Medical large Vision-Language Models (Med-LVLMs) have shown strong potential in multimodal clinical applications such as medical visual question answering and report generation. However, Med-LVLMs remain challenged by hallucinations caused by modality misalignment, where models prioritize textual knowledge over visual evidence and generate outputs that conflict with medical images. To mitigate this issue, recent studies have explored preference optimization to improve image–text alignment, achieving promising results. Despite these advances, existing preference-based methods still face two limitations in medical settings: (1) overfitting to superficial cues, and (2) pseudo convergence of the preference signal. In this paper, we propose Dynamic Evidence-Guided Preference Optimization (DEPO), a new framework that enables evidence-aware and adaptive preference learning for Med-LVLMs. DEPO introduces Multi-Modal Evidence Perturbation (MEP) to suppress non-causal textual and visual shortcuts, and Dispreferred Evidence Resampling (DER) to continuously update dispreferred responses as hallucination patterns evolve. Experiments on multiple medical VQA and report generation benchmarks demonstrate consistent improvements over existing methods, with strong robustness across datasets and architectures. All Codes and data will be released after review.

Digital collage is an artistic practice that combines image cutouts to tell stories. However, preparing cutouts from a set of photos remains a tedious and time-consuming task. A formative study identified three main challenges: 1) inefficient search for relevant photos, 2) manual image cutout, and 3) difficulty in organizing large sets of cutouts. To meet these challenges and facilitate asset preparation for collage, we propose Collaposer, a tool that transforms a collection of photos into organized, ready-to-use visual cutouts based on user-provided story descriptions. Collaposer tags, detects, and segments photos, and then uses an LLM to select central and related labels based on the user-provided story description. Collaposer presents the resulting visuals in varying sizes, clustered according to semantic hierarchy. Our evaluation shows that Collaposer effectively automates the preparation process to produce diverse sets of visual cutouts adhering to the storyline, allowing users to focus on collaging these assets for storytelling.

Emergency evacuation in VR must balance realism with clear guidance. However, most prior studies strengthen either sensory or social factors in isolation, leaving equal-geometry causal estimates of load versus crowd still lacking. We present SAFE-VR, a controlled testbed that orthogonally varies Environmental Load (low vs. high) and Crowd Dynamics (orderly vs. chaotic) while keeping layout, signage, and spawn constant. In a preregistered 2 x 2 between-subjects experiment (N=80), we analyzed time-to-exit, frame-coded behavior, presence, and workload to distangle sensory from social effects. Both factors impaired egress, with the High×Chaotic condition performing worst overall. For time-to-exit, effects were additive (no reliable Load×Crowd interaction); in contrast, Temporal demand showed a crossed interaction. High load increased effort, frustration, and object contacts; while chaotic flow increased route deviations, human contacts, and slowed exits. These patterns align with reliability-weighted cueing: as guidance becomes harder to perceive, participants may shift toward crowd-following, especially when flow is unstable. SAFE-VR thus delineates how load and crowd structure jointly shape route fidelity, collisions, and evacuation time, and highlights conditions where subjective time pressure diverges from objective delay.

Surfacing peers' workload and speed can reshape collaboration in VR, yet prior work often conflates social cues with environmental load. We isolate the social channel with TRACE-VR, which holds geometry, physics, rules, and timing constant while independently manipulating effort identifiability (traceable vs. anonymous) and peer effort (pace) (high vs. low). In a 2×2 within-subjects study $(n=32)$, each participant worked with nine scripted co-actors for three minutes, transporting 16 crates along a self-chosen path between fixed pickup/drop points; co-actors followed fixed pre-authored routes. We measured intrinsic motivation, social-evaluative load, NASA-TLX, and behavioral/process outcomes. Our findings show that identifiability and higher peer effort (pace) each modestly increased completion rates, with asymmetric effects suggesting partial substitution between accountability cues and normative pace. Perceived monitoring and temporal demand depended on their combination-identifiability raised monitoring at low pace and amplified time pressure at high pace-while composite TLX and finishers' times remained comparable across conditions. Thus, these cues chiefly determine who finishes rather than how fast finishers move. Motivation did not uniformly increase under higher pace. We frame identifiability and pace as partially substitutable levers for shaping social-evaluative experience and facet-level motivation, and outline tempo-aware, accountability-aware guidance for collaborative VR.

