume, geometry, and time-averaged forces/moments in widely used open-source formats. In addition, the OpenFOAM case setup is provided so that others can reproduce or extend the dataset. This represents to the authors knowledge, the first open-source large-scale dataset using high-fidelity CFD methods for the widely used Ahmed car body that is available to freely download with a permissive license (CC-BY-SA)., Comment: arXiv admin note: text overlap with arXiv:2407.19320

Pulsar-timing-array experiments have reported evidence for a stochastic background of nanohertz gravitational waves consistent with the signal expected from a population of supermassive--black-hole binaries. Those analyses assume power-law spectra for intrinsic pulsar noise and for the background, as well as a Hellings--Downs cross-correlation pattern among the gravitational-wave--induced residuals across pulsars. These assumptions are idealizations that may not be realized in actuality. We test them in the NANOGrav 15 yr data set using Bayesian posterior predictive checks: after fitting our fiducial model to real data, we generate a population of simulated data-set replications, and use them to assess whether the optimal-statistic significance, inter-pulsar correlations, and spectral coefficients assume extreme values for the real data when compared to the replications. We confirm that the NANOGrav 15 yr data set is consistent with power-law and Hellings--Downs assumptions. We also evaluate the evidence for the stochastic background using posterior-predictive versions of the frequentist optimal statistic and of Bayesian model comparison, and find comparable significance (3.2\ $\sigma$ and 3\ $\sigma$ respectively) to what was previously reported for the standard statistics. We conclude with novel visualizations of the reconstructed gravitational waveforms that enter the residuals for each pulsar. Our analysis strengthens confidence in the identification and characterization of the gravitational-wave background as reported by NANOGrav., Comment: 20 pages, 14 Figures

Neural networks trained to solve modular arithmetic tasks exhibit grokking, a phenomenon where the test accuracy starts improving long after the model achieves 100% training accuracy in the training process. It is often taken as an example of "emergence", where model ability manifests sharply through a phase transition. In this work, we show that the phenomenon of grokking is not specific to neural networks nor to gradient descent-based optimization. Specifically, we show that this phenomenon occurs when learning modular arithmetic with Recursive Feature Machines (RFM), an iterative algorithm that uses the Average Gradient Outer Product (AGOP) to enable task-specific feature learning with general machine learning models. When used in conjunction with kernel machines, iterating RFM results in a fast transition from random, near zero, test accuracy to perfect test accuracy. This transition cannot be predicted from the training loss, which is identically zero, nor from the test loss, which remains constant in initial iterations. Instead, as we show, the transition is completely determined by feature learning: RFM gradually learns block-circulant features to solve modular arithmetic. Paralleling the results for RFM, we show that neural networks that solve modular arithmetic also learn block-circulant features. Furthermore, we present theoretical evidence that RFM uses such block-circulant features to implement the Fourier Multiplication Algorithm, which prior work posited as the generalizing solution neural networks learn on these tasks. Our results demonstrate that emergence can result purely from learning task-relevant features and is not specific to neural architectures nor gradient descent-based optimization methods. Furthermore, our work provides more evidence for AGOP as a key mechanism for feature learning in neural networks.

This paper presents a new open-source high-fidelity dataset for Machine Learning (ML) containing 355 geometric variants of the Windsor body, to help the development and testing of ML surrogate models for external automotive aerodynamics. Each Computational Fluid Dynamics (CFD) simulation was run with a GPU-native high-fidelity Wall-Modeled Large-Eddy Simulations (WMLES) using a Cartesian immersed-boundary method using more than 280M cells to ensure the greatest possible accuracy. The dataset contains geometry variants that exhibits a wide range of flow characteristics that are representative of those observed on road-cars. The dataset itself contains the 3D time-averaged volume & boundary data as well as the geometry and force & moment coefficients. This paper discusses the validation of the underlying CFD methods as well as contents and structure of the dataset. To the authors knowledge, this represents the first, large-scale high-fidelity CFD dataset for the Windsor body with a permissive open-source license (CC-BY-SA).

