Search

As the cultural heritage sector increasingly adopts technologies like Retrieval-Augmented Generation (RAG) to provide more personalised search experiences and enable conversations with collections data, the demand for specialised evaluation datasets has grown. While end-to-end system testing is essential, it's equally important to assess individual components. We target the final, answering task, which is well-suited to Machine Reading Comprehension (MRC). Although existing MRC datasets address general domains, they lack the specificity needed for cultural heritage information. Unfortunately, the manual creation of such datasets is prohibitively expensive for most heritage institutions. This paper presents a cost-effective approach for generating domain-specific MRC datasets with increased difficulty using Reinforcement Learning from Human Feedback (RLHF) from synthetic preference data. Our method leverages the performance of existing question-answering models on a subset of SQuAD to create a difficulty metric, assuming that more challenging questions are answered correctly less frequently. This research contributes: (1) A methodology for increasing question difficulty using PPO and synthetic data; (2) Empirical evidence of the method's effectiveness, including human evaluation; (3) An in-depth error analysis and study of emergent phenomena; and (4) An open-source codebase and set of three llama-2-chat adapters for reproducibility and adaptation., Comment: is to be published in NLP4DH 2024

This study presents scaling results and a performance analysis across different supercomputers and compilers for the Met Office weather and climate model, LFRic. The model is shown to scale to large numbers of nodes which meets the design criteria, that of exploitation of parallelism to achieve good scaling. The model is written in a Domain-Specific Language, embedded in modern Fortran and uses a Domain-Specific Compiler, PSyclone, to generate the parallel code. The performance analysis shows the effect of choice of algorithm, such as redundant computation and scaling with OpenMP threads. The analysis can be used to motivate a discussion of future work to improve the OpenMP performance of other parts of the code. Finally, an analysis of the performance tuning of the I/O server, XIOS is presented.

A major, ongoing social transition is the inclusion of autonomous agents into human organizations. For example, in defence and security applications, robots may be used alongside human operatives to reduce risk or add capability. But a key barrier to the transition to successful human-autonomous agent collectives is the need for sufficient trust between team members. A critical enabling factor for this trust will be a suitably designed dynamic allocation of function (AoF). We consider AoF in terms of a 'ladder of trust' (from low to high) with individual team members adjusting trust in their teammates based on variation in 'score' over time. The score is derived by the ability of team member to perceive and understand its situation based on the gathered information and act to acheive team or self goals. Combining these trust scores gives a system-level perspective on how AoF might be adjusted during a mission. That is, the most suitable teammate for a function might have a low trust rating from its fellow teammates, so it might be preferable to choose the next most suitable teammate for the function at that point in time. Of course, this is only in the situation where the next most suitable teammate is also likely to perform within the set framework of moral, ethical, and legal constraints. The trade-offs between trust in the individual agent's capability and predictability need to be considered within the broader context of the agent's integrity and accountability. From this perspective, the Allocation Space is defined by more than ability of each agent to perform a function.

We develop the foundations of a general framework for producing optimal upper and lower bounds on the sum $\sum_p a_p$ over primes $p$, where $(a_n)_{x/2

The nascent topic of fake news requires automatic detection methods to quickly learn from limited annotated samples. Therefore, the capacity to rapidly acquire proficiency in a new task with limited guidance, also known as few-shot learning, is critical for detecting fake news in its early stages. Existing approaches either involve fine-tuning pre-trained language models which come with a large number of parameters, or training a complex neural network from scratch with large-scale annotated datasets. This paper presents a multimodal fake news detection model which augments multimodal features using unimodal features. For this purpose, we introduce Cross-Modal Augmentation (CMA), a simple approach for enhancing few-shot multimodal fake news detection by transforming n-shot classification into a more robust (n $\times$ z)-shot problem, where z represents the number of supplementary features. The proposed CMA achieves SOTA results over three benchmark datasets, utilizing a surprisingly simple linear probing method to classify multimodal fake news with only a few training samples. Furthermore, our method is significantly more lightweight than prior approaches, particularly in terms of the number of trainable parameters and epoch times. The code is available here: \url{https://github.com/zgjiangtoby/FND_fewshot}

The energy-momentum tensor form factors are studied in $\Phi^4$ theory to one-loop order with particular focus on the $D$-term, a particle property which has attracted a lot of attention in the recent literature. It is shown that the free Klein-Gordon theory value of the $D$-term $D_{\text{free}}=-1$ is reduced to $D_{\text{one-loop}} =-\frac13$ even if the $\Phi^4$ interaction is infinitesimally small. A companion work in $\Phi^3$ theory confirms this result which may indicate that it is independent of the type of interaction as long as the scalar theory is renormalizable. Dispersion relations are studied. Various definitions of mean square radii including the mechanical radius are investigated. The findings contribute to a better understanding of the EMT properties of particles and their interpretation., Comment: 20 pages, 2 figures

