Search

We assess the difficulty of gender resolution in literary-style dialogue settings and the influence of gender stereotypes. Instances of the test suite contain spoken dialogue interleaved with external meta-context about the characters and the manner of speaking. We find that character and manner stereotypes outside of the dialogue significantly impact the gender agreement of referents within the dialogue.

This work provides an explanatory view of how LLMs can apply moral reasoning to both criticize and defend sexist language. We assessed eight large language models, all of which demonstrated the capability to provide explanations grounded in varying moral perspectives for both critiquing and endorsing views that reflect sexist assumptions. With both human and automatic evaluation, we show that all eight models produce comprehensible and contextually relevant text, which is helpful in understanding diverse views on how sexism is perceived. Also, through analysis of moral foundations cited by LLMs in their arguments, we uncover the diverse ideological perspectives in models' outputs, with some models aligning more with progressive or conservative views on gender roles and sexism. Based on our observations, we caution against the potential misuse of LLMs to justify sexist language. We also highlight that LLMs can serve as tools for understanding the roots of sexist beliefs and designing well-informed interventions. Given this dual capacity, it is crucial to monitor LLMs and design safety mechanisms for their use in applications that involve sensitive societal topics, such as sexism., Comment: To be published at EMNLP2024

The correspondence principle bridges the quantum and classical worlds by establishing a direct link between their dynamics. This well-accepted tenant of quantum physics has been explored in quantum systems wherein the number of particles is increased to macroscopic scales. However, theoretical investigations of nanoscale structures have revealed discrepancies when attempting to bridge classical and quantum physics. Here, we report on the experimental observation of anomalous classical-to-quantum transitions in open many-body optical systems. We demonstrate, for the first time, the lack of classical-to-quantum correspondence between a macroscopic optical system and its constituent quantum multiphoton subsystems. In contrast to common belief, we demonstrate that the coherence dynamics of many-body quantum subsystems with up to forty particles can indeed be opposite to that exhibited by the hosting macroscopic system. By employing complex-Gaussian statistics, we show that these effects are universal for open many-body systems. Consequently, our work can have important implications for other fields of physics ranging from condensed matter to nuclear physics.

We present a statistical modelling framework for implementing Distributed Lag Models (DLMs), encompassing several extensions of the approach to capture the temporally distributed effect from covariates via regression. We place DLMs in the context of penalised Generalized Additive Models (GAMs) and illustrate that implementation via the R package \texttt{mgcv}, which allows for flexible and interpretable inference in addition to thorough model assessment. We show how the interpretation of penalised splines as random quantities enables approximate Bayesian inference and hierarchical structures in the same practical setting. We focus on epidemiological studies and demonstrate the approach with application to mortality data from Cyprus and Greece. For the Cyprus case study, we investigate for the first time, the joint lagged effects from both temperature and humidity on mortality risk with the unexpected result that humidity severely increases risk during cold rather than hot conditions. Another novel application is the use of the proposed framework for hierarchical pooling, to estimate district-specific covariate-lag risk on morality and the use of posterior simulation to compare risk across districts., Comment: Submitted for review in "Statistics in Medicine" on 4/January/ 2024

Large language models (LLMs) have advanced to a point that even humans have difficulty discerning whether a text was generated by another human, or by a computer. However, knowing whether a text was produced by human or artificial intelligence (AI) is important to determining its trustworthiness, and has applications in many domains including detecting fraud and academic dishonesty, as well as combating the spread of misinformation and political propaganda. The task of AI-generated text (AIGT) detection is therefore both very challenging, and highly critical. In this survey, we summarize state-of-the art approaches to AIGT detection, including watermarking, statistical and stylistic analysis, and machine learning classification. We also provide information about existing datasets for this task. Synthesizing the research findings, we aim to provide insight into the salient factors that combine to determine how "detectable" AIGT text is under different scenarios, and to make practical recommendations for future work towards this significant technical and societal challenge.

It is thought that schemes for quantum imaging are fragile against realistic environments in which the background noise is often stronger than the nonclassical signal of the imaging photons. Unfortunately, it is unfeasible to produce brighter quantum light sources to alleviate this problem. Here, we overcome this paradigmatic limitation by developing a quantum imaging scheme that relies on the use of natural sources of light. This is achieved by performing conditional detection on the photon number of the thermal light field scattered by a remote object. Specifically, the conditional measurements in our scheme enable us to extract quantum features of the detected thermal photons to produce quantum images with improved signal-to-noise ratios. This technique shows a remarkable exponential enhancement in the contrast of quantum images. Surprisingly, this measurement scheme enables the possibility of producing images from the vacuum fluctuations of the light field. This is experimentally demonstrated through the implementation of a single-pixel camera with photon-number-resolving capabilities. As such, we believe that our scheme opens a new paradigm in the field of quantum imaging. It also unveils the potential of combining natural light sources with nonclassical detection schemes for the development of robust quantum technologies.