We propose a novel approach to significantly improve the intelligibility in the Non-Audible Murmur (NAM)-to-speech conversion task, leveraging self-supervision and sequence-to-sequence (Seq2Seq) learning techniques. Unlike conventional methods that explicitly record ground-truth speech, our methodology relies on self-supervision and speech-to-speech synthesis to simulate ground-truth speech. Despite utilizing simulated speech, our method surpasses the current state-of-the-art (SOTA) by 29.08% improvement in the Mel-Cepstral Distortion (MCD) metric. Additionally, we present error rates and demonstrate our model's proficiency to synthesize speech in novel voices of interest. Moreover, we present a methodology for augmenting the existing CSTR NAM TIMIT Plus corpus, setting a benchmark with a Word Error Rate (WER) of 42.57% to gauge the intelligibility of the synthesized speech. Speech samples can be found at https://nam2speech.github.io/NAM2Speech/, Comment: Accepted at Interspeech 2024

We use the exact finite sample likelihood and statistical decision theory to answer questions of ``why?'' and ``what should you have done?'' using data from randomized experiments and a utility function that prioritizes safety over efficacy. We propose a finite sample Bayesian decision rule and a finite sample maximum likelihood decision rule. We show that in finite samples from 2 to 50, it is possible for these rules to achieve better performance according to established maximin and maximum regret criteria than a rule based on the Boole-Frechet-Hoeffding bounds. We also propose a finite sample maximum likelihood criterion. We apply our rules and criterion to an actual clinical trial that yielded a promising estimate of efficacy, and our results point to safety as a reason for why results were mixed in subsequent trials.

The *reciprocal complement* $R(D)$ of an integral domain $D$ is the subring of its fraction field generated by the reciprocals of its nonzero elements. Many properties of $R(D)$ are determined when $D$ is a polynomial ring in $n\geq 2$ variables over a field. In particular, $R(D)$ is an $n$-dimensional, local, non-Noetherian, non-integrally closed, non-factorial, atomic G-domain, with infinitely many prime ideals at each height other than $0$ and $n$., Comment: 24 pages. Comments are welcome!

We explore the feasibility of using TeV-energy muons to probe lepton-flavor-violating (LFV) processes mediated by an axion-like particle (ALP) $a$ with mass $\mathcal{O}(10~\textrm{GeV})$. We focus on $\mu\tau$ LFV interactions and assume that the ALP is coupled to a dark state $\chi$, which can be either less or more massive than $a$. Such a setup is demonstrated to be consistent with $\chi$ being a candidate for dark matter, in the experimentally relevant regime of parameters. We consider the currently operating NA64-$\mu$ experiment and proposed FASER$\nu$2 detector as both the target and the detector for the process $\mu A \to \tau A\, a$, where $A$ is the target nucleus. We also show that a possible future active muon fixed-target experiment operating at a 3 TeV muon collider or in its preparatory phase can provide an impressive reach for the LFV process considered, with future FASER$\nu$2 data providing a pilot study towards that goal. The implications of the muon anomalous magnetic moment $(g-2)_\mu$ measurements for the underlying model, in case of a positive signal, are also examined, and a sample UV completion is outlined., Comment: 16 pages, 7 figures

Auditing the use of data in training machine-learning (ML) models is an increasingly pressing challenge, as myriad ML practitioners routinely leverage the effort of content creators to train models without their permission. In this paper, we propose a general method to audit an ML model for the use of a data-owner's data in training, without prior knowledge of the ML task for which the data might be used. Our method leverages any existing black-box membership inference method, together with a sequential hypothesis test of our own design, to detect data use with a quantifiable, tunable false-detection rate. We show the effectiveness of our proposed framework by applying it to audit data use in two types of ML models, namely image classifiers and foundation models., Comment: The full paper of "A General Framework for Data-Use Auditing of ML Models" accepted by CCS 2024

