arXiv subjects

Juho Kim

Publications and source records attributed to Juho Kim.

At least 37 records · Page 2Linked to original sources

arXiv2026

Parallelizing Counterfactual Regret Minimization

Parallelization has played an instrumental role in the field of artificial intelligence (AI), drastically reducing the time taken to train and evaluate large AI models. In contrast to its impact in the broader field of AI, applying parallelization to computational game solving is relatively unexplored, despite its great potential. In this paper, we parallelize the family of counterfactual regret minimization (CFR) algorithms, which were central to important breakthroughs for solving large imperfect-information games. We present a generalized parallelization framework, reframing CFR as a series of linear algebra operations. Then, existing techniques for parallelizing linear algebra operations can be applied to accelerate CFR. We also describe how our technique can be applied to other tabular members of the CFR family of algorithms, including the state-of-the-art, such as CFR+, discounted CFR, and predictive variants of CFR. Experimentally, we show that our CFR implementation on a GPU is up to four orders of magnitude faster than Google DeepMind OpenSpiel's CFR implementations on a CPU.

cs.AI↗

arXiv2026

Watermarking Game-Playing Agents in Perfect-Information Extensive-Form Games

Watermarking techniques for large language models (LLMs), which encode hidden information in the output so its source can be verified, have gained significant attention in recent days, thanks to their potential capability to detect accidental or deliberate misuse. Similar challenges involving model misuse also exist in the context of game-playing, such as when detecting the unauthorized use of AI tools in gaming platforms (e.g., cheating in online chess). In this paper, we initiate the study of how game-playing strategies can be watermarked. We show how the KGW watermark for LLMs can be adapted to watermark game-playing agents in perfect-information extensive-form games. The watermark can then be detected using a statistical test. We show that the degradation in the quality of the watermarked strategy profile, quantified by the expected utility, can be bounded, but there is a tradeoff between detectability and quality. In our experiments, we bootstrap the watermarking framework to various chess engines and demonstrate that a) the impact of the watermark on the quality of the strategy is negligible and b) the watermark can be detected with just a handful of games.

cs.GT↗

arXiv2026

DiscoverLLM: From Executing Intents to Discovering Them

To handle ambiguous and open-ended requests, Large Language Models (LLMs) are increasingly trained to interact with users to surface intents they have not yet expressed (e.g., ask clarification questions). However, users are often ambiguous because they have not yet formed their intents: they must observe and explore outcomes to discover what they want. Simply asking "what kind of tone do you want?" fails when users themselves do not know. We introduce DiscoverLLM, a novel and generalizable framework that trains LLMs to help users form and discover their intents. Central to our approach is a novel user simulator that models cognitive state with a hierarchy of intents that progressively concretize as the model surfaces relevant options -- where the degree of concretization serves as a reward signal that models can be trained to optimize. Resulting models learn to collaborate with users by adaptively diverging (i.e., explore options) when intents are unclear, and converging (i.e., refine and implement) when intents concretize. Across proposed interactive benchmarks in creative writing, technical writing, and SVG drawing, DiscoverLLM achieves over 10% higher task performance while reducing conversation length by up to 40%. In a user study with 75 human participants, DiscoverLLM improved conversation satisfaction and efficiency compared to baselines.

cs.AI↗

arXiv2026

IdeaBlocks: Expressing and Reusing Divergent Intents for Graphic Design Exploration using Generative AI

While designers increasingly leverage Generative AI for divergent exploration, current interaction is optimized for convergent refinement, forcing users to specify fixed targets rather than open-ended search spaces. Based on a formative study (N=7), we define the anatomy of Divergent Intent, comprising property, direction, and range, and identified two critical barriers: the lack of mechanisms to explicitly shape the parametric boundaries of exploration and the difficulty of reusing successful search strategies. We present IdeaBlocks, where users can modularize divergent intents into Exploration Blocks. Users can reuse prior intents at multiple levels (block, path, and project) with options for literal or context-adaptive reuse. In our comparative study (N=12), participants using IdeaBlocks explored 2.13 times more images with 12.5% greater visual diversity than the baseline, demonstrating how structured intent expression and reuse support effective divergence. A three-day deployment study (N=6) further revealed how different reuse mechanisms allowed distinct creative strategies, offering design implications for future intent-aware creativity supports.

cs.HC↗

arXiv2026

Personalizing Cancer Models under Data Scarcity via Parameter Decomposition

Personalized cancer modeling for clinical applications requires robust and efficient parameter calibration, particularly in settings with limited patient data. This need is especially critical for medical digital twins (MDTs), which are virtual representations of disease continuously updated using longitudinal patient measurements. In this work, we propose a novel parameter personalization framework for dynamical cancer models under data scarcity. Our approach decomposes selected model parameters into a common component, shared across patients, and a personalized component, which is patient-specific and can be updated as new data become available. The common component captures population-level structure and is estimated once, providing an informed prior that enables rapid and accurate personalization. We demonstrate the effectiveness of this framework using synthetic data generated from canonical dynamical systems, such as logistic growth models with optimized treatment interventions. Our results show that parameter decomposition significantly improves calibration performance in limited-data regimes, facilitating fast and reliable personalization and supporting the development of patient-specific cancer models and MDTs.

