VR locomotion often entails a trade-off: active control enhances user agency but can divert attention from content, whereas passive travel supports focused viewing at the cost of autonomy. We investigated whether semi-active control can mediate this dichotomy by comparing active, passive, and semi-active locomotion in a user study ($N$ = 24) conducted in a structured, route-guided VR museum. Eye-tracking analyses showed that semi-active locomotion sustained exhibit-focused visual attention at a level comparable to passive travel, while subjective agency remained comparable to active control. These findings suggest that minimal, intentional input can effectively balance agency and attention in focused, predefined VR experiences.

Radiance field methods, such as Neural Radiance Field or 3D Gaussian Splatting, have emerged as seminal 3D representations for synthesizing realistic novel views. For practical applications, there is ongoing research on flexible scene editing techniques, among which object removal is a representative task. However, removing objects exposes occluded regions, often leading to unnatural appearances. Thus, studies have employed image inpainting techniques to replace such regions with plausible content - a task referred to as 3D scene inpainting. However, image inpainting methods produce one of many plausible completions for each view, leading to inconsistencies between viewpoints. A widely adopted approach leverages perceptual cues to blend inpainted views smoothly. However, it is prone to detail loss and can fail when there are perceptual inconsistencies across views. In this paper, we propose a novel 3D scene inpainting method that reliably produces realistic and perceptually consistent results even for complex scenes by leveraging a reference view. Given the inpainted reference view, we estimate the inpainting similarity of the other views to adjust their contribution in constructing an accurate geometry tailored to the reference. This geometry is then used to warp the reference inpainting to other views as pseudo-ground truth, guiding the optimization to match the reference appearance. Comparative evaluation studies have shown that our approach improves both the geometric fidelity and appearance consistency of inpainted scenes.

In in-car virtual reality (VR), ambiguous vehicle-related non-visual signals may increase reliance on top-down interpretation. We conducted a pilot study in a stationary vehicle using a headset-delivered low-frequency tone as a controllable surrogate cue, and manipulated instruction context to frame it as either aligned with VR forward motion (Aligned) or unrelated background (Not-Aligned). Phase 1 varied onset asynchrony to estimate temporal binding windows, and Phase 2 used a fixed asynchrony during repeated exposure to probe perceived coherence and motion sickness. This design offers an initial testbed for examining how contextual framing may shape multisensory inference and discomfort in in-car VR.

In-situ VR surveys offer ecological validity compared to poststimulus questionnaires. However, surveying during the ongoing task can cause interruption and bias measured variables. Thus, insitu data collection is designed to be simple and quick - typically relying on controller-based Likert-scale responses to short 2D pop-up questions. This pilot study proposes using voice communication instead, validating its effectiveness in maintaining gradual immersion without disrupting user experience. Results indicate voice interaction enabled a more stable, stress-free immersive state compared to the controller-based method.

Weight reduction in a headset is critical for improving XR usability. This poster investigates whether components related to thermal management (e.g., heat sinks) can be reduced or eliminated, in a PCVR (or thin client type) headset through display brightness and thus temperature modulation. We share findings from several tests, measuring temperature ranges (and differences) among various operating conditions, including the existence of the heat sink, especially on the display. Results have shown a potential for reduction of approximately $2-3 {\%}$ of weight for the lightest PCVR headset (about 125 g).

Managing multiple activities in virtual reality (VR) is often hindered by fragmented workflows and disruptive application switching. We present JourneyVR, a metaphorical interaction model designed for spatial and sequential workflows that reframes tasks as continuous journeys rather than isolated sessions. Users construct a Journey Map as a layout of islands (tasks) and bridges (transitions), which expands into an immersive world where activities unfold as a coherent, embodied narrative. Through a formative study, a controlled comparison, and an expert evaluation, JourneyVR was shown to enhance experiential continuity, intention to use, and overall satisfaction compared to using a conventional app launcher. Participants highlighted how the metaphor fosters motivation and achievement, while we identify boundaries regarding task type, scalability, and flexibility. Our findings demonstrate that framing sequential activities as navigable journeys can transform fragmented tasks into meaningful narratives, offering concrete guidelines for sustained engagement and more flexible workflows in immersive environments.

