As many entities aim to participate in the ongoing AI race to gain competitive advantages, there is a risk of creating knowledge gaps by overlooking fundamental steps in the research and development processes. This paper aims to bridge the knowledge gap in the domain of large language model (LLM) integrations for knowledge elicitation by performing a systematic literature review using the PRISMA workflow. Through an analysis of 17 research papers, this study identifies new directions, including tools for education, knowledge curation, and factual information. Additionally, the research highlights key benefits and concerns associated with each direction, providing further understanding of the potential and challenges of LLM integrations for knowledge elicitation.

The swift growth of artificial intelligence has led to the development of large language models, revolutionising various scientific domains and professional fields. This research explores collaborative, cooperative, and competitive game designs, that enhance knowledge elicitation using LLMs in games with a purpose. Through a systematic literature review using the PRISMA workflow, we identify that collaborative and cooperative game designs are more effective for knowledge elicitation than competitive designs. We also address the potential and limitations of incorporating LLMs into GWAPs, such as their use as Non-Player Characters to create engaging interactions. This study provides design principles for GWAPs leveraging LLMs, offering insights for researchers and developers, and discusses ethical considerations and future research directions.

The process of knowledge elicitation is crucial to the field of artificial intelligence because of the lack of data on commonsense knowledge. This paper explores the potential of using large language models (LLM) to enhance knowledge elicitation in games with a purpose (GWAP). By analyzing the capability of LLMs to play games and generate game content, this research shows how LLMs can increase accessibility, efficiency and engagement in GWAPs for knowledge elicitation. The findings show that LLMs can play games with great performance and show much promise in generating game content that facilitates knowledge elicitation. Through a comprehensive literature survey, this research highlights potential ways of enhancing knowledge elicitation in GWAPs using LLMs and offers recommendations for future research in this field.

This research investigates the types of knowledge that can be elicited through the integration of Large Language Models (LLMs) into Games With A Purpose (GWAPs). By using a literature survey using the PRISMA framework, we synthesize findings from different studies to find patterns and gaps in the existing research. The survey focuses on the utilization of LLMs within GWAPs and examines how these models diversify the types of knowledge elicited. The findings indicate that LLMs significantly enhance the knowledge elicitation capabilities of GWAPs, transforming them into more interactive and effective tools. We found that different types of knowledge can be elicited, such as contextual insights, factual information, semantic associations and experimental knowledge.

Live-cell imaging captures dynamic cellular behaviors and aims to maximize both spatial and temporal resolution while minimizing sample damage, enabling advancements in fundamental cell biology. However, spatial resolution is limited by the diffraction barrier of optical lenses, which prevents the visualization of many subcellular structures. Single-molecule localization microscopy (SMLM) overcomes this barrier, achieving resolutions as fine as 10 nm, but it typically requires millions of frames and higher illumination, which reduces temporal resolution and can damage the sample. Super-resolution Optical Fluctuation Imaging (SOFI) operates at lower illumination levels and requires hundreds of frames, leveraging the statistical relationships of blinking fluorophores to achieve n-fold spatial resolution based on the nth-order SOFI calculation. Despite its benefits, SOFI still demands too many frames and involves extensive post-processing, making it impractical for real-time live-cell imaging. Without real-time imaging, researchers are unable to make immediate decisions, ultimately costing valuable time for the researchers.

To address this limitation, we introduce a supervised deep learning model designed to accelerate second-order SOFI. Our model reconstructs super-resolved second-order SOFI images using just 20 frames, compared to the hundreds typically required, while maintaining a 2-fold improvement in spatial resolution and showing minimal background artifacts. We demonstrate that after being trained on real fixed-cell (static) mitochondria data, the model is able to reconstruct super-resolved images in a dynamic environment by moving the microscope stage. The model achieves real-time temporal resolutions of up to 4.85 fps, unlocking new possibilities for real-time studies of live-cell dynamics.