We prove new bounds for how often Dirichlet polynomials can take large values. This gives improved estimates for a Dirichlet polynomial of length $N$ taking values of size close to $N^{3/4}$, which is the critical situation for several estimates in analytic number theory connected to prime numbers and the Riemann zeta function. As a consequence, we deduce a zero density estimate $N(\sigma,T)\le T^{30(1-\sigma)/13+o(1)}$ and asymptotics for primes in short intervals of length $x^{17/30+o(1)}$., Comment: 48 pages

We prove a quantitative version of the Duffin-Schaeffer conjecture with an almost sharp error term. Precisely, let $\psi:\mathbb{N}\to[0,1/2]$ be a function such that the series $\sum_{q=1}^\infty \varphi(q)\psi(q)/q$ diverges. In addition, given $\alpha\in\mathbb{R}$ and $Q\geqslant1$, let $N(\alpha;Q)$ be the number of coprime pairs $(a,q)\in\mathbb{Z}\times\mathbb{N}$ with $q\leqslant Q$ and $|\alpha-a/q|<\psi(q)/q$. Lastly, let $\Psi(Q)=\sum_{q\leqslant Q}2\varphi(q)\psi(q)/q$, which is the expected value of $N(\alpha;Q)$ when $\alpha$ is uniformly chosen from $[0, 1]$. We prove that $N(\alpha;Q)=\Psi(Q)+O_{\alpha,\varepsilon}(\Psi(Q)^{1/2+\varepsilon})$ for almost all $\alpha$ (in the Lebesgue sense) and for every fixed $\varepsilon>0$. This improves upon results of Koukoulopoulos-Maynard and of Aistleitner-Borda-Hauke., Comment: 41 pages. Final version. To appear in Duke Math. J. arXiv admin note: text overlap with arXiv:1907.04593

A study of chiral-odd generalized parton distributions (GPDs) of the nucleon is presented in the bag model demonstrating that in this model all four chiral-odd GPDs are non-zero contrary to other claims in literature. The bag model results for the GPDs $H_T^q(x,\xi,t)$, $E_T^q(x,\xi,t)$, $\tilde{H}_T^q(x,\xi,t)$ agree with other models within a typical quark model accuracy. We present one of the few quark model calculations where polynomiality is satisfied and the sum rule $\int dx\,\tilde{E}_T^q(x,\xi,t)=0$ holds. We confront our results with predictions from the large-$N_c$ limit, and with lattice QCD calculations. We conclude that the bag model successfully catches the main features of chiral-odd GPDs., Comment: 21 pages, 9 figures, 2 tables v2: minor corrections, references added

Agonism plays a vital role in democratic dialogue by fostering diverse perspectives and robust discussions. Within the realm of online conflict there is another type: hateful antagonism, which undermines constructive dialogue. Detecting conflict online is central to platform moderation and monetization. It is also vital for democratic dialogue, but only when it takes the form of agonism. To model these two types of conflict, we collected Twitter conversations related to trending controversial topics. We introduce a comprehensive annotation schema for labelling different dimensions of conflict in the conversations, such as the source of conflict, the target, and the rhetorical strategies deployed. Using this schema, we annotated approximately 4,000 conversations with multiple labels. We then trained both logistic regression and transformer-based models on the dataset, incorporating context from the conversation, including the number of participants and the structure of the interactions. Results show that contextual labels are helpful in identifying conflict and make the models robust to variations in topic. Our research contributes a conceptualization of different dimensions of conflict, a richly annotated dataset, and promising results that can contribute to content moderation., Comment: To appear

The use of abusive language online has become an increasingly pervasive problem that damages both individuals and society, with effects ranging from psychological harm right through to escalation to real-life violence and even death. Machine learning models have been developed to automatically detect abusive language, but these models can suffer from temporal bias, the phenomenon in which topics, language use or social norms change over time. This study aims to investigate the nature and impact of temporal bias in abusive language detection across various languages and explore mitigation methods. We evaluate the performance of models on abusive data sets from different time periods. Our results demonstrate that temporal bias is a significant challenge for abusive language detection, with models trained on historical data showing a significant drop in performance over time. We also present an extensive linguistic analysis of these abusive data sets from a diachronic perspective, aiming to explore the reasons for language evolution and performance decline. This study sheds light on the pervasive issue of temporal bias in abusive language detection across languages, offering crucial insights into language evolution and temporal bias mitigation.