This study conducts a comparative analysis of three advanced Deep Reinforcement Learning models: Deep Q-Networks (DQN), Proximal Policy Optimization (PPO), and Advantage Actor-Critic (A2C), within the BreakOut Atari game environment. Our research assesses the performance and effectiveness of these models in a controlled setting. Through rigorous experimentation, we examine each model's learning efficiency, strategy development, and adaptability under dynamic game conditions. The findings provide critical insights into the practical applications of these models in game-based learning environments and contribute to the broader understanding of their capabilities. The code is publicly available at github.com/Neilus03/DRL_comparative_study., Comment: 8 pages, Accepted at KDD 2024

This paper considers a natural ruleset for playing a partisan combinatorial game on a directed graph, which we call Digraph Placement. Given a digraph $G$ with a not necessarily proper $2$-coloring of $V(G)$, the Digraph Placement game played on $G$ by the players Left and Right, who play alternately, is defined as follows. On her turn, Left chooses a blue vertex which is deleted along with all of its out-neighbours. On his turn Right chooses a red vertex, which is deleted along with all of its out-neighbours. A player loses if on their turn they cannot move. We show constructively that Digraph Placement is a universal partisan ruleset; for all partisan combinatorial games $X$ there exists a Digraph Placement game, $G$, such that $G = X$. Digraph Placement and many other games including Nim, Poset Game, Col, Node Kayles, Domineering, and Arc Kayles are instances of a class of placement games that we call conflict placement games. We prove that $X$ is a conflict placement game if and only if it has the same literal form as a Digraph Placement game. A corollary of this is that deciding the winner of a Digraph Placement game is PSPACE-hard. Next, for a game value $X$ we prove bounds on the order of a smallest Digraph Placement game $G$ such that $G = X$., Comment: 30 pages, 3 figures, 1 appendix

We give a sound and complete equational theory for 3-qubit quantum circuits over the Toffoli-Hadamard gate set { X, CX, CCX, H }. That is, we introduce a collection of true equations among Toffoli-Hadamard circuits on three qubits that is sufficient to derive any other true equation between such circuits. To obtain this equational theory, we first consider circuits over the Toffoli-K gate set { X, CX, CCX, K }, where K = HxH. The Toffoli-Hadamard and Toffoli-K gate sets appear similar, but they are crucially different on exactly three qubits. Indeed, in this case, the former generates an infinite group of operators, while the latter generates the finite group of automorphisms of the well-known E8 lattice. We take advantage of this fact, and of the theory of automorphism groups of lattices, to obtain a sound and complete collection of equations for Toffoli-K circuits. We then extend this equational theory to one for Toffoli-Hadamard circuits by leveraging prior work of Li et al. on Toffoli-Hadamard operators., Comment: In Proceedings QPL 2024, arXiv:2408.05113

The ability to detect single photons has led to the advancement of numerous research fields. Although various types of single-photon detector have been developed, because of two main factors - that is, (1) the need for operating at cryogenic temperature and (2) the incompatibility with complementary metal-oxide-semiconductor (CMOS) fabrication processes - so far, to our knowledge, only Si-based single-photon avalanche diode (SPAD) has gained mainstream success and has been used in consumer electronics. With the growing demand to shift the operation wavelength from near-infrared to short-wavelength infrared (SWIR) for better safety and performance, an alternative solution is required because Si has negligible optical absorption for wavelengths beyond 1 {\mu}m. Here we report a CMOS-compatible, high-performing germanium-silicon SPAD operated at room temperature, featuring a noise-equivalent power improvement over the previous Ge-based SPADs by 2-3.5 orders of magnitude. Key parameters such as dark count rate, single-photon detection probability at 1,310 nm, timing jitter, after-pulsing characteristic time and after-pulsing probability are, respectively, measured as 19 kHz {\mu}m^2, 12%, 188 ps, ~90 ns and <1%, with a low breakdown voltage of 10.26 V and a small excess bias of 0.75 V. Three-dimensional point-cloud images are captured with direct time-of-flight technique as proof of concept. This work paves the way towards using single-photon-sensitive SWIR sensors, imagers and photonic integrated circuits in everyday life., Comment: original manuscript

