17837 research outputs found
Sort by
Towards Logically Sound Natural Language Reasoning with Logic-Enhanced Language Model Agents
Full text linkLarge language models (LLMs) are increasingly explored as general-purpose reasoners, particularly in agentic contexts. However, their outputs remain prone to mathematical and logical errors. This is especially challenging in open-ended tasks, where unstructured outputs lack explicit ground truth and may contain subtle inconsistencies. To address this issue, we propose Logic-Enhanced Language Model Agents (LELMA), a framework that integrates LLMs with formal logic to enable validation and refinement of natural language reasoning. LELMA comprises three components: an LLM-Reasoner, an LLM-Translator, and a Solver, and employs autoformalization to translate reasoning into logic representations, which are then used to assess logical validity. Using game-theoretic scenarios such as the Prisoner’s Dilemma as testbeds, we highlight the limitations of both less capable (Gemini 1.0 Pro) and advanced (GPT4o) models in generating logically sound reasoning. LELMA achieves high accuracy in error detection and improves reasoning correctness via self-refinement, particularly in GPT-4o. The study also highlights challenges in autoformalization accuracy and in evaluation of inherently ambiguous open-ended reasoning tasks
Loneliness, office space arrangement and mental well-being of Gen Z PR professionals.:Falling into the trap of an agile office?
Full text linkPurpose – The purpose of the study was to assess how the well-being and loneliness of public relations and communication professionals are impacted by the post-pandemic characteristics of the work environment: flexible work schemes, non-territorial office arrangements and video communication technologies. It was hypothesised that the post-pandemic workplace landscape poses several new challenges to the practice of PR – an industry which invariably relies on working with other people and demands a good level of social resilience. Loneliness and well-being both depend on the experience of having good and efficient social relationships, but the pandemic has directly and indirectly led to their deterioration.Design/methodology/approach – The project employed a correlational design and used an online survey system to collect responses from Gen Z professionals employed in the public relations and communications industry in the UK and the US via the Prolific platform. Demographical and workplace- related characteristics were assessed to investigate links with loneliness (measured using a three-item scale adopted from Russell et al., 1980 in Hughes, 2004) and well-being (using a short Warwick- Edinburgh Mental Well-being Scale scale). Causal relationships between data were tested using regression analysis for continuous variables and analysis of covariance for categorical factors. Bootstrapping was used to test mediated relationships that explain loneliness, job satisfaction and the well-being of Gen Z PR professionals.Findings – Several types of flexible working schemes, defined as the ability to work from home on any number of weeks, showed an impact on loneliness and job satisfaction but not on well-being. However, all remaining aspects of the post-pandemic office did manifest as important predictors. In the sample, 30% of Gen Z PR professionals showed signs of mild to clinical levels of depression, and the best protection from this state was the presence of a significant other. Lower levels of loneliness were related to non-territorial office arrangements and job satisfaction. The use of hot desks and open-plan arrangements led to a significantly lower level of job satisfaction than a traditional, cellular office. Both excessive online meetings and face-to-face only interactions led to marginally lower levels of loneliness and job satisfaction.Research limitations/implications – The present research is limited in several aspects. Firstly, while the project evaluated loneliness, job satisfaction and mental well-being (with each of these elements including a component of the requirement for building effective relationships), the quality of relationships built by PR professionals was not measured. Secondly, the project focused only on post-pandemic aspects of the workplace and did not cover other important components of job satisfaction. Lastly, the measure of online meetings was declarative rather than behavioural, and greater control of the number of online meetings held would be required to show more reliable links between variables.Practical implications – This study calls for proposing recommendations for employers to develop organisational-level measures and programmes to counteract loneliness. While traditionally intimate relationships of employees were not a direct focus of HR programmes, employers should develop elements of organisational culture that would support employees in building effective intimate relationships. Separately from this, despite immediate financial benefits, employers should avoid using open-space and hot desk policies, as they contribute negatively to job satisfaction (and indirectly to well-being). The sample of UK and US professionals was chosen for analysis because in these countries employers have more capacity to introduce changes to tangible characteristics of the workplace and work culture, which may positively impact the well-being of their employees.Social implications – It is expected that both employers and employees will revisit their approach to post- pandemic financial and logistic challenges related to the workplace. A lower level of job satisfaction and well- being is linked to the lack of assigned office space, but the ability to work exclusively from home leads to loneliness. Employees – when offered this possibility – should work in offices they are provided. Employers must appreciate the negative link between open and hot-desking policies and job satisfaction and well-being of their employees.Originality/value – This study is the first to examine the post-pandemic workplace and personal characteristics of public relations and communications professionals in the UK and US and show how they impact job satisfaction and well-being. The study shows that 30% of employed in the PR industry are at risk of depression or anxiety. The connecting factor between personal and work-related characteristics that explains this problem is loneliness