Designing processors for implantable closed-loop neuromodulation systems presents a formidable challenge owing to the constrained operational environment, which requires low latency and high energy efficacy. Previous benchmarks have provided limited insights into power consumption and latency. However, this study introduces algorithmic metrics that capture the potential and limitations of neural decoders for closed-loop intra-cortical brain-computer interfaces in the context of energy and hardware constraints. This study benchmarks common decoding methods for predicting a primate’s finger kinematics from the motor cortex and explores their suitability for low latency and high energy efficient neural decoding. The study found that ANN-based decoders provide superior decoding accuracy, requiring high latency and many operations to effectively decode neural signals. Spiking neural networks have emerged as a solution, bridging this gap by achieving competitive decoding performance within sub-10ms while utilizing a fraction of computational resources. These distinctive advantages of neuromorphic spiking neural networks make them highly suitable for the challenging closed-loop neural modulation environment. Their capacity to balance decoding accuracy and operational efficiency offers immense potential in reshaping the landscape of neural decoders, fostering greater understanding, and opening new frontiers in closed-loop intra-cortical human-machine interaction.

Background: The advancements in Virtual Reality (VR) technology have opened up new possibilities for studying human dynamics and conducting experiments in immersive environments. To gain insights into collaborative learning and how it can be enhanced, an experiment was conducted to investigate the effects of visualizations of activities in VR on social modes of co-construction, which is to what extent learners refer to their partner's contribution.

Methods: A maze, specifically designed for collaboration, has been chosen to use for this study and visualization tools such as laser pointing and vision cones were made available for certain sessions.

Results.: The findings from this maze experiment did not provide conclusive evidence regarding the impact of visualization tools on social modes of co-construction, mainly due to the limited comparison material available. An interesting finding however, is that sessions with visualization tools tend to have more distracted participants compared to the absence of them.

Conclusion: Further research is needed to examine the relationship between visualization tools and social modes of co-construction, as well as to explore whether the observed distractions are specific to the participants in this study or representative of a bigger spectrum. By addressing these aspects, we can gain a better understanding of the role of visualization tools in collaborative learning and uncover strategies to mitigate any potential distractions they may introduce.

Background
Learning is a core part of how we grow as humans. Over the last few years it has been shown increasingly that learning in groups tends to be more effective than learning individually. This research aims to show a link between an individual’s Situational awareness and the degree of collaboration.
Method
An experiment was conducted where groups of participants navigated a maze. The maze was designed to require collaboration to reach the end. Data was gathered about both the situational awareness of the participants and their levels ofSocial Modes of Co-Construction.
Conclusion
The current data suggests no strong correlation between Situational Awareness and Levels of Social Modes of Co-Construction. However, more data will need to be gathered to increase the reliability of this conclusion.

Background
Virtual Reality is an increasingly hot topic these days. As a user, you can play more and more high-quality games and strides are being made to use this medium for education and new kinds of workspaces. Previous research on Virtual Reality shows that it could improve teamwork by letting users share information easier. Users in Virtual Reality can communicate in ways exclusive to the medium, like highlighting exact locations, so this paper analyzed how effective verbal communication is in this situation.

Methods
To analyze how Virtual Reality influences group interaction, an experiment was set up that ascertained in the presence and absence of visualizations of user actions whether a group being more aware influences how they communicate in Virtual Reality. To approach this, this paper used a maze in Virtual Reality where participants need to communicate hints of their color.

Results
The data encoding of the experiment showed that when a group's Shared Situational Awareness decreased between sessions, their level of Social Modes of Co-Construction did as well to a similar degree. When the awareness remained around the same level, only decreasing slightly, the level of social modes remained as well, going slightly up or down depending on the reading of the data.

Conclusions
From the research experiment, this paper observes there is a positive correlation between group Shared Situational Awareness and their Social Modes of Co-Construction.
Because of limited time and data sources, it is suggested that this experiment is reproduced over a longer period on a larger scale to support these findings.

Despite Virtual Reality being a relatively new field, it is steadily being introduced into numerous disciplines, such as education. Within these systems, particularly in collaborative environments, it is crucial to have a high Share Situational Awareness (SSA) in order to be aware of ones’ surroundings, encourage efficient decision making and effective team coordination. This project aims to study the change in the SSA of a group of players by allowing them (or not) to
visualize their teammates’ actions.