Mitigation of gender bias in NLP has a long history tied to debiasing static word embeddings. More recently, attention has shifted to debiasing pre-trained language models. We study to what extent the simplest projective debiasing methods, developed for word embeddings, can help when applied to BERT's internal representations. Projective methods are fast to implement, use a small number of saved parameters, and make no updates to the existing model parameters. We evaluate the efficacy of the methods in reducing both intrinsic bias, as measured by BERT's next sentence prediction task, and in mitigating observed bias in a downstream setting when fine-tuned. To this end, we also provide a critical analysis of a popular gender-bias assessment test for quantifying intrinsic bias, resulting in an enhanced test set and new bias measures. We find that projective methods can be effective at both intrinsic bias and downstream bias mitigation, but that the two outcomes are not necessarily correlated. This finding serves as a warning that intrinsic bias test sets, based either on language modeling tasks or next sentence prediction, should not be the only benchmark in developing a debiased language model.

The modification of the quantum properties of coherence of photons through their interaction with matter lies at the heart of the quantum theory of light. Indeed, the absorption and emission of photons by atoms can lead to different kinds of light with characteristic quantum statistical properties. As such, different types of light are typically associated with distinct sources. Here, we report on the observation of the modification of quantum coherence of multiphoton systems in free space. This surprising effect is produced by the scattering of thermal multiphoton wavepackets upon propagation. The modification of the excitation mode of a photonic system and its associated quantum fluctuations result in the formation of different light fields with distinct quantum coherence properties. Remarkably, we show that these processes of scattering can lead to multiphoton systems with sub-shot-noise quantum properties. Our observations are validated through the nonclassical formulation of the emblematic van Cittert-Zernike theorem. We believe that the possibility of producing quantum systems with modified properties of coherence, through linear propagation, can have dramatic implications for diverse quantum technologies.

Star-product graphs are a natural extension of the Cartesian product, but have not been well-studied. We show that many important established and emerging network topologies, including HyperX, SlimFly, BundleFly, PolarStar, mesh, and torus, are in fact star-product graphs. While this connection was known for BundleFly and PolarStar, it was not for the others listed. We extend a method of constructing maximal and near-maximal sets of edge-disjoint spanning trees on Cartesian products to the star product, thus obtain maximal or near-maximal sets of edge-disjoint spanning trees on new networks of importance, where such sets can improve bandwidth of collective operations and therefore accelerate many important workloads in high-performance computing.

The quantum theory of the electromagnetic field uncovered that classical forms of light were indeed produced by distinct superpositions of nonclassical multiphoton wavepackets. Specifically, partially coherent light represents the most common kind of classical light. Here, for the first time, we demonstrate the extraction of the constituent multiphoton quantum systems of a partially coherent light field. We shift from the realm of classical optics to the domain of quantum optics via a quantum representation of partially coherent light using its complex-Gaussian statistical properties. Our formulation of the quantum Gaussian-Schell model unveils the possibility of performing photon-number-resolving detection to isolate the constituent quantum multiphoton wavepackets of a classical light field. We experimentally verified the coherence properties of isolated vacuum systems and wavepackets with up to sixteen photons. Our findings not only demonstrate the possibility of observing quantum properties of classical macroscopic objects, but also establish a fundamental bridge between the classical and quantum worlds., Comment: 11 pages, 2 figures

For $\mathbf{b}=(b_1,\dots,b_n)\in \mathbb{Z}_{>0}^n$, a $\mathbf{b}$-parking function is defined to be a sequence $(\beta_1,\dots,\beta_n)$ of positive integers whose nondecreasing rearrangement $\beta'_1\leq \beta'_2\leq \cdots \leq \beta'_n$ satisfies $\beta'_i\leq b_1+\cdots + b_i$. The $\mathbf{b}$-parking-function polytope $\mathfrak{X}_n(\mathbf{b})$ is the convex hull of all $\mathbf{b}$-parking functions of length $n$ in $\mathbb{R}^n$. Geometric properties of $\mathfrak{X}_n(\mathbf{b})$ were previously explored in the specific case where $\mathbf{b}=(a,b,b,\dots,b)$ and were shown to generalize those of the classical parking-function polytope. In this work, we study $\mathfrak{X}_n(\mathbf{b})$ in full generality. We present a minimal inequality and vertex description for $\mathfrak{X}_n(\mathbf{b})$, prove it is a generalized permutahedron, and study its $h$-polynomial. Furthermore, we investigate $\mathfrak{X}_n(\mathbf{b})$ through the perspectives of building sets and polymatroids, allowing us to identify its combinatorial types and obtain bounds on its combinatorial and circuit diameters., Comment: 27 pages, 4 figures, Comments welcomed!