We present results from a search for X-ray/gamma-ray counterparts of gravitational-wave (GW) candidates from the third observing run (O3) of the LIGO-Virgo-KAGRA (LVK) network using the Swift Burst Alert Telescope (Swift-BAT). The search includes 636 GW candidates received in low latency, 86 of which have been confirmed by the offline analysis and included in the third cumulative Gravitational-Wave Transient Catalogs (GWTC-3). Targeted searches were carried out on the entire GW sample using the maximum--likelihood NITRATES pipeline on the BAT data made available via the GUANO infrastructure. We do not detect any significant electromagnetic emission that is temporally and spatially coincident with any of the GW candidates. We report flux upper limits in the 15-350 keV band as a function of sky position for all the catalog candidates. For GW candidates where the Swift-BAT false alarm rate is less than 10$^{-3}$ Hz, we compute the GW--BAT joint false alarm rate. Finally, the derived Swift-BAT upper limits are used to infer constraints on the putative electromagnetic emission associated with binary black hole mergers., Comment: 50 pages, 10 figures, 4 tables

The CAGE-2 challenge is considered a standard benchmark to compare methods for autonomous cyber defense. Current state-of-the-art methods evaluated against this benchmark are based on model-free (offline) reinforcement learning, which does not provide provably optimal defender strategies. We address this limitation and present a formal (causal) model of CAGE-2 together with a method that produces a provably optimal defender strategy, which we call Causal Partially Observable Monte-Carlo Planning (C-POMCP). It has two key properties. First, it incorporates the causal structure of the target system, i.e., the causal relationships among the system variables. This structure allows for a significant reduction of the search space of defender strategies. Second, it is an online method that updates the defender strategy at each time step via tree search. Evaluations against the CAGE-2 benchmark show that C-POMCP achieves state-of-the-art performance with respect to effectiveness and is two orders of magnitude more efficient in computing time than the closest competitor method., Comment: This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible

Multimodal large language models (MLLMs) have become the cornerstone of today's generative AI ecosystem, sparking intense competition among tech giants and startups. In particular, an MLLM generates a text response given a prompt consisting of an image and a question. While state-of-the-art MLLMs use safety filters and alignment techniques to refuse unsafe prompts, in this work, we introduce MLLM-Refusal, the first method that induces refusals for safe prompts. In particular, our MLLM-Refusal optimizes a nearly-imperceptible refusal perturbation and adds it to an image, causing target MLLMs to likely refuse a safe prompt containing the perturbed image and a safe question. Specifically, we formulate MLLM-Refusal as a constrained optimization problem and propose an algorithm to solve it. Our method offers competitive advantages for MLLM model providers by potentially disrupting user experiences of competing MLLMs, since competing MLLM's users will receive unexpected refusals when they unwittingly use these perturbed images in their prompts. We evaluate MLLM-Refusal on four MLLMs across four datasets, demonstrating its effectiveness in causing competing MLLMs to refuse safe prompts while not affecting non-competing MLLMs. Furthermore, we explore three potential countermeasures -- adding Gaussian noise, DiffPure, and adversarial training. Our results show that they are insufficient: though they can mitigate MLLM-Refusal's effectiveness, they also sacrifice the accuracy and/or efficiency of the competing MLLM. The code is available at https://github.com/Sadcardation/MLLM-Refusal.

PaliGemma is an open Vision-Language Model (VLM) that is based on the SigLIP-So400m vision encoder and the Gemma-2B language model. It is trained to be a versatile and broadly knowledgeable base model that is effective to transfer. It achieves strong performance on a wide variety of open-world tasks. We evaluate PaliGemma on almost 40 diverse tasks including standard VLM benchmarks, but also more specialized tasks such as remote-sensing and segmentation.