LED-VP is now mainstream in high-end studios, yet the capital cost of real-time volume stages renders training with this technology prohibitively expensive. Most film institutions simply cannot afford to build one LED-VP set for their students. This paper explores an alternative solution to the training of LED-VP by developing a virtual counterpart. However, significant challenges arise in simulating a virtual environment where students can operate in an LED-VP. Our study proposes the virtual environment of LED-VP, a virtual reality Collaborative Learning (VRCL) system. We developed an interface featuring visual symbols, namely MetaCineMoji , for film operation to facilitate smooth communication and learning processes in LED-VP workflows. MetaCineMoji demonstrates the feasibility of multi-person collaborative learning in a virtual filming studio, translating film set communication to visual symbols for lighting design, scene construction, and collaborative work with key stakeholders, e.g., directors, cinematographers, and gaffers. We explore the impact of the film-operation interfaces containing visual symbols on social interaction factors within the virtual studio. We conducted evaluations with 24 participants. Our findings show that our system equipped with the film-operation interface, enabled by visual symbols, significantly enhances social interaction among learners and results in significantly higher learning outcomes than systems without the interface. • MetaCineMoji VR system enables collaborative LED production training, significantly enhancing team communication and learning outcomes. • Visual symbols in MetaCineMoji streamline film set operations, facilitating efficient role-based interactions among directors, cinematographers, and gaffers. • Participants using MetaCineMoji reported improved workflow understanding, and strengthened team presence compared to non-symbolic systems.

Recent research within the HCI community has illuminated the challenges faced by marginalized groups on algorithm-driven livestreaming platforms. However, there is a notable gap in understanding how elderly livestreamers interact with the platform content moderation and algorithmic (in)visibility. This study investigates the perceptions of the algorithm-moderated (in)visibility and the coping strategies of 16 elderly streamers on Douyin. We find that, contrary to stereotypes of elderly users as digitally uninformed, these streamers actively engage with the platform to facilitate their understanding about platform algorithm. This engagement involves official guidance, peer learning, and personal experimentation. The streamers adopt various strategies to align with the perceived algorithmic preferences. Despite their rich knowledge about the platform's visibility moderation, many elderly streamers face significant challenges, such as physical and psychological strain and low viewer traffic. We conclude with design implications for livestreaming platforms to foster fairness and promote engagement among elderly streamers.

Detecting hidden stigmatization in social media poses significant challenges due to semantic misalignments between textual and visual modalities, as well as the subtlety of implicit stigmatization. Traditional approaches often fail to capture these complexities in real-world, multimodal content. To address this gap, we introduce ContextAware, an agent-based framework that leverages specialized modules to collaboratively process and analyze images, textual context, and social interactions. Our approach begins by clustering image embeddings to identify recurring content, activating high-likes agents for deeper analysis of images receiving substantial user engagement, while comprehensive agents handle lower-engagement images. By integrating case-based learning, textual sentiment, and vision-language models (VLMs), ContextAware refines its detection of harmful content. We evaluate ContextAware on a self-collected Douyin dataset focused on interracial relationships, comprising 871 short videos and 885,502 comments—of which a notable portion are image-based. Experimental results show that ContextAware not only outperforms state-of-the-art methods in accuracy and F1 score but also effectively detects implicit stigmatization within the highly contextual environment of social media. Our findings underscore the importance of agent-based architectures and multimodal alignment in capturing nuanced, culturally specific forms of harmful content.

In virtual reality (VR) education, especially in creative fields like film production, avatar design and narrative style extend beyond appearance and aesthetics. This study explores how the interaction between avatar gender, the dominant narrative actor's gender, and the learner's gender influences film production learning in VR, focusing on gaze dynamics and gender perspectives. Using a$2 \times 2 \times 2$experimental design, 48 participants operated avatars of different genders and interacted with male or female-dominant narratives. The results show that the consistency between the avatar and gender affects presence, and learners' control over the avatar is also influenced by gender matching. Learners using avatars of the opposite gender reported stronger control, suggesting gender incongruity prompted more focus on the avatar. Additionally, female participants with female avatars were more likely to adopt a “female gaze,” favoring soft lighting and emotional shots, while male participants with male avatars were more likely to adopt a “male gaze,” choosing dynamic shots and high contrast. When male participants used female avatars, they favored “female gaze,” while female participants with male avatars focused on “male gaze”. These findings advance our understanding of how avatar design and narrative style in VRbased education influence creativity and the cultivation of gender perspectives, and they offer insights for developing more inclusive and diverse VR teaching tools going forward.

Impulsive buying tendencies exist on Douyin, the most popular Chinese social media platform, primarily due to the users’ exposure to live sales events. This study delved into examining impulsive buying behaviour, specifically triggered by external stimuli, through the lens of the Stimulus-Organism-Response framework model. Thus, our study implemented a tailored Douyin client, namely Douyin X, that contains four interventions: Visualizing the wallet’s balances, enhancing payment friction, Prolonging the duration of purchase decision-making, and imposing browsing time limits and usage statistics. Our user study with 20 participants implies individuals’ impulsive buying due to external stimuli, combined with the proliferation of impulsive buying-promoting designs, which has led to the excessive prevalence of impulsive buying in live e-commerce. Our research offers a comprehensive framework that can effectively mitigate the likelihood of impulsive buying behaviours, specifically from a design-oriented standpoint.

