arXiv subjects

Heesoo Kim

Publications and source records attributed to Heesoo Kim.

4 recordsLinked to original sources

arXiv2025

Measuring Data Quality for Project Lighthouse

In this paper, we first situate the challenges for measuring data quality under Project Lighthouse in the broader academic context. We then discuss in detail the three core data quality metrics we use for measurement--two of which extend prior academic work. Using those data quality metrics as examples, we propose a framework, based on machine learning classification, for empirically justifying the choice of data quality metrics and their associated minimum thresholds. Finally we outline how these methods enable us to rigorously meet the principle of data minimization when analyzing potential experience gaps under Project Lighthouse, which we term quantitative data minimization.

stat.AP↗

arXiv2025

Core Mondrian: Basic Mondrian beyond k-anonymity

We present Core Mondrian, a scalable extension of the Original Mondrian partition-based anonymization algorithm. A modular strategy layer supports k-anonymity, allowing new privacy models to be added easily. A hybrid recursive/queue execution engine exploits multi-core parallelism while maintaining deterministic output. Utility-preserving enhancements include NaN-pattern pre-partitioning, metric-driven cut scoring, and dynamic suppression budget management. Experiments on the 48k-record UCI ADULT dataset and synthetically scaled versions up to 1M records achieve lower Discernibility Metric scores than Original Mondrian for numeric quasi-identifier sets while parallel processing delivers up to 4x speedup vs. sequential Core Mondrian. Core Mondrian enables privacy-compliant equity analytics at production scale.

cs.CR↗

arXiv2025

An Exploratory Study on Multi-modal Generative AI in AR Storytelling

Storytelling in AR has gained attention due to its multi-modality and interactivity. However, generating multi-modal content for AR storytelling requires expertise and efforts for high-quality conveyance of the narrator's intention. Recently, Generative-AI (GenAI) has shown promising applications in multi-modal content generation. Despite the potential benefit, current research calls for validating the effect of AI-generated content (AIGC) in AR Storytelling. Therefore, we conducted an exploratory study to investigate the utilization of GenAI. Analyzing 223 AR videos, we identified a design space for multi-modal AR Storytelling. Based on the design space, we developed a testbed facilitating multi-modal content generation and atomic elements in AR Storytelling. Through two studies with N=30 experienced storytellers and live presenters, we 1. revealed participants' preferences for modalities, 2. evaluated the interactions with AI to generate content, and 3. assessed the quality of the AIGC for AR Storytelling. We further discussed design considerations for future AR Storytelling with GenAI.

cs.HC↗

arXiv2023

"I see models being a whole other thing": An Empirical Study of Pre-Trained Model Naming Conventions and A Tool for Enhancing Naming Consistency

As innovation in deep learning continues, many engineers are incorporating Pre-Trained Models (PTMs) as components in computer systems. Some PTMs are foundation models, and others are fine-tuned variations adapted to different needs. When these PTMs are named well, it facilitates model discovery and reuse. However, prior research has shown that model names are not always well chosen and can sometimes be inaccurate and misleading. The naming practices for PTM packages have not been systematically studied, which hampers engineers' ability to efficiently search for and reliably reuse these models. In this paper, we conduct the first empirical investigation of PTM naming practices in the Hugging Face PTM registry. We begin by reporting on a survey of 108 Hugging Face users, highlighting differences from traditional software package naming and presenting findings on PTM naming practices. The survey results indicate a mismatch between engineers' preferences and current practices in PTM naming. We then introduce DARA, the first automated DNN ARchitecture Assessment technique designed to detect PTM naming inconsistencies. Our results demonstrate that architectural information alone is sufficient to detect these inconsistencies, achieving an accuracy of 94% in identifying model types and promising performance (over 70%) in other architectural metadata as well. We also highlight potential use cases for automated naming tools, such as model validation, PTM metadata generation and verification, and plagiarism detection. Our study provides a foundation for automating naming inconsistency detection. Finally, we envision future work focusing on automated tools for standardizing package naming, improving model selection and reuse, and strengthening the security of the PTM supply chain.

cs.SE↗