Selected work

Representative Papers

Memory-Guided B-Roll Generation from User Video Collections

Oct 01, 2026

This study addresses the challenge of extracting visual evidence from massive user videos to generate multi-shot B-roll sequences with consistent characters, scenes, and styles. To this end, it proposes MemComposer, a system employing a three-stage architecture that transforms raw videos into entity-centric structured memories. By integrating natural language instruction planning, conditional retrieval, and iterative generation alongside an automated critique-feedback mechanism, the approach enables precise reference-frame-based generation and consistency verification. This work transcends the limitations of conventional text-to-video paradigms through memory grounding. User preference studies demonstrate that the system achieves 60.0% prompt adherence and up to 92.8% visual alignment, significantly outperforming ungrounded baselines and thereby validating the effectiveness of the proposed memory-grounding strategy.

0 citationsRead paper

How Should We Teach Robots? A Comparison of Kinesthetic, Joystick, and Gesture-Based Teaching

May 27, 2026

This study systematically investigates the trade-offs between usability and task performance across three prevalent robot teaching modalities: kinesthetic guidance, joystick-based teleoperation, and gesture-based demonstration. Through a user study involving three manipulation tasks of increasing complexity, the methods are quantitatively evaluated using the NASA-TLX workload scale, task reproduction success rates, and error-type analysis. Results indicate that kinesthetic guidance significantly outperforms the alternatives in direction-sensitive and contact-intensive tasks, while joystick control achieves the highest efficiency in simple pick-and-place operations. Although gesture-based teaching exhibits lower overall reliability, it approaches the performance of kinesthetic guidance in specific scenarios. These findings delineate the operational boundaries of each teaching modality and provide empirical support for task-driven selection of robot programming interfaces.

0 citationsRead paper

Recent publications

Latest Papers

Memory-Guided B-Roll Generation from User Video Collections

Oct 01, 2026

This study addresses the challenge of extracting visual evidence from massive user videos to generate multi-shot B-roll sequences with consistent characters, scenes, and styles. To this end, it proposes MemComposer, a system employing a three-stage architecture that transforms raw videos into entity-centric structured memories. By integrating natural language instruction planning, conditional retrieval, and iterative generation alongside an automated critique-feedback mechanism, the approach enables precise reference-frame-based generation and consistency verification. This work transcends the limitations of conventional text-to-video paradigms through memory grounding. User preference studies demonstrate that the system achieves 60.0% prompt adherence and up to 92.8% visual alignment, significantly outperforming ungrounded baselines and thereby validating the effectiveness of the proposed memory-grounding strategy.

0 citationsRead paper

How Should We Teach Robots? A Comparison of Kinesthetic, Joystick, and Gesture-Based Teaching

May 27, 2026

This study systematically investigates the trade-offs between usability and task performance across three prevalent robot teaching modalities: kinesthetic guidance, joystick-based teleoperation, and gesture-based demonstration. Through a user study involving three manipulation tasks of increasing complexity, the methods are quantitatively evaluated using the NASA-TLX workload scale, task reproduction success rates, and error-type analysis. Results indicate that kinesthetic guidance significantly outperforms the alternatives in direction-sensitive and contact-intensive tasks, while joystick control achieves the highest efficiency in simple pick-and-place operations. Although gesture-based teaching exhibits lower overall reliability, it approaches the performance of kinesthetic guidance in specific scenarios. These findings delineate the operational boundaries of each teaching modality and provide empirical support for task-driven selection of robot programming interfaces.

0 citationsRead paper