Livestream shopping platforms often overlook the accessibility needs of the Deaf and Hard of Hearing (DHH) community, leading to barriers such as information inaccessibility and overload. To tackle these challenges, we developed EchoAid, a mobile app designed to improve the livestream shopping experience for DHH users. EchoAid utilizes advanced speech-to-text conversion, Rapid Serial Visual Presentation (RSVP) technology, and Large Language Models (LLMs) to simplify the complex information flow in live sales environments. We conducted exploratory studies with eight DHH individuals to identify design needs and iteratively developed the EchoAid prototype based on feedback from three participants. We then evaluate the performance of this system in a user study workshop involving 38 DHH participants. Our findings demonstrate the successful design and validation process of EchoAid, highlighting its potential to enhance product information extraction, leading to reduced cognitive overload and more engaging and customized shopping experiences for DHH users.

Despite the growing popularity of data videos, guidance on designing narrative climaxes that maximise viewers' emotional engagement remains scarce. To address this gap, our work leverages emotional theory to derive patterns for crafting emotionally resonant climaxes of data videos. We first analyzed the climaxes of 96 data videos, categorizing them into eight emotional dimensions based on Plutchik's basic emotion model. Based on data analysis, we then formulated 40 patterns for creating narrative climaxes. To evaluate the patterns when applied as design hints, we conducted a user study with 48 participants, where Group A created data video climaxes using our patterns, Group B created them without our patterns, and Group C used other patterns as the baseline. Evaluations by two experts and 20 general audiences revealed that the climaxes created with the patterns were more emotionally engaging. The participants also praised the clarity and practicality of the patterns.

Data videos increasingly becoming a popular data storytelling form represented by visual and audio integration. In recent years, more and more researchers have explored many narrative structures for effective and attractive data storytelling. Meanwhile, the Hero's Journey provides a classic narrative framework specific to the Hero's story that has been adopted by various mediums. There are continuous discussions about applying Hero's Journey to data stories. However, so far, little systematic and practical guidance on how to create a data video for a specific story type like the Hero's Journey, as well as how to manipulate its sound and visual designs simultaneously. To fulfill this gap, we first identified 48 data videos aligned with the Hero's Journey as the common storytelling from 109 high-quality data videos. Then, we examined how existing practices apply Hero's Journey for creating data videos. We coded the 48 data videos in terms of the narrative stages, sound design, and visual design according to the Hero's Journey structure. Based on our findings, we proposed a design space to provide practical guidance on the narrative, visual, and sound custom design for different narrative segments of the hero's journey (i.e., Departure, Initiation, Return) through data video creation. To validate our proposed design space, we conducted a user study where 20 participants were invited to design data videos with and without our design space guidance, which was evaluated by two experts. Results show that our design space provides useful and practical guidance for data storytellers effectively creating data videos with the Hero's Journey.

In virtual reality (VR) education, particularly in creative fields like film production, the role of different virtual environments in shaping learning outcomes remains underexplored. This study investigates how three distinct environments-baseline, a dynamic beach setting, and a familiar office space-affect students' ability to learn film lighting techniques and whether team-based learning offers advantages over individual learning. We conducted a 3×2 factorial experiment with 36 participants to examine the effects of these environments on learning performance. Our results show for individual learners, the dynamic and potentially distracting beach environment increased frustration and effort but also heightened their sense of engagement and perceived performance. In contrast, team-based learning in familiar environments like the office significantly reduced frustration and fostered collaboration, leading to improved performance. Interestingly, team-based learning excelled in the baseline environment, whereas individual learners performed better in more challenging settings like the beach. These findings provide practical insights into optimizing virtual environments to enhance both individual and collaborative learning in VR education.

Lighting education is a foundational component of cinematography education. However, many art schools do not have expensive soundstages for traditional cinematography lessons. Migrating physical setups to virtual experiences is a potential solution driven by metaverse initiatives. Yet there is still a lack of knowledge on the design of a VR system for teaching cinematography. We first analyzed the educational needs for cinematography lighting education by conducting interviews with six cinematography professionals from academia and industry. Accordingly, we presented Art Mirror, a VR soundstage for teachers and students to emulate cinematography lighting in virtual scenarios. We evaluated Art Mirror from the aspects of usability, realism, presence, sense of agency, and collaboration. Sixteen participants were invited to take a cinematography lighting course and assess the design elements of Art Mirror. Our results demonstrate that Art Mirror is usable and useful for cinematography lighting education, which sheds light on the design of VR cinematography education.