For this scope, an experiment was carried out: a VR game was played twice by two groups of three participants, at a distance of roughly two weeks, testing both visualizing and not visualizing their teammates’ actions. During
such event, their SSA was measured and analyzed through two techniques; namely Situation Awareness Rating Technique (SART) and Situational Awareness Linked Indicators Adapted to Novel Tasks (SALIANT).

SART and SALIANT produced opposing viewpoints. According to the former, both groups showed higher degrees of SSA in their second sessions. According to the latter, both groups’ first session had the highest level of awareness. Both of these results were unrelated to the addition of actions’ visualizations. Possible explanations include the players’ confidence rising in the second session (biasing SART), and the encoders not having an accurate inter-reliability analysis for SALIANT.

Overall, the experiment produced inconclusive results on the impact of visualizing teammates’ actions on the groups’ SSA. Nonetheless, some recommendations for future research may be made: to choose the SSA measurement tool in accordance to the layout of the system that will be examined, to extend the time between sessions, and to ensure the system is suited for the experiment.

Simulating lighting is one of the most important parts of rendering 3D scenes. While lighting coming directly from a light source is easy to simulate in real-time, indirect illumination is more difficult. One of the methods used to convincingly approximate indirect illumination is ambient occlusion, where the brightness of objects is reduced when there is geometry nearby, most visible in corners and crevices.

To render ambient occlusion in real-time often requires shortcuts that leave out some important lighting cues when geometry is not directly visible on screen. Our method aims to improve this specifically for character animation by pre-calculating the occlusion and storing it in a compact data format using cubemaps. This method has the potential to improve both visual quality and rendering performance.

Simulating visually compelling water is difficult especially in Augmented Reality environments where the water has to interact with the user’s surroundings. In this paper, implementations of reflections, refractions and transparency effects that are physically inaccurate but result in visually compelling water simulation in Augmented Reality environments are presented and tested. There are many works on simulating water efficiently; however, they focus on doing this in a scene that is fully virtual. The contribution of this paper is to extend some of these effects so that they also work in Augmented Reality environments which are non-virtual.
The effects are tested by measuring the frames per second. In the end, the methods described in the paper make the water look aesthetically pleasing at around 26 frames per second. It is important to mention that the application is locked at 30 frames per second. This means the water shader is improved with the visual effects for the cost of 4 frames. The frames were measured on an OPPO mobile phone with the model CPH2195.

Aligning sketches to their corresponding painting could give more insight into the creative process of an artist. This is a difficult task that cannot be solved directly with classical image registration techniques. Typically, features such as cracks and brushstrokes are used to match different modalities. A sketch does not contain such artifacts and is often roughly similar to the final painting. Therefore the following question rises what suitable feature detection, feature extraction/description, and transform model estimation methods can be used to align sketches to their corresponding painting. This paper provides a proof of concept by taking a manual feature detection, and a histogram of orientations as a feature description. We demonstrate that our algorithm can automatically align sketches with up to 1.4 percent of the accuracy of manual alignment.

CycleGANs [1] and CIConv [2] are both relatively new approaches to their respective applications. For CycleGANs this application is unpaired image-to-image domain adaptation and for CIConv this application is making images more
robust to illumination changes. We investigate whether CycleGANs in combination with CIConv can be used to improve the day-night domain adaptation. The resulting images could then be used during the training of CNNs that can be found in self-driving cars. Attempts were made to get the CycleGANs in combination with CIConv to train in a stable manner. These attempts included a variety of hyperparameter combinations, a number of architecture alterations and training procedure adjustments, and most significantly two different loss functions. Both these loss functions apply a CycleConsistency Loss, one applies an additional Adversarial Loss [1] and the other an additional Wasserstein Distance and Gradient Penalty [3]. In this paper we show that CycleGANs with CIConv as the first layer in either the Discriminators or the Generators resulted in unstable training. We conclude that the root of the instability issues lies in the CIConv layer causing exploding gradients resulting in unsuccessful training of the model. Finally, we propose an adjustment to the CIConv layer which shows promise in resolving these issues for the architecture with CIConv in the Generators. However, no extensive testing has been done.

