arXiv subjects

Yuzhe You

Publications and source records attributed to Yuzhe You.

4 recordsLinked to original sources

arXiv2026

Surprise2Refine: Axis-Centered Exploration-To-Refinement for Agent-Assisted Creative Scaffolding

Designers require different design spaces across creative stages: broad during exploration, and targeted during refinement. Yet existing agent-driven tools assume a fixed or continuously expanding space, leaving designers to manage and navigate it themselves. Informed by a formative study with five designers, we propose an axis-centered workflow that adaptively broadens and narrows the design space to support structured exploration and refinement. We implemented this workflow in Surprise2Refine, a prototype that allows users to build and reshape an nxn design space through a set of axis-centered interactions as their creative intent evolves. A within-subjects study with 14 designers shows that Surprise2Refine enhances users' sense of control, supports tracking of scaffolding paths, and improves the perceived creativity of design outcomes. We further distill design insights to guide future agent-assisted tools for creative scaffolding.

cs.HC↗

arXiv2024

Panda or not Panda? Understanding Adversarial Attacks with Interactive Visualization

Adversarial machine learning (AML) studies attacks that can fool machine learning algorithms into generating incorrect outcomes as well as the defenses against worst-case attacks to strengthen model robustness. Specifically for image classification, it is challenging to understand adversarial attacks due to their use of subtle perturbations that are not human-interpretable, as well as the variability of attack impacts influenced by diverse methodologies, instance differences, and model architectures. Through a design study with AML learners and teachers, we introduce AdvEx, a multi-level interactive visualization system that comprehensively presents the properties and impacts of evasion attacks on different image classifiers for novice AML learners. We quantitatively and qualitatively assessed AdvEx in a two-part evaluation including user studies and expert interviews. Our results show that AdvEx is not only highly effective as a visualization tool for understanding AML mechanisms, but also provides an engaging and enjoyable learning experience, thus demonstrating its overall benefits for AML learners.

cs.HC↗

arXiv2024

Gamifying XAI: Enhancing AI Explainability for Non-technical Users through LLM-Powered Narrative Gamifications

Artificial intelligence (AI) has become tightly integrated into modern technology, yet existing exploratory visualizations for explainable AI (XAI) are primarily designed for users with technical expertise. This leaves everyday users, who also regularly interact with AI systems, with limited resources to explore or understand AI technologies they use. We propose a novel framework that enables non-technical users to collect insights by conversing directly with visualization elements via LLM-powered narrative gamifications. We implemented a prototype that utilizes such gamification to facilitate non-technical users' exploration of AI embedding projections. We conducted a comparative study with 10 participants to assess our prototype quantitatively and qualitatively. Our study results indicate that although our prototype effectively enhances non-technical users' AI/XAI knowledge, and users believe they learn more through the gamification feature, it remains inconclusive whether the gamification itself leads to further improvements in understanding. In addition, opinions among participants regarding the framework's engagement are mixed: some believe it enhances their exploration of the visualizations, while others feel it disrupts their workflow.

cs.HC↗

arXiv2024

CoPrompt: Supporting Prompt Sharing and Referring in Collaborative Natural Language Programming

Natural language (NL) programming has become more approachable due to the powerful code-generation capability of large language models (LLMs). This shift to using NL to program enhances collaborative programming by reducing communication barriers and context-switching among programmers from varying backgrounds. However, programmers may face challenges during prompt engineering in a collaborative setting as they need to actively keep aware of their collaborators' progress and intents. In this paper, we aim to investigate ways to assist programmers' prompt engineering in a collaborative context. We first conducted a formative study to understand the workflows and challenges of programmers when using NL for collaborative programming. Based on our findings, we implemented a prototype, CoPrompt, to support collaborative prompt engineering by providing referring, requesting, sharing, and linking mechanisms. Our user study indicates that CoPrompt assists programmers in comprehending collaborators' prompts and building on their collaborators' work, reducing repetitive updates and communication costs.

cs.HC↗