In film education, high expenses and limited space significantly challenge teaching synchronized sound recording (SSR). Traditional methods, which emphasize theory with limited practical experience, often fail to bridge the gap between theoretical understanding and practical application. As such, we introduce MetaEcho, an educational virtual reality leveraging the presence theory for teaching SSR. MetaEcho provides realistic simulations of various recording equipment and facilitates communication between learners and instructors, offering an immersive learning experience that closely mirrors actual practices. An evaluation with 24 students demonstrated that MetaEcho surpasses the traditional method in presence, collaboration, usability, realism, comprehensibility, and creativity. Three experts also commented on the benefits of MetaEcho and the opportunities for promoting SSR education in the metaverse era.

We examine the social media discourse surrounding interracial relationships in China, specifically on the popular platform Douyin. By analyzing comments on short video posts, the study focuses on four types of interracial relationships: Black men and Chinese women, Black women and Chinese men, White men and Chinese women, and White women and Chinese men. The study also explores potential regional differences in these discourses, using IP geolocation data made available to the public since April 2022. Our content analysis revealed that the Black men and Chinese women couples attracted the most negative comments and the White women and Chinese men couples received the least negative comments. We also observed substantial regional differences in the discourses towards these interracial relationships. We investigated several regional socioeconomic development indicators and noted that local GDP, population sizes, and the level of openness to Western cultures explained the variation in the negative sentiment level. This work advances our understanding of the interplay of race, gender, and immigration in constructing public discourses on social media and offers important insights into how these discourses evolve along with socioeconomic development.

Data videos are becoming increasingly popular in society and academia. Yet little is known about how to create endings that strengthen a lasting impression and persuasion. To fulfill the gap, this work aims to develop guidelines for data video endings by drawing inspiration from cinematic arts. To contextualize cinematic endings in data videos, 111 film endings and 105 data video endings are first analyzed to identify four common styles using the framework of ending punctuation marks. We conducted expert interviews (N=11) and formulated 20 guidelines for creating cinematic endings in data videos. To validate our guidelines, we conducted a user study where 24 participants were invited to design endings with and without our guidelines, which are evaluated by experts and the general public. The participants praise the clarity and usability of the guidelines, and results show that the endings with guidelines are perceived to be more understandable, impressive, and reflective.

The high cost and limited availability of soundstages for cinematography lighting education pose significant challenges for art institutions. Traditional teaching methods, combining basic lighting equipment operation with slide lectures, often yield unsatisfactory results, hindering students' mastery of cinematography lighting techniques. Therefore, we propose Metashadow, a virtual reality (VR) cinematography lighting education system demonstrating the feasibility of learning in a virtual soundstage. Based on the presence theory, Metashadow features high-fidelity lighting devices that enable users to adjust multiple parameters, providing a quantifiable learning approach. We evaluated Metashadow with 24 participants and found that it provides better learning outcomes than traditional teaching methods regarding presence, collaboration, usability, realism, creativity, and flexibility. Six experts also praised the Metashadow's expressiveness and its learning outcomes. Our study demonstrates the potential of VR technology to enhance cinematography lighting education while imposing a smaller cost burden and space requirement.

Data storytelling explores how to communicate data insights to the general public engagingly and effectively. It combines the power of data visualizations and storytelling techniques and is popular in various media such as newspapers, interactive websites, and videos. Recently, virtual reality has brought new opportunities to enhance data storytelling with an incomparable sense of immersion. However, there exists a limited understanding of data stories in virtual reality (VR) as they are still in the early stage. In this paper, we investigated the idea of VR data videos by drawing inspiration from popular 3D data videos and studying how to transfer them from screens to VR. We systematically analyzed 100 highly-watched 3D data videos from Youtube and Tiktok channels to derive their design space. We then conducted a user study with 12 participants to explore the effects of four design factors on user experience, including varying camera angles, showing chart overview, animation, and using anchors. Specifically, participants watched 3D data videos in desktop and VR environments. We collected and analyzed their quantitative and qualitative feedback regarding the story’s understandability, memorability, engagement, and emotional effects. Results suggested that data videos in VR were significantly more appreciated than on desktops. We concluded with design implications for future applications and research on VR data videos.

Data videos are an increasingly popular storytelling form. The opening of a data video critically influences its success as the opening either attracts the audience to continue watching or bores them to abandon watching. However, little is known about how to create an attractive opening. We draw inspiration from the openings of famous films to facilitate designing data video openings. First, by analyzing over 200 films from several sources, we derived six primary cinematic opening styles adaptable to data videos. Then, we consulted eight experts from the film industry to formulate 28 guidelines. To validate the usability and effectiveness of the guidelines, we asked participants to create data video openings with and without the guidelines, which were then evaluated by experts and the general public. Results showed that the openings designed with the guidelines were perceived to be more attractive, and the guidelines were praised for clarity and inspiration.