Emotional states influence a wide range of cognitive processes, such as attention, memory, and decision-making, and play an increasingly recognized role in human-computer interaction (HCI). Although most prior research has focused on high-level effects of emotion, the impact of incidental emotional states on fine-grained motor behaviors such as pointing and selection remains understudied. Addressing this gap, the present study examines how affective priming with validated emotional stimuli modulates user performance in a Fitts’ Law-based pointing paradigm. By systematically varying task difficulty, we quantitatively assess the effects of emotion on core performance measures: movement time, reaction time, error rate, and throughput. Our findings demonstrate that emotion can modulate low-level motor interactions in digital environments, with important implications for the design of adaptive, emotion-aware interfaces. These results advance the theoretical understanding of the interplay between emotion, cognition, and motor control and offer actionable insights to develop more robust and personalized interactive systems.

In virtual reality, weight perception can be framed through visual cues only, yet effects of different types of visual framings has not been fully explored. We experimentally investigate the individual/interactive effects of implicit (by visual surface) and explicit framing (overt depiction of deformation) and when combined in a congruent or possibly conflicting way. Preliminary results demonstrated that explicit framing dominated the implicit. Correct selection rates peaked when framings aligned congruently. These findings characterize how explicit and implicit visual framings can bias illusory relative weight judgments in VR and provide insights into how to apply them in conjunction with pseudo-haptic techniques.

Remote, vs. in situ, instruction may be regarded to decrease trainee engagement and concentration, potentially reducing training effectiveness. As such, local evaluative observers are often deployed to create the situated atmosphere. However, these observers can also have a negative effect on the trainees’ mental state and performance. This study investigates the impact of a local human observer’s presence on trainees’ mental state and task performance during military training conducted in a mixed reality (MR) environment, where a tele-presence avatar, controlled by the remote instructor, leads the training. An experiment was conducted comparing three conditions: remote training with (1) no observer, (2) a real observer, and (3) a virtual observer. The study found that although the observer, real or virtual, indeed negatively impacted the trainee’s mental state, the remote trainer avatar helped maintain the immersion/concentration, ensuring the trainees achieved the performance comparable to the no observer condition.

IntroductionOphthalmic involvement occurs in up to 40% of patients with granulomatosis with polyangiitis (GPA), usually confined to the anterior segment. Herein, we describe patients presenting with optic neuropathy as an early manifestation of GPA, without other signs of ocular or adnexa involvement.MethodsWe report a case of isolated optic neuropathy without other ocular or adnexal involvement and examine the reported clinical features of 17 additional patients through a literature review. We analyzed clinical characteristics and neuro-ophthalmological findings and discuss the clinical implications for the early detection of GPA-associated optic neuropathy.ResultsAmong the 17 patients, 10 had optic neuropathy confined to one eye, three exhibited simultaneous bilateral optic neuropathies at initial presentation, and four had unilateral involvement initially; however, the fellow eye was subsequently affected during follow-up. Nine patients had optic neuropathy as the first clinical presentation and no prior diagnosis of GPA (9/17, 53%). Among the 21 eyes (15 patients, excluding two without descriptions), disc edema was observed in five eyes (24%). Visual impairment was often profound; the measurements of 23 affected eyes at the initial presentation showed that the patient’s acuity was to count fingers or worse (14/23, 61%). The final visual outcome was often poor, with significant visual recovery in only eight eyes (8/23, 35%). Other constitutional symptoms or systemic involvements were found in most patients (15/16, 94%), mostly affecting the lung (n = 10), sinus (n = 9), and pachymeninges (n = 8). Furthermore, 88% of the patients (15/17) showed positive results on antineutrophil cytoplastic antibody. Elevated CRP (n = 6) or ESR (n = 5) was found in 56% of cases.DiscussionOur case and literature review indicates that optic neuropathy can present in the context of systemic inflammation of GPA, without any other signs of ocular or orbital involvement. Catching other clinical, imaging, and laboratory signs of systemic inflammation is important in cases of GPA-associated optic neuropathy with atypical presentations.