Stirling permutations are parking functions, and we investigate two parking function statistics in the context of these objects: lucky cars and displacement. Among our results, we consider two extreme cases: extremely lucky Stirling permutations (those with maximally many lucky cars) and extremely unlucky Stirling permutations (those with exactly one lucky car). We show that the number of extremely lucky Stirling permutations of order $n$ is the Catalan number $C_n$, and the number of extremely unlucky Stirling permutations is $(n-1)!$. We also give some results for luck that lies between these two extremes. Further, we establish that the displacement of any Stirling permutation of order $n$ is $n^2$, and we prove several results about displacement composition vectors. We conclude with directions for further study., Comment: 17 pages, 3 tables

In this paper, we introduce a new notion of curvature on the edges of a graph that is defined in terms of effective resistances. We call this the Ricci--Foster curvature. We study the Ricci flow resulting from this curvature. We prove the existence of solutions to Ricci flow on short time intervals, and prove that Ricci flow preserves graphs with nonnegative (resp. positive) curvature., Comment: 14 pages, 8 figures, comments welcome!

Devriendt and Lambiotte recently introduced the \emph{node resistance curvature}, a notion of graph curvature based on the effective resistance matrix. In this paper, we begin the study of the behavior of the node resistance curvature under the operation of the Cartesian graph product. We study the natural question of global positivity of node resistance curvature of the Cartesian product of positively-curved graphs, and prove that, whenever $m,n\ge3$, the node resistance curvature of the interior vertices of a $m\times n$ grid is always nonpositive, while it is always nonnegative on the boundary of such grids. For completeness, we also prove a number of results on node resistance curvature in $2\times n$ grids and exhibit a counterexample to a generalization. We also give generic bounds and suggest several further questions for future study.

In this article we compare the known dynamical polynomial time algorithm for the game-over attack strategy, to that of the brute force approach; of checking all the ordered rooted subtrees of a given tree that represents a given computer network. Our approach is purely enumerative and combinatorial in nature. We first revisit known results about a doubly exponential sequence and generalize them. We then consider both finite and locally finite ordered rooted trees (LFOR-trees), and the class of their finite ordered rooted subtrees of bounded height, describing completely the LFOR-trees with no leaves where the number of ordered rooted subtrees of height at most $h$ are bounded by a polynomial in $h$. We finally consider general LFOR-trees where each level can have leaves and determine conditions for the number of ordered rooted subtrees of height at most $h$ to be bounded by a polynomial in $h$., Comment: 35 pages, 7 figures

We prove best-possible edge density conditions sufficient to guarantee traceability, Hamiltonicity, chorded pancyclicity, Hamiltonian-connectedness, $k$-path Hamiltonicity, $k$-Hamiltonicity, $k$-Hamiltonian-connectedness, and $k$-connectedness in $K_{r+1}$-free graphs. We characterize the extremal graphs. Then we extend these edge density conditions to clique density conditions. Equivalently, we introduce variants of the extremal number ex$(n,F)$ and the generalized extremal number ex$(n,H,F)$ -- that is, the maximum numbers of edges and copies of $H$, respectively, in $n$-vertex, $F$-free graphs -- in which we require that these graphs not be traceable, Hamiltonian, chorded pancyclic, Hamiltonian-connected, $k$-path Hamiltonian, $k$-Hamiltonian, $k$-Hamiltonian-connected, or $k$-connected, and we determine their values for $F=K_{r+1}$ and $H=K_t$., Comment: 49 pages

The 1$^{\text{st}}$ Workshop on Maritime Computer Vision (MaCVi) 2023 focused on maritime computer vision for Unmanned Aerial Vehicles (UAV) and Unmanned Surface Vehicle (USV), and organized several subchallenges in this domain: (i) UAV-based Maritime Object Detection, (ii) UAV-based Maritime Object Tracking, (iii) USV-based Maritime Obstacle Segmentation and (iv) USV-based Maritime Obstacle Detection. The subchallenges were based on the SeaDronesSee and MODS benchmarks. This report summarizes the main findings of the individual subchallenges and introduces a new benchmark, called SeaDronesSee Object Detection v2, which extends the previous benchmark by including more classes and footage. We provide statistical and qualitative analyses, and assess trends in the best-performing methodologies of over 130 submissions. The methods are summarized in the appendix. The datasets, evaluation code and the leaderboard are publicly available at https://seadronessee.cs.uni-tuebingen.de/macvi., Comment: MaCVi 2023 was part of WACV 2023. This report (38 pages) discusses the competition as part of MaCVi