q-bio.OT↗

arXiv2026

Hint-Writing with Deferred AI Assistance: Fostering Critical Engagement in Data Science Education

Generating hints for incorrect code is a cognitively demanding task that fosters learning and metacognitive development. This study investigates three designs for personalized, scalable, and reflective hint-writing activities within a data science course: (i) writing a hint independently, (ii) writing a hint with on-demand AI assistance, and (iii) deferred AI assistance, in which students first write a hint independently and then revise it with the help of an AI-generated one. We examine how AI support can scaffold the learning process without diminishing students' productive cognitive effort. Through a randomized controlled experiment with graduate-level students (N=97), we found that deferring AI assistance leads to the highest-quality hints. Further, this design helps students identify a wide range of mistakes they otherwise struggle to identify without any AI assistance. Students valued these activities as opportunities to practice debugging and critically engage with AI outputs--skills that are now critical for learners to acquire as programming becomes increasingly automated and the use of AI for learning grows. Our findings also highlight key considerations for designing student-AI collaborative learning experiences to sustain student engagement, maintain appropriate cognitive load, and mitigate negative effects of AI, such as introducing redundancies and extraneous information into student work.

cs.HC↗

arXiv2026

Evalet: Evaluating Large Language Models through Functional Fragmentation

Practitioners increasingly rely on Large Language Models (LLMs) to evaluate generative AI outputs through "LLM-as-a-Judge" approaches. However, these methods produce holistic scores that obscure which specific elements influenced the assessments. We propose functional fragmentation, a method that dissects each output into key fragments and interprets the rhetoric functions that each fragment serves relative to evaluation criteria -- surfacing the elements of interest and revealing how they fulfill or hinder user goals. We instantiate this approach in Evalet, an interactive system that visualizes fragment-level functions across many outputs to support inspection, rating, and comparison of evaluations. A user study (N=10) found that, while practitioners struggled to validate holistic scores, our approach helped them identify 48% more evaluation misalignments. This helped them calibrate trust in LLM evaluations and rely on them to find more actionable issues in model outputs. Our work shifts LLM evaluation from quantitative scores toward qualitative, fine-grained analysis of model behavior.

cs.HC↗

arXiv2026

GUIDE: A Benchmark for Understanding and Assisting Users in Open-Ended GUI Tasks

Graphical User Interface (GUI) agents have the potential to assist users in interacting with complex software (e.g., PowerPoint, Photoshop). While prior research has primarily focused on automating user actions through clicks and keystrokes, this paradigm overlooks human intention, where users value the ability to explore, iterate, and refine their ideas while maintaining agency. To move beyond automation and toward collaboration, GUI agents must understand what users are doing and why. We introduce GUIDE (GUI User Intent Detection Evaluation), a benchmark that evaluates AI models on their ability to perceive user behavior, infer intent, and provide assistance in open-ended GUI tasks. GUIDE consists of 67.5 hours of screen recordings from 120 novice user demonstrations with think-aloud narrations, across 10 software. GUIDE defines three tasks - (i) Behavior State Detection, (ii) Intent Prediction, and (iii) Help Prediction that test a model's ability to recognize behavior state, reason about goals, and decide when and how to help. Evaluations across eight state-of-the-art multimodal models reveal that all models struggled, achieving only 44.6% and 55.0% accuracy on behavior state and help prediction. However, providing user context significantly improved the performance, raising help prediction by up to 50.2pp, highlighting the critical role of structured user understanding in effective assistance. Our dataset is available at https://guide-bench.github.io.

cs.CV↗

arXiv2026

Precision Proactivity: Measuring Cognitive Load in Real-World AI-Assisted Work

Systems like ChatGPT and Claude assist billions through proactive dialogue-offering unsolicited, task-relevant information. Drawing on Cognitive Load Theory, we study how cognitive load shapes performance in AI-assisted knowledge work. We recruited 34 financial professionals to complete a complex valuation task using GPT-4o and developed a transcript-based framework estimating intrinsic and extraneous load from computational indicators anchored in a task decomposition and knowledge graph. Across 1,178 participant-subtask observations, AI-generated content usage is positively associated with quality, while extraneous load shows the largest negative association-roughly three times that of intrinsic load. Mediation reveals a compensatory pathway partially offsetting but not eliminating load-related deficits. Extraneous load persists within speakers and spills asymmetrically to model responses. Model-initiated task switching is the strongest predictor of decline. Expertise moderates these dynamics: less experienced professionals face larger penalties and derive greater marginal gains from AI-generated content, yet are not those who most increase uptake under load.