Co-creating conditions for social justice in digital societies:modes of resistance in HCI collaborative endeavors and evolving socio-technical landscapes
Full text linkIn this paper, we report on a three-year endeavour that fostered 18 collaborations between academic and non-academic organizations to co-create responses to social (in)justice issues in digital societies. The projects and range of individuals and organisations connected to this programme offer a snapshot of the state of social justice thinking within the UK digital economy research sector. Our analysis shows how the programme’s constellations of actions enacted different modes of resistance attempting to reshape people’s relationship to power dynamics, addressing institutions and exposing systems, and developing and restoring values for social justice. We explore how these efforts invite nuanced understanding of what constitutes resistance in knowledge co-production endeavours and how they helped surface tensions at the intersection of agencies and the distribution of responsibilities. Drawing from our insights and experience, we discuss implications for HCI concerned with the creation of the conditions for social justice in our digital societies
Stability of Electron and X-ray Emission in a LiTaO3 crystal-based pyroelectric accelerator driven by a periodically varying temperature
Full text linkConventional X-ray sources are bulky and require a high DC voltage. Pyroelectric X-ray generator technology has enabled us to develop portable, low-voltage X-ray sources for use in materials analysis, imaging, and other applications. The development of intense and reliable sources of charged particle beams is a current within accelerator physics in its own right. Changing the temperature of a single crystal of Lithium Tantalate (LiTaO3) in moderate vacuum conditions leads to generation of a strong electric field. If a metal target is placed nearby facing the crystal, the uncompensated polarization generated during the heating or cooling of the crystal causes the ejection of electrons from either the dielectric layer on the surface of the crystal or from the metal target depending on the polarity. These electrons are than accelerated by the strong electric field gaining an energy of up to 100 keV. The energy of these electrons can be determined by measuring the end-point energy of the X-ray spectrum that results from the interaction of the electrons with the target. It has been experimentally confirmed that a pyroelectric crystal installed in a chamber with a residual gas pressure of about 2 mTorr could be used to generate electrons with energy of up to 35 keV. Here, we present studies of the features of the electron flux in a pyroelectric accelerator and how they are affected by the pressure of the residual gas, and the distance between the crystal and the target. The connection between monoenergetic electron production in a pyroelectric generator and avalanche discharge in a gas is discussed. It is demonstrated that using a pair of crystals enables us to double the acceleration potential. Using the same setup an unknown sample was fluoresced using a pyroelectric accelerator in order to analyse its elemental content. In this report we demonstrate how pyroelectric accelerators can complement conventional X-ray tubes and radioisotopes or even large central facilities. Pyroelectric X-ray generator technology is currently being developed to provide a reliable, compact, stable, and reproducible X-ray source with controllable parameters, which does not require a high-voltage DC voltage or the use of hazardous (radioactive) materials.<br/
Cluster search optimisation of deep neural networks for audio emotion classification
Full text linkAutomated patient monitoring solutions greatly benefit from audio emotion classification, although the considerable variance in individual expression and interpretation of emotions poses a challenge. Current approaches often employ standard Audio Spectrogram Transformer (AST) and deep learning models such as Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN)-based networks. However, their performance can be enhanced by integrating neural architecture search techniques using swarm optimisation algorithms. In this research, we explore AST with hyperparameter optimisation for speech emotion recognition. Three deep learning architectures with optimisable -block structures and variable filter numbers, i.e. 1DCNN, bidirectional LSTM (BiLSTM) and CNN-BiLSTM, are also proposed, enabling the optimisation of network depth and width. A novel Cluster Search Optimisation (CSO) algorithm is introduced. It incorporates Cluster Centroid Search, a Cluster Distance Improvement metric and reinforcement learning to dispatch different search actions based on clustering convergence and -learning strategies, respectively. A novel Noise Tempered K-means (NTKM) clustering model is also proposed with the integration of Gaussian-based noise insertion and cluster compactness-separation measurement, to further fine-tune the cluster centriods obtained using OPTICS clustering. CSO is used for hyperparameter and architecture search for AST and aforementioned deep networks. Attention mechanisms are also integrated with CSO-optimised networks to further enhance feature learning. We evaluate the resulting models against those devised by other optimisation algorithms across the EMO-DB, SAVEE, and TESS datasets. The empirical results demonstrate that CSO-optimised AST and CNN-BiLSTM with attention mechanisms outperform other architectures and yield favourable comparison results against those from existing state-of-the-art audio emotion classification methods