The mechanism underlying magnetoreception has long eluded explanation. A popular hypothesis attributes this sense to the quantum coherent spin dynamics of spin-selective recombination reactions of radical pairs in the protein cryptochrome. However, concerns about the validity of the hypothesis have been raised as unavoidable inter-radical interactions, such as strong electron-electron dipolar coupling, appear to suppress its sensitivity. We demonstrate that this can be overcome by driving the spin system through a modulation of the inter-radical distance. It is shown that this dynamical process markedly enhances geomagnetic field sensitivity in strongly coupled radical pairs via a Landau-Zener type transition between singlet and triplet states. These findings suggest that a "live" harmonically driven magnetoreceptor can be more sensitive than its "dead" static counterpart., Comment: 7 pages, 4 figures, in addition to Supporting Material of 15 pages and 12 figures

A ubiquitous learning problem in today's digital market is, during repeated interactions between a seller and a buyer, how a seller can gradually learn optimal pricing decisions based on the buyer's past purchase responses. A fundamental challenge of learning in such a strategic setup is that the buyer will naturally have incentives to manipulate his responses in order to induce more favorable learning outcomes for him. To understand the limits of the seller's learning when facing such a strategic and possibly manipulative buyer, we study a natural yet powerful buyer manipulation strategy. That is, before the pricing game starts, the buyer simply commits to "imitate" a different value function by pretending to always react optimally according to this imitative value function. We fully characterize the optimal imitative value function that the buyer should imitate as well as the resultant seller revenue and buyer surplus under this optimal buyer manipulation. Our characterizations reveal many useful insights about what happens at equilibrium. For example, a seller with concave production cost will obtain essentially 0 revenue at equilibrium whereas the revenue for a seller with convex production cost is the Bregman divergence of her cost function between no production and certain production. Finally, and importantly, we show that a more powerful class of pricing schemes does not necessarily increase, in fact, may be harmful to, the seller's revenue. Our results not only lead to an effective prescriptive way for buyers to manipulate learning algorithms but also shed lights on the limits of what a seller can really achieve when pricing in the dark., Comment: Neural Information Processing Systems (NeurIPS 2021)

We observe an instance of gender-induced bias in a downstream application, despite the absence of explicit gender words in the test cases. We provide a test set, SoWinoBias, for the purpose of measuring such latent gender bias in coreference resolution systems. We evaluate the performance of current debiasing methods on the SoWinoBias test set, especially in reference to the method's design and altered embedding space properties. See https://github.com/hillarydawkins/SoWinoBias., Comment: 7+2 pages. Published in the proceedings of the 3rd Workshop on Gender Bias in Natural Language Processing (co-located with ACL-IJCNLP 2021)

Reporting and providing test sets for harmful bias in NLP applications is essential for building a robust understanding of the current problem. We present a new observation of gender bias in a downstream NLP application: marked attribute bias in natural language inference. Bias in downstream applications can stem from training data, word embeddings, or be amplified by the model in use. However, focusing on biased word embeddings is potentially the most impactful first step due to their universal nature. Here we seek to understand how the intrinsic properties of word embeddings contribute to this observed marked attribute effect, and whether current post-processing methods address the bias successfully. An investigation of the current debiasing landscape reveals two open problems: none of the current debiased embeddings mitigate the marked attribute error, and none of the intrinsic bias measures are predictive of the marked attribute effect. By noticing that a new type of intrinsic bias measure correlates meaningfully with the marked attribute effect, we propose a new postprocessing debiasing scheme for static word embeddings. The proposed method applied to existing embeddings achieves new best results on the marked attribute bias test set. See https://github.com/hillary-dawkins/MAB., Comment: 9+4 pages. Published in Findings of the ACL (ACL-IJCNLP 2021)

Diffusion source identification on networks is a problem of fundamental importance in a broad class of applications, including rumor controlling and virus identification. Though this problem has received significant recent attention, most studies have focused only on very restrictive settings and lack theoretical guarantees for more realistic networks. We introduce a statistical framework for the study of diffusion source identification and develop a confidence set inference approach inspired by hypothesis testing. Our method efficiently produces a small subset of nodes, which provably covers the source node with any pre-specified confidence level without restrictive assumptions on network structures. Moreover, we propose multiple Monte Carlo strategies for the inference procedure based on network topology and the probabilistic properties that significantly improve the scalability. To our knowledge, this is the first diffusion source identification method with a practically useful theoretical guarantee on general networks. We demonstrate our approach via extensive synthetic experiments on well-known random network models and a mobility network between cities concerning the COVID-19 spreading.