Experiencing VR inside autonomous vehicles can cause both VR-induced cybersickness and vehicle-induced motion sickness, which negatively affects user experience. This study investigates a mitigation approach by providing synchronized visual and haptic feedback to counteract the effects of such sickness. Specifically, visual feedback reflects the linear and rotational motion of the vehicle to the VR content using flying particles and rotating grid in the background. The haptic feedback, compensating for the vehicle's acceleration, is delivered through a tiltable seat, according to the vehicle's motion. A user study was conducted to evaluate the effectiveness of each feedback individually and their combined effect on reducing overall sickness. Additionally, the study examined the impact on user comfort and presence. The results indicated that the visual feedback enhanced the sense of movement, while haptic feedback improved comfort. Moreover, when the visual and haptic feedback were combined, a marked and synergistic decrease in sickness was observed.

Difficulty arises in virtual reality (VR) due to cybersickness. The discrepancy in motion information between the visual and vestibular feedback causes sensory conflict, a widely accepted cause of cybersickness. Previous research has shown reduced VR sickness by augmenting the content with motion patterns in the opposite direction of the virtual motion. However, it can also cause significant content intrusion. This poster aims to mitigate the latter by applying content-aware textured motion patterns vs. simple white animated feature points. Positive results were obtained with a higher preference for the proposed visualization while showing a similar degree of sickness reduction.

Concept erasure has emerged as a promising technique for mitigating the risk of harmful content generation in diffusion models by selectively unlearning undesirable concepts. The common principle of previous works to remove a specific concept is to map it to a fixed generic concept, such as a neutral concept or just an empty text prompt. In this paper, we demonstrate that this fixed-target strategy is suboptimal, as it fails to account for the impact of erasing one concept on the others. To address this limitation, we model the concept space as a graph and empirically analyze the effects of erasing one concept on the remaining concepts. Our analysis uncovers intriguing geometric properties of the concept space, where the influence of erasing a concept is confined to a local region. Building on this insight, we propose the Adaptive Guided Erasure (AGE) method, which dynamically selects optimal target concepts tailored to each undesirable concept, minimizing unintended side effects. Experimental results show that AGE significantly outperforms state-of-the-art erasure methods on preserving unrelated concepts while maintaining effective erasure performance. Our code is published at https://github.com/tuananhbui89/Adaptive-Guided-Erasure.

Multi-Agent Reinforcement Learning (MARL) is vulnerable to Adversarial Machine Learning (AML) attacks and needs adequate defences before it can be used in real world applications. We have conducted a survey into the use of execution-time AML attacks against MARL and the defences against those attacks. We surveyed related work in the application of AML in Deep Reinforcement Learning (DRL) and Multi-Agent Learning (MAL) to inform our analysis of AML for MARL. We propose a novel perspective to understand the manner of perpetrating an AML attack, by defining Attack Vectors. We develop two new frameworks to address a gap in current modelling frameworks, focusing on the means and tempo of an AML attack against MARL, and identify knowledge gaps and future avenues of research.

Cybersickness (CS) is a serious usability problem in virtual reality (VR). Postural instability theory has emerged as one major hypothesis for the cause of CS. Based on such a hypothesis, we present two experiments to observe the trends in users’ trained balance ability with respect to their susceptibilities to CS. The first experiment (as a preliminary study) evaluated the effects of 2-week balance training under three different operational conditions: training in VR (VRT), training in non-immersive media with a 2D projection display (2DT), and VR exposure without any training (VRO; Baseline). The effect toward CS was tested not only in the training space but also in a different VR content to observe for any transfer effect. As results clearly indicated that the non-VR 2DT was ineffective in gaining any significant tolerance to CS, we conducted a follow-up second experiment with 1-week balance training, focusing on comparing only the VRT and VRO conditions. Overall, the experimental findings have shown, aside from the obvious improvement in balance performance itself, that accompanying balance training had the stronger effect of increasing tolerance to CS than mere exposure to VR. Furthermore, the tolerance to CS developed through VR balance training exhibited a transfer effect, that is, with reduced levels of CS in another VR content (not used during the training sessions).

