2025
State Your Intention to Steer Your Attention: An AI assistant for Intentional Digital Living
Juheon Choi, Juyong Lee, Jian Kim, Chanyoung Kim, Taewon Min, W. Bradley Knox, Min Kyung Lee, Kimin Lee
Accepted to ACM Conference on Human Factors in Computing Systems (CHI) 2026
People often face distractions when using digital devices. To address this, we propose an AI assistant that infers user intentions, monitors alignment with them, and provides gentle nudges when deviations occur. In a three-week study with 22 participants, the system outperformed rule-based and passive baselines in helping users stay focused and aligned with their goals.
State Your Intention to Steer Your Attention: An AI assistant for Intentional Digital Living
Juheon Choi, Juyong Lee, Jian Kim, Chanyoung Kim, Taewon Min, W. Bradley Knox, Min Kyung Lee, Kimin Lee
Accepted to ACM Conference on Human Factors in Computing Systems (CHI) 2026
People often face distractions when using digital devices. To address this, we propose an AI assistant that infers user intentions, monitors alignment with them, and provides gentle nudges when deviations occur. In a three-week study with 22 participants, the system outperformed rule-based and passive baselines in helping users stay focused and aligned with their goals.
Subtle Risks, Critical Failures: A Framework for Diagnosing Physical Safety of LLMs for Embodied Decision Making
Yejin Son, Minseo Kim, Sungwoong Kim, Seungju Han, Jian Kim, Dongju Jang, Youngjae Yu, Chanyoung Park
Accepted to Empirical Methods in Natural Language Processing (EMNLP) Main 2025
Evaluation across 13 state-of-the-art LLMs reveals that while models often reject clearly unsafe commands, they struggle to anticipate and mitigate subtle, situational risks, and this work introduces SAFEL, the framework for systematically evaluating the physical safety of LLMs in embodied decision making.
Subtle Risks, Critical Failures: A Framework for Diagnosing Physical Safety of LLMs for Embodied Decision Making
Yejin Son, Minseo Kim, Sungwoong Kim, Seungju Han, Jian Kim, Dongju Jang, Youngjae Yu, Chanyoung Park
Accepted to Empirical Methods in Natural Language Processing (EMNLP) Main 2025
Evaluation across 13 state-of-the-art LLMs reveals that while models often reject clearly unsafe commands, they struggle to anticipate and mitigate subtle, situational risks, and this work introduces SAFEL, the framework for systematically evaluating the physical safety of LLMs in embodied decision making.
sudo rm ‑rf agentic_security
Sejin Lee*, Jian Kim*, Haon Park, Ashkan Yousefpour, Sangyoon Yu, Min Song (* equal contribution)
Accepted to 63rd Annual Meeting of the Association for Computational Linguistics(ACL) (Volume 6: Industry Track) 2025
SUDO (Screen-based Universal Detox2Tox Offense), a novel attack framework that systematically bypasses refusal-trained safeguards in commercial computer-use agents, such as Claude for Computer Use, is presented.
sudo rm ‑rf agentic_security
Sejin Lee*, Jian Kim*, Haon Park, Ashkan Yousefpour, Sangyoon Yu, Min Song (* equal contribution)
Accepted to 63rd Annual Meeting of the Association for Computational Linguistics(ACL) (Volume 6: Industry Track) 2025
SUDO (Screen-based Universal Detox2Tox Offense), a novel attack framework that systematically bypasses refusal-trained safeguards in commercial computer-use agents, such as Claude for Computer Use, is presented.