Going Beyond the Ordinary — User Perceptions of the Impact of Multisensory Elements on Presence in Virtual Reality at the Royal Opera House
Full text linkThis exploratory study investigates the relative impacts of incorporating additional sensory- and embodiment-enhancing elements into virtual reality (VR) experiences beyond standard headset features, including vibrating floors, blowing wind, accurately rendered hands, free-roam walking and seeing avatars of real people; the outcome is sometimes called a hyper-reality experience. After taking part in the ‘Current Rising’ immersive experience at the Royal Opera House, 726 participants completed a survey examining the different perceived impacts the various additional elements were thought to have on presence. Blowing wind and free-roam walking were thought to be most impactful on presence, followed by floor vibration (contrary to expectations), along with seeing avatars. Conversely, virtual hands were thought to exhibit the least influence, despite being rendered with greater detail and precision than those commonly found in standard VR applications. Past VR experience only minimally affected these reported impacts, suggesting that hyper-reality experiences introduce novel elements even to experienced users. By looking at the perceived impact on presence over a rich, holistic range of factors (multisensory elements, virtual bodies, prior experience and enjoyment) in a real-world cultural experience, these findings offer practical guidance for immersive experience designers and researchers to optimise presence. Future research should explore more nuanced assessments of presence and consider non-correlational experimental designs that mitigate various highlighted potential biases and confounding factors
The vocabulary barrier in the General Certificate of Secondary Education (GCSE) in English Literature
Full text linkEvery year in the United Kingdom hundreds of thousands of pupils in their last year of secondary education take a General Certificate of Secondary Education (GCSE) exam in English Literature. Yet, every year, attainment is strikingly low: one quarter of those sitting the exam fail to achieve the grade 4 required for a standard pass. This paper sought to understand the reasons for this low attainment by comparing the vocabulary used in texts on the GCSE English Literature specifications with the vocabulary encountered in books that British teenagers read for pleasure. Our analysis shows that the GCSE texts have varied but dense vocabulary, and feature many words that are not encountered in popular books or in a typical spoken language environment. Many of these unfamiliar words are new roots whose meanings cannot be derived from their parts, suggesting that readers will need to rely on context or turn to a dictionary to interpret these. Together, our findings indicate that the GCSE texts will challenge even those pupils who read avidly in their free time, while their less able peers will be unable to access the texts. Our work suggests that a specification review is in order, and that it is critical that this review takes into account the wide variation in reading and language skills that pupils bring into the classroom
Illusory speeding-up and slowing-down of objects moving at constant speed emerges from natural motion detection algorithms
Full text linkThe footsteps illusion is a perceptual illusion in which two bars moving atthe same constant speed on a stripey background are seen as alternatelyaccelerating and decelerating like footsteps. The cortical mechanisms thatgive rise to footsteps and similar illusions remain to be fully understoodand may reveal important neural computations. Using an implementationof the biologically inspired correlational model of motion detection,the 2-Dimensional Motion Detector, this study had three aims. First,reproducing perceptual speed oscillations in model simulations. Second,mapping empirical reports of multiple illusion configurations onto modeloutputs. Third, inferring from the successful model, the perceptual roleof multi-scale spatio-temporal channels. We developed a 2-DimensionalMotion Detector implementation adding a global (single value) frame-by-framedynamic readout to quantify continuous and oscillating responsecomponents. We confirmed that an expected signature oscillatory motionresponse corresponded to the footsteps illusion, demonstrating that itsamplitude varied according to empirically measured illusion strength.We showed that with a global readout, the inherent pattern and contrastdependence of correlation detectors is sufficient to reproduce the surprisingperceptual illusion. This evidence suggests spacetime correlation may bea fundamental sensory computation across species, with complementaryfiltering and global pooling operations adapted for various complex phenomena
Data sources:
Access Repository Dashboard
Do you manage Royal Holloway - Pure? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!