cs.AI↗

arXiv2026

When Scaffolding Breaks: Investigating Student Interaction with LLM-Based Writing Support in Real-Time K-12 EFL Classrooms

Large language models (LLMs) are promising tools for scaffolding students' English writing skills, but their effectiveness in real-time K-12 classrooms remains underexplored. Addressing this gap, our study examines the benefits and limitations of using LLMs as real-time learning support, considering how classroom constraints, such as diverse proficiency levels and limited time, affect their effectiveness. We conducted a deployment study with 157 eighth-grade students in a South Korean middle school English class over six weeks. Our findings reveal that while scaffolding improved students' ability to compose grammatically correct sentences, this step-by-step approach demotivated lower-proficiency students and increased their system reliance. We also observed challenges to classroom dynamics, where extroverted students often dominated the teacher's attention, and the system's assistance made it difficult for teachers to identify struggling students. Based on these findings, we discuss design guidelines for integrating LLMs into real-time writing classes as inclusive educational tools.

cs.HC↗

arXiv2026

ClearFairy: Capturing Creative Workflows through Decision Structuring, In-Situ Questioning, and Rationale Inference

Capturing professionals' decision-making in creative workflows (e.g., UI/UX) is essential for reflection, collaboration, and knowledge sharing, yet existing methods often leave rationales incomplete and implicit decisions hidden. To address this, we present the CLEAR approach, which structures reasoning into cognitive decision steps-linked units of actions, artifacts, and explanations making decisions traceable with generative AI. Building on CLEAR, we introduce ClearFairy, a think-aloud AI assistant for UI design that detects weak explanations, asks lightweight clarifying questions, and infers missing rationales. In a study with twelve professionals, 85% of ClearFairy's inferred rationales were accepted (as-is or with revisions). Notably, the system increased "strong explanations"-rationales providing sufficient causal reasoning-from 14% to 83% without adding cognitive demand. Furthermore, exploratory applications demonstrate that captured steps can enhance generative AI agents in Figma, yielding predictions better aligned with professionals and producing coherent outcomes. We release a dataset of 417 decision steps to support future research.

cs.HC↗

arXiv2026

ChoiceMates: Supporting Unfamiliar Online Decision-Making with Multi-Agent Conversational Interactions

From deciding on a PhD program to buying a new camera, unfamiliar decisions--decisions without domain knowledge--are frequent and significant. The complexity and uncertainty of such decisions demand unique approaches to information seeking, understanding, and decision-making. Our formative study highlights that users want to start by discovering broad and relevant domain information evenly and simultaneously, quickly address emerging inquiries, and gain personalized standards to assess information found. We present ChoiceMates, an interactive multi-agent system designed to address these needs by enabling users to engage with a dynamic set of LLM agents each presenting a unique experience in the domain. Unlike existing multi-agent systems that automate tasks with agents, the user orchestrates agents to assist their decision-making process. Our user evaluation (n=12) shows that ChoiceMates enables a more confident, satisfactory decision-making with better situation understanding than web search, and higher decision quality and confidence than a commercial multi-agent framework. This work provides insights into designing a more controllable and collaborative multi-agent system.

cs.HC↗

arXiv2026

Authorship Drift: How Self-Efficacy and Trust Evolve During LLM-Assisted Writing

Large language models (LLMs) are increasingly used as collaborative partners in writing. However, this raises a critical challenge of authorship, as users and models jointly shape text across interaction turns. Understanding authorship in this context requires examining users' evolving internal states during collaboration, particularly self-efficacy and trust. Yet, the dynamics of these states and their associations with users' prompting strategies and authorship outcomes remain underexplored. We examined these dynamics through a study of 302 participants in LLM-assisted writing, capturing interaction logs and turn-by-turn self-efficacy and trust ratings. Our analysis showed that collaboration generally decreased users' self-efficacy while increasing trust. Participants who lost self-efficacy were more likely to ask the LLM to edit their work directly, whereas those who recovered self-efficacy requested more review and feedback. Furthermore, participants with stable self-efficacy showed higher actual and perceived authorship of the final text. Based on these findings, we propose design implications for understanding and supporting authorship in human-LLM collaboration.

cs.HC↗

arXiv2026

"I Can't Keep Up": Accessibility Barriers in Video-Based Learning for Individuals with Borderline Intellectual Functioning