Defining and measuring trust in dynamic, multiagent teams is important in a range of contexts, particularly in defense and security domains. Team members should be trusted to work towards agreed goals and in accordance with shared values. In this paper, our concern is with the definition of goals and values such that it is possible to define 'trust' in a way that is interpretable, and hence usable, by both humans and robots. We argue that the outcome of team activity can be considered in terms of 'goal', 'individual/team values', and 'legal principles'. We question whether alignment is possible at the level of 'individual/team values', or only at the 'goal' and 'legal principles' levels. We argue for a set of metrics to define trust in human-robot teams that are interpretable by human or robot team members, and consider an experiment that could demonstrate the notion of 'satisficing trust' over the course of a simulated mission.

Information and knowledge leakage has become a significant security risk to Australian organizations. Each security incident in Australian business cost an average US$\$$2.8 million. Furthermore, Australian organisations spend the second most worldwide (US$\$$1.2 million each on average) on investigating and assessing information breaches. The leakage of sensitive organizational information occurs through different avenues, such as social media, cloud computing and mobile devices. In this study, we (1) analyze the knowledge leakage risk (KLR) caused by the use of mobile devices in knowledge-intensive Australian organizations, (2) present a conceptual research model to explain the determinants that influence KLR through the use of mobile devices grounded in the literature, (3) conduct interviews with security and knowledge managers to understand what strategies they use to mitigate KLR caused by the use of mobile devices and (4) use content analysis and the conceptual model to frame the preliminary findings from the interviews. Keywords: Knowledge leakage, mobile devices, mobile contexts, knowledge leakage risk, Comment: Pages 14. arXiv admin note: text overlap with arXiv:1606.01450

The use of mobile devices in knowledge-intensive organizations while effective and cost-efficient also pose a challenging management problem. Often employees whether deliberately or inadvertently are the cause of knowledge leakage in organizations and the use of mobile devices further exacerbates it. This problem is the result of overly focusing on technical controls while neglecting human factors. Knowledge leakage is a multidimensional problem, and in this paper, we highlight the different dimensions that constitute it. In this study, our contributions are threefold. First, we study knowledge leakage risk (KLR) within the context of mobile devices in knowledge-intensive organizations in Australia. Second, we present a conceptual framework to explain and categorize the mitigation strategies to combat KLR through the use of mobile devices grounded in the literature. And third, we apply the framework to the findings from interviews with security and knowledge managers. Keywords: Knowledge Leakage, Knowledge Risk, Knowledge intensive, Mobile device., Comment: 22 pages, ECIS full paper 2018

The Skolem Problem asks to determine whether a given integer linear recurrence sequence has a zero term. This problem arises across a wide range of topics in computer science, including loop termination, formal languages, automata theory, and control theory, amongst many others. Decidability of the Skolem Problem is notoriously open. The state of the art is a decision procedure for recurrences of order at most 4: an advance achieved some 40 years ago, based on Baker's theorem on linear forms in logarithms of algebraic numbers. A new approach to the Skolem Problem was recently initiated via the notion of a Universal Skolem Set: a set $S$ of positive integers such that it is decidable whether a given non-degenerate linear recurrence sequence has a zero in $S$. Clearly, proving decidability of the Skolem Problem is equivalent to showing that $\mathbb{N}$ itself is a Universal Skolem Set. The main contribution of the present paper is to construct a Universal Skolem Set that has lower density at least $0.29$. We show moreover that this set has density one subject to the Bateman-Horn conjecture. The latter is a central unifying hypothesis concerning the frequency of prime numbers among the values of systems of polynomials.

Brillouin spectroscopy was used to probe the viscoelastic properties of diluted snail mucus at GHz frequencies over the range -11 $^\circ$C $\leq T \leq$ 52 $^\circ$C and of dehydrated mucus as a function of time. Two peaks were observed in the spectra for diluted mucus: the longitudinal acoustic mode of the liquid mucus peak varies with dilution but fluctuates around the typical value of 8.0 GHz. A second peak due to ice remained unchanged with varying dilution and was seen at 18.0 GHz and appeared below the dilutions "freezing" point depression. Only a single peak was found in all the dehydrated mucus spectra and was also attributed to the longitudinal acoustic mode of liquid mucus. Anomalous changes in the protein concentration dependence of the frequency shift, linewidth, and ``freezing" point depression and consequently, hypersound velocity, compressibility, and apparent viscosity suggest that the viscoelastic properties of this system is influenced by the presence of water. Furthermore, this research uncovered three unique transitions within the molecular structure. These transitions included the first stage of glycoprotein cross-linking, followed by the steady depletion of free water in the system, and eventually resulted in the creation of a gel-like state when all remaining free water was evaporated., Comment: 15 pages, 14 Figures