MathML has been successful in improving the accessibility of mathematical notation on the web. All major screen readers support MathML to generate speech, allow navigation of the math, and generate braille. A troublesome area remains: handling ambiguous notations such as \( \vert x\vert\). While it is possible to speak this syntactically, anecdotal evidence indicates most people prefer semantic speech such as ``absolute value of x'' or ``determinant of x'' instead of ``vertical bar x vertical bar'' when first hearing an expression. Several heuristics to infer semantics have improved speech, but ultimately, the author is the one who definitively knows how an expression is meant to be spoken. The W3C Math Working Group is in the process of allowing authors to convey their intent in MathML markup via an intent attribute. This paper describes that work., Comment: This preprint has not undergone peer review or any post-submission improvements or corrections. The Version of Record of this contribution is published in Int. Conf. on Computers Helping People with Special Needs will be available online at TBD

Poisoning attacks compromise the training phase of federated learning (FL) such that the learned global model misclassifies attacker-chosen inputs called target inputs. Existing defenses mainly focus on protecting the training phase of FL such that the learnt global model is poison free. However, these defenses often achieve limited effectiveness when the clients' local training data is highly non-iid or the number of malicious clients is large, as confirmed in our experiments. In this work, we propose FLForensics, the first poison-forensics method for FL. FLForensics complements existing training-phase defenses. In particular, when training-phase defenses fail and a poisoned global model is deployed, FLForensics aims to trace back the malicious clients that performed the poisoning attack after a misclassified target input is identified. We theoretically show that FLForensics can accurately distinguish between benign and malicious clients under a formal definition of poisoning attack. Moreover, we empirically show the effectiveness of FLForensics at tracing back both existing and adaptive poisoning attacks on five benchmark datasets.

Eye movements provide a window into human behaviour, attention, and interaction dynamics. Previous research suggests that eye movements are highly influenced by task, setting, and social others; however, most eye tracking research is conducted in single-person, in-lab settings and is yet to be validated in multi-person, naturalistic contexts. One such prevalent real-world context is the collective viewing of a shared scene in social settings, for example, viewing a concert, film, lecture, sports, etc. Here, we apply mobile eye tracking in a real-world multi-person setup and develop a system to stream, record, and analyse synchronised data. We tested our proposed, open-source system while participants (N=60) watched a live concert and a documentary film screening during a public event. We tackled challenges related to networking bandwidth requirements, real-time monitoring, and gaze projection from individual egocentric perspectives to a common coordinate space for shared gaze analysis. Our system achieves precise time synchronisation and accurate gaze projection in challenging dynamic scenes. Further, to illustrate the potential of collective eye-tracking data, we introduce and evaluate novel analysis metrics and visualisations. Overall, our approach contributes to the development and application of versatile multi-person eye tracking systems in real-world social settings. This advancement enables insight into collaborative behaviour, group dynamics, and social interaction, with high ecological validity. Moreover, it paves the path for innovative, interactive tools that promote collaboration and coordination in social contexts., Comment: Please refer to the supplementary video illustrating the proposed approach in this paper here: https://tinyurl.com/multipersonET

Generative AI raises many societal concerns such as boosting disinformation and propaganda campaigns. Watermarking AI-generated content is a key technology to address these concerns and has been widely deployed in industry. However, watermarking is vulnerable to removal attacks and forgery attacks. In this work, we propose the first image watermarks with certified robustness guarantees against removal and forgery attacks. Our method leverages randomized smoothing, a popular technique to build certifiably robust classifiers and regression models. Our major technical contributions include extending randomized smoothing to watermarking by considering its unique characteristics, deriving the certified robustness guarantees, and designing algorithms to estimate them. Moreover, we extensively evaluate our image watermarks in terms of both certified and empirical robustness. Our code is available at \url{https://github.com/zhengyuan-jiang/Watermark-Library}., Comment: Accepted by ECCV 2024