Too many young people cycle in and out of prison, jails, and detention centers and face probation and parole conditions that keep them locked out of opportunity. These interactions with the criminal justice system demand the need for both equitable practices and programs that support second chances and large-scale investments in decarceration. They also require a critical analysis and undoing of historical policies that manifest in an unequal and unjust criminal justice system. "From Surviving to Thriving: Supporting Transformation, Reentry and Connections to Employment for Young Adults," from FHI 360 and CLASP, offers practical programmatic solutions that support second chances for young people and raise policy and systems considerations to address equity, collateral consequences, and opportunity. The report features insights of best practices from nine communities that are part of the Compass Rose Collaborative (CRC). Launched in 2017, the CRC began as a three-year program funded by the U.S. Department of Labor (DOL). The CRC connects young adults ages 18-24 who have had contact with the juvenile justice or criminal justice system to employment, education pathways, and supportive services across nine communities. Policymakers and practitioners can use this analysis to increase equity and access to jobs and education for young people in need of another chance and in navigating the converging fields of workforce development, education, and criminal justice. Ultimately, the lessons shared here can offer a variety of stakeholders, including public agencies and employers, a roadmap to better understand strategies to support dismantling structural barriers and implementing strategies that support young adults on a journey of transformation and connectedness.

The characterization of errors in a quantum system is a fundamental step for two important goals. First, learning about specific sources of error is essential for optimizing experimental design and error correction methods. Second, verifying that the error is below some threshold value is required to meet the criteria of threshold theorems. We consider the case where errors are dominated by the generalized damping channel (encompassing the common intrinsic processes of amplitude damping and dephasing) but may also contain additional unknown error sources. We demonstrate the robustness of standard $T_1$ and $T_2$ estimation methods and provide expressions for the expected error in these estimates under the additional error sources. We then derive expressions that allow a comparison of the actual and expected results of fine-grained randomized benchmarking experiments based on the damping parameters. Given the results of this comparison, we provide bounds that allow robust estimation of the thresholds for fault-tolerance.

Ab initio nuclear physics tackles the problem of strongly interacting four-component fermions. The same setting could foreseeably be probed experimentally in ultracold atomic systems, where two- and three-component experiments have led to major breakthroughs in recent years. Both due to the problem's inherent interest and as a pathway to nuclear physics, in this Letter we study four-component fermions at unitarity via the use of quantum Monte Carlo methods. We explore novel forms of the trial wave function and find one which leads to a ground state of the eight-particle system whose energy is almost equal to that of two four-particle systems. We investigate the clustering properties involved and also extrapolate to the zero-range limit. In addition to being experimentally testable, our results impact the prospects of developing nuclear physics as a perturbation around the unitary limit., Comment: 6+1 pages, 3+1 figures; v2 corresponds to published version

Animals use a wide variety of strategies to reduce or avoid aggression in conflicts over resources. These strategies range from sharing resources without outward signs of conflict to the development of dominance hierarchies, in which initial fighting is followed by the submission of subordinates. Although models have been developed to analyze specific strategies for resolving conflicts over resources, little work has focused on trying to understand why particular strategies are more likely to arise in certain situations. In this paper, we use a model based on an iterated Hawk--Dove game to analyze how resource holding potentials (RHPs) and other factors affect whether sharing, dominance relationships, or other behaviours are evolutionarily stable. We find through extensive numerical simulations that sharing is stable only when the cost of fighting is low and the animals in a contest have similar RHPs, whereas dominance relationships are stable in most other situations. We also explore what happens when animals are unable to assess each other's RHPs without fighting, and we compare a range of strategies for this problem using simulations. We find (1) that the most successful strategies involve a limited period of assessment followed by a stable relationship in which fights are avoided and (2) that the duration of assessment depends both on the costliness of fighting and on the difference between the animals' RHPs. Along with our direct work on modeling and simulations, we develop extensive software to facilitate further testing; it is available at \url{https://bitbucket.org/CameronLHall/dominancesharingassessmentmatlab/}.