Video-based learning (VBL) has become a dominant method for learning practical skills, yet accessibility guidelines provide limited guidance for users with cognitive differences. In particular, challenges that individuals with Borderline Intellectual Functioning (BIF) encounter in video-based learning remain largely underexplored, despite VBL's potential to support their learning through features like self-paced viewing and visual demonstration. To address this gap, we conducted a series of studies with BIF individuals and caretakers to comprehensively understand their VBL challenges. Our analysis revealed challenges stemming from misalignment between user cognitive characteristics and video elements (e.g., overwhelmed by pacing and density, difficulty inferring omitted content), and experiential factors intensifying challenges (e.g., low self-efficacy). While participants employed coping strategies such as repetitive viewing to address these challenges, these strategies could not overcome fundamental gaps with video. We further discuss the design implications on both content and UI-level features for BIF and broader groups with cognitive diversities.

cs.HC↗

arXiv2026

Benchmarking Large Language Models for Diagnosing Students' Cognitive Skills from Handwritten Math Work

Students' handwritten math work provides a rich resource for diagnosing cognitive skills, as it captures intermediate reasoning beyond final answers. We investigate how current large language models (LLMs) perform in diagnosing cognitive skills from such work. However, student responses vary widely, often omitting steps or providing only vague, contextually implicit evidence. Despite recent advances in LLMs' multimodal and reasoning capabilities, their performance under such conditions remains underexplored. To address this gap, we constructed MathCog, a benchmark dataset containing 3,036 diagnostic verdicts across 639 student responses to 110 math problems, annotated by teachers using TIMSS-grounded cognitive skill checklists with evidential strength labels (Evident/Vague). Evaluating 18 LLMs, we find that (1) all models underperform (F1 < 0.5) regardless of capability, and (2) performance degrades sharply under vague evidence. Error analysis reveals systematic patterns: models frequently misattribute Vague evidence as Evident, overthink minimal cues, and hallucinate nonexistent evidence. We discuss implications for evidence-aware, teacher-in-the-loop designs for LLM-based cognitive diagnosis in educational settings.

cs.AI↗

arXiv2026

DiaryPlay: AI-Assisted Authoring of Interactive Vignettes for Everyday Storytelling

An interactive vignette is a popular and immersive visual storytelling approach that invites viewers to role-play a character and influences the narrative in an interactive environment. However, it has not been widely used by everyday storytellers yet due to authoring complexity, which conflicts with the immediacy of everyday storytelling. We introduce DiaryPlay, an AI-assisted authoring system for interactive vignette creation in everyday storytelling. It takes a natural language story as input and extracts the three core elements of an interactive vignette (environment, characters, and events), enabling authors to focus on refining these elements instead of constructing them from scratch. Then, it automatically transforms the single-branch story input into a branch-and-bottleneck structure using an LLM-powered narrative planner, which enables flexible viewer interactions while freeing the author from multi-branching. A technical evaluation (N=16) shows that DiaryPlay-generated character activities are on par with human-authored ones regarding believability. A user study (N=16) shows that DiaryPlay effectively supports authors in creating interactive vignette elements, maintains authorial intent while reacting to viewer interactions, and provides engaging viewing experiences.

cs.HC↗

arXiv2026

IntentFlow: Investigating Fluid Dynamics of Intent Communication in Generative AI

Generative AI shifts interaction toward intent-based outcome specification, despite user intents being inherently vague, fluid, and evolving. While a growing body of HCI research has proposed diverse interaction techniques to support this process, there is limited understanding of what the key aspects of intent communication are and how they interplay to shape users' workflows. To bridge this gap, we first conduct a systematic literature review of 46 HCI papers and identify four core aspects of intent communication support: intent articulation, exploration, management, and synchronization. To investigate how these aspects interplay in practice, we developed IntentFlow, a research probe that embodies all four aspects for a writing task, and conducted a comparative study (N=12). Our action-level behavioral analysis reveals that comprehensive support enables verification-driven refinement and progressive intent curation, reduces cognitive effort, and improves users' sense of control and understanding of intent-output alignment. We conclude with design implications for building generative AI systems that support intent communication as a dynamic, iterative process.

cs.HC↗

arXiv2025

Understanding EFL Learners' Code-Switching and Teachers' Pedagogical Approaches in LLM-Supported Speaking Practice

For English as a Foreign Language (EFL) learners, code-switching (CSW), or alternating between their native language and the target language (English), can lower anxiety and ease communication barriers. Large language models (LLMs), with their multilingual abilities, offer new opportunities to support CSW in speaking practice. Yet, the pedagogical design of LLM-based tutors remains underexplored. To this end, we conducted a six-week study of LLM-mediated speaking practice with 20 Korean EFL learners, alongside a qualitative study with nine English teachers who designed and refined responses to learner CSW. Findings show that learners used CSW not only to bridge lexical gaps but also to express cultural and emotional nuance, prompting teachers to employ selective interventions and dynamic scaffolding strategies. We conclude with design implications for bilingual LLM-powered tutors that leverage teachers' expertise to transform CSW into meaningful learning opportunities.

cs.HC↗