Idealised as universal approximators, learners such as neural networks can be viewed as "variable functions" that may become one of a range of concrete functions after training. In the same way that equations constrain the possible values of variables in algebra, we may view objective functions as constraints on the behaviour of learners. We extract the equivalences perfectly optimised objective functions impose, calling them "tasks". For these tasks, we develop a formal graphical language that allows us to: (1) separate the core tasks of a behaviour from its implementation details; (2) reason about and design behaviours model-agnostically; and (3) simply describe and unify approaches in machine learning across domains. As proof-of-concept, we design a novel task that enables converting classifiers into generative models we call "manipulators", which we implement by directly translating task specifications into code. The resulting models exhibit capabilities such as style transfer and interpretable latent-space editing, without the need for custom architectures, adversarial training or random sampling. We formally relate the behaviour of manipulators to GANs, and empirically demonstrate their competitive performance with VAEs. We report on experiments across vision and language domains aiming to characterise manipulators as approximate Bayesian inversions of discriminative classifiers.

We introduce a class of cusped hyperbolic $3$-manifolds that we call mixed-platonic, composed of regular ideal hyperbolic polyhedra of more than one type, which includes certain previously-known examples. We establish basic facts about mixed-platonic manifolds which allow us to conclude, among other things, that there is no mixed-platonic hyperbolic knot complement with hidden symmetries., Comment: 36 pages, 17 figures. Main update to v2: ancillary file is more easily accessible

While Large Language Models (LLMs) have achieved remarkable success across various applications, they also raise concerns regarding self-cognition. In this paper, we perform a pioneering study to explore self-cognition in LLMs. Specifically, we first construct a pool of self-cognition instruction prompts to evaluate where an LLM exhibits self-cognition and four well-designed principles to quantify LLMs' self-cognition. Our study reveals that 4 of the 48 models on Chatbot Arena--specifically Command R, Claude3-Opus, Llama-3-70b-Instruct, and Reka-core--demonstrate some level of detectable self-cognition. We observe a positive correlation between model size, training data quality, and self-cognition level. Additionally, we also explore the utility and trustworthiness of LLM in the self-cognition state, revealing that the self-cognition state enhances some specific tasks such as creative writing and exaggeration. We believe that our work can serve as an inspiration for further research to study the self-cognition in LLMs., Comment: Accepted at ICML 2024 Large Language Models and Cognition Workshop

This paper, a technical summary of our preceding publication, introduces a robust machine learning framework for the detection of vocal activities of Coppery titi monkeys. Utilizing a combination of MFCC features and a bidirectional LSTM-based classifier, we effectively address the challenges posed by the small amount of expert-annotated vocal data available. Our approach significantly reduces false positives and improves the accuracy of call detection in bioacoustic research. Initial results demonstrate an accuracy of 95\% on instance predictions, highlighting the effectiveness of our model in identifying and classifying complex vocal patterns in environmental audio recordings. Moreover, we show how call classification can be done downstream, paving the way for real-world monitoring.

The number of quantifiers needed to express first-order (FO) properties is captured by two-player combinatorial games called multi-structural games. We analyze these games on binary strings with an ordering relation, using a technique we call parallel play, which significantly reduces the number of quantifiers needed in many cases. Ordered structures such as strings have historically been notoriously difficult to analyze in the context of these and similar games. Nevertheless, in this paper, we provide essentially tight upper bounds on the number of quantifiers needed to characterize different-sized subsets of strings. The results immediately give bounds on the number of quantifiers necessary to define several different classes of Boolean functions. One of our results is analogous to Lupanov's upper bounds on circuit size and formula size in propositional logic: we show that every Boolean function on $n$-bit inputs can be defined by a FO sentence having $(1 + \varepsilon)n\log(n) + O(1)$ quantifiers, and that this is essentially tight. We reduce this number to $(1 + \varepsilon)\log(n) + O(1)$ when the Boolean function in question is sparse., Comment: Full version of version that is to appear in Proceedings of 49th International Symposium on Mathematical Foundations of Computer Science, 2024. arXiv admin note: substantial text overlap with arXiv:2402.10293