While deep neural networks show great potential for being part of safety-critical applications such as autonomous driving, covering their sensitivity to illumination shifts by adding training data is of- ten non-trivial. The undesired illumination shift between train and test data can be addressed by domain adaptation methods. Recent work [9] has demonstrated performance improvements with a novel zero-shot domain adaptation setting by in- troducing a physics-based visual inductive prior - a trainable Color Invariant Convolution (CIConv) layer - aiming to transform its input to a more do- main invariant representation.
We compare the performance of image classifica- tion for day-night domain adaptation in the zero- shot and the unsupervised setting, and explore the effectiveness of using CIConv in both settings. We show that unsupervised domain adaptation reduces the day-night distribution shift similarly to CIConv in the zero-shot setting. We demonstrate improved performance when CIConv and unsupervised day- night domain adaptation are combined.

The possibility to improve an existing method by making (part of) it learnable is explored in this research. The work that this research extends added prior knowledge to a Convolutional Neural Network (CNN) to improve its performance when dealing with an illumination shift. The method used for the preprocessing, is the color invariant. The method was used in a zero-domain adaptation setting, where the network is trained without having access to the target domain. The research demonstrated improved performance, motivating further improvements.

The Color Invariant Convolution (CIConv) layer implements the color invariant edge detectors. The layer converts the RGB input of each pixel to spectral differential quotients, which are used to determine the color invariant representation. This is done through two fixed linear transformations that only approximate these values. This indicates that an even better approximation can be obtained by making this transition learnable.

Two methods are used to make this transition learnable; a linear learning method and a non-linear learning method. The linear learning method uses the original transformation but allows for change and the non-linear method replaces the linear transform with a neural network. Both methods show potential for achieving better results than the fixed transformation, but only the linear learning method actually does perform better in a classification experiment. All experiments are done following the zero-shot day to night domain adaptation on a synthetic dataset.

Moral values play a crucial role in our decision-making process by defining what is right and wrong. With the emergence of political activism and moral discourse on social media, and the latest developments in Natural Language Processing, we are looking at an opportunity to analyze moral values to observe trends as they form. Recent studies have extensively examined the performance of different NLP models for estimating moral values from text, but none of them has tackled the problem of transfer learning. Our study provides a comprehensive look into the cross-domain performance of three state-of-the-art models. We find that BERT, the current most used model in Natural Language Processing, offers the best results. For reproducibility, we publicly release our code on GitHub.

Personal moral values represent the motivation behind individuals' actions and opinions. Understanding these values is helpful both in predicting individuals' actions, such as violent protests, and building AI that can better collaborate with humans. Predicting moral values is a challenging problem due to the abstract and subjective essence of moral values. With the help of seven Twitter datasets corresponding to different domains, we train state-of-the-art Natural Language Processing models in predicting moral values. An interesting limitation of the models is that they all suffer from catastrophic forgetting. Catastrophic forgetting is the degree to which models worsen their performance on older data after being trained on new data. We conclude that catastrophic forgetting occurs irrespective of the models being trained and can be mitigated by not only training on new data but by training on a combination of old and new data. This is all possible under one assumption: old data is available. We deliver an evaluation of catastrophic forgetting for each model, explain the differences between the models, and suggest possible future work that can be built upon this research.

Moral values are instrumental in understanding people's beliefs and behaviors. Estimating such values from text would facilitate the interaction between humans and computers. To date, no comparison between NLP models for predicting moral values from text exists. This paper addresses this by comparing LSTM and more novel models such as BERT and fastText to evaluate their capabilities for predicting moral values. Twitter Corpus, a collection of 35000 Tweets containing relevant recent political and social events, is chosen for this purpose. The results show that novel solutions outperform long-established ones. BERT is proven to be the best model for this task, but long training times hinder its practicality. By contrast, fastText offers similar performance while being orders of magnitude faster.