One popular form of a tele-presence guidance VR/MR system features local trainees utilizing immersive interfaces (such as headset and controllers), while the remote trainer employs the desktop interface, for the sake of familiarity, efficiency and convenience, to control the presence and actions of one’s avatar in the third person perspective. Nowadays, the desktop (or non-VR) interfaces may extend from the mere keyboard/mouse to include speech and upper body/arm/hand gestures as well. Given this, however, it is not clear whether there exist any guideline in terms of what kind of interfaces might offer the best usability for effective avatar control. This poster presents a pilot study investigating such a problem with regards to using the keyboard, speech, and hand gestures (and properly mixing them) for a MR-based military training environment where the user controlled a squad leader character for making tactical signals to following squad members. Results indicated that appropriately mixing the interfaces by the characteristics of the subtask produced the highest usability and preference, reduced world load, and even the sense of embodiment with their avatar.

In AR applications, there may be situations in which the visual target is not clearly visible/legible because it is too far or small. The unclear part of the imagery can be captured and magnified, but the image quality can still be problematic with the aliasing artifacts by the limited resolution. This poster proposes to apply deep learning based upscaling to enhance the low-resolution images. We developed a prototype system that can capture an image and upscale/present it to the user. The pilot study demonstrated that upscaled imagery improved image clarity, the ability to find hidden information more quickly, and user experience.

Various measurement tools/methods have been developed to assess cybersickness induced in virtual environments, e.g., using the controller, dial device, and verbal input. In this poster, we propose PianoFMS as a cybersickness measurement tool that allows users to directly input absolute scores using the five piano keys without tampering with visual content. The preliminary study revealed that the levels of cybersickness measured using both the dial and PianoFMS were similar, and they each exhibited a significant correlation with that of the conventional post-experiment questionnaire scores. However, the PianoFMS exhibited a markedly enhanced level of usability in comparison to the dial.

While controllers can support many generic 3D interactions in virtual reality (VR), it alone may fall short of eliciting the core experience for certain actions. One such task is the "object throwing'', ubiquitous in many sports contents, involving intricately timed actions of aiming, arm swinging, and object releasing. With the increasing availability of eye tracking, we propose to combine gaze-based targeting with the controller swing gesture to model the object throw. The target is aimed/locked by gaze, and the throw is enacted by the controller swing/button press with the object let-gone by the button release. We compare and evaluate the proposed interface against the conventional controller-only based interaction through two typical baseball tasks – Pitcher and Outfielder. The findings indicated that the task accuracy was similar, but the gaze-based targeting allowed for faster completion. More importantly, the gaze-based method showed significantly higher usability and richer VR experience.

Attacks and defences in adversarial machine learning literature have primarily focused on supervised learning. However, it remains an open question whether existing methods and strategies can be adapted to unsupervised learning approaches. In this paper we explore the challenges and strategies in attacking a $k$ -means clustering algorithm and in enhancing its robustness against adversarial manipulations. We evaluate the vulnerability of clustering algorithms to adversarial attacks on two datasets (MNIST and Fashion-MNIST), emphasising the associated security risks. Our study investigates the impact of incremental attack strength on training, introduces the concept of transferability between supervised and unsupervised models, and highlights the sensitivity of unsupervised models to sample distributions. We additionally introduce and evaluate an adversarial training method that improves testing performance in adversarial scenarios, and we highlight the importance of various parameters in the proposed training method, such as continuous learning, centroid initialisation, and adversarial step-count. Overall, our study emphasises the vulnerability of unsupervised learning and clustering algorithms to adversarial attacks and provides insights into potential defence mechanisms.