In 2015, Abt Associates received a grant from the Institutes for Education Sciences (IES) for a five-year "Development and Innovation" study of PTC. The purposes of the study were to gauge progress in implementing PTC and to develop and test improvements where needed. Fein et al. (2020) summarize the IES study's approach and findings. A subsequent grant from Arnold Ventures provided support for extending the two analyses--to three follow-up years for Study 1 and to four years for Study 2. This report provides findings from these longer-term analyses. Study 1 found no difference in average earnings or months enrolled in college in follow-up Years 2 and 3 between young adults assigned to PTC and their control group counterparts. (As expected, the PTC group earned less and spent more time in college than the control group in Year 1, when participants were still in the program.) The results also show modest increases in receipt of credentials (mostly short-term certificates based on credit earned at partner college during PTC).

Brillouin spectroscopy was used to probe the viscoelastic properties of a natural gastropod mucus at GHz frequencies over the range -11 $^\circ$C $\leq T \leq$ 52 $^\circ$C. Anomalies in the temperature dependence of mucus longitudinal acoustic mode peak parameters and associated viscoelastic properties at $T = -2.5^\circ$C, together with the appearance of a peak due to ice at this temperature, suggest that the mucus undergoes a phase transition from a viscous liquid state to one in which liquid mucus and solid ice phases coexist. Failure of this transition to proceed to completion even at -11 $^\circ$C is attributed to glycoprotein-water interaction. The temperature dependence of the viscoelastic properties and the phase behaviour suggest that water molecules bind to glycoprotein at a temperature above the onset of freezing and that the reduced ability of this bound water to take on a configuration that facilitates freezing is responsible for the observed freezing point depression and gradual nature of the liquid-solid transition.

This paper studies inference in predictive quantile regressions when the predictive regressor has a near-unit root. We derive asymptotic distributions for the quantile regression estimator and its heteroskedasticity and autocorrelation consistent (HAC) t-statistic in terms of functionals of Ornstein-Uhlenbeck processes. We then propose a switching-fully modified (FM) predictive test for quantile predictability. The proposed test employs an FM style correction with a Bonferroni bound for the local-to-unity parameter when the predictor has a near unit root. It switches to a standard predictive quantile regression test with a slightly conservative critical value when the largest root of the predictor lies in the stationary range. Simulations indicate that the test has a reliable size in small samples and good power. We employ this new methodology to test the ability of three commonly employed, highly persistent and endogenous lagged valuation regressors - the dividend price ratio, earnings price ratio, and book-to-market ratio - to predict the median, shoulders, and tails of the stock return distribution.

Deconvolution of cell mixtures in "bulk" transcriptomic samples from homogenate human tissue is important for understanding the pathologies of diseases. However, several experimental and computational challenges remain in developing and implementing transcriptomics-based deconvolution approaches, especially those using a single cell/nuclei RNA-seq reference atlas, which are becoming rapidly available across many tissues. Notably, deconvolution algorithms are frequently developed using samples from tissues with similar cell sizes. However, brain tissue or immune cell populations have cell types with substantially different cell sizes, total mRNA expression, and transcriptional activity. When existing deconvolution approaches are applied to these tissues, these systematic differences in cell sizes and transcriptomic activity confound accurate cell proportion estimates and instead may quantify total mRNA content. Furthermore, there is a lack of standard reference atlases and computational approaches to facilitate integrative analyses, including not only bulk and single cell/nuclei RNA-seq data, but also new data modalities from spatial -omic or imaging approaches. New multi-assay datasets need to be collected with orthogonal data types generated from the same tissue block and the same individual, to serve as a "gold standard" for evaluating new and existing deconvolution methods. Below, we discuss these key challenges and how they can be addressed with the acquisition of new datasets and approaches to analysis., Comment: 28 pages; 4 figures

The quality of electron bunches accelerated by laser wakefields is highly dependant on the temporal and spatial features of the laser driver. Analysis of experiments performed at APOLLON PW-class laser facility shows that spatial instabilities of the focal spot, such as shot-to-shot pointing fluctuations or asymmetry of the transverse fluence, lead to charge and energy degradation of the accelerated electron bunch. It is shown that PIC simulations can reproduce experimental results with a significantly higher accuracy when the measured laser asymmetries are included in the simulated laser's transverse profile, compared to simulations with ideal, symmetric laser profile. A method based on a modified Gerchberg-Saxton iterative algorithm is used to retrieve the laser electric field from fluence measurements in vacuum in the focal volume, and accurately reproduce experimental results using PIC simulations, leading to simulated electron spectra in close agreement with experimental results, for the accelerated charge, energy distribution and pointing of the electron beam at the exit of the plasma.