Selected work

Representative Papers

Cautious optimism for deep parameterized quantum circuits

Jul 23, 2026

This study investigates how the generalization performance of parameterized quantum circuits (PQCs) evolves as model size increases. Challenging the conventional wisdom that larger models generalize worse, the work demonstrates that deep PQCs trained with gradient-based methods can exhibit a double descent phenomenon—where increasing the number of parameters actually improves generalization. Through rigorous theoretical analysis grounded in perturbation theory and random matrix spectral theory, combined with systematic experiments on re-uploading PQCs across multiple datasets, this paper provides the first theoretical explanation and empirical validation of double descent in quantum machine learning. The findings consistently hold across diverse datasets and training scales, contesting classical notions of generalization and offering both theoretical grounding and practical confidence for scalable quantum machine learning.

0 citationsRead paper

Understanding Cross-Rig Generalization in Automotive Perception: a Multi-Rig Benchmark and Rig Variation Metrics

Jun 25, 2026

This work addresses the challenge of cross-rig generalization in autonomous driving perception systems, which are typically trained under fixed sensor configurations and struggle to adapt to diverse real-world camera layouts due to geometric domain shifts. To this end, the authors introduce the Plentiful CARLA Camera Rigs benchmark, which renders data from 14 systematically designed camera rigs within identical driving scenes. They propose the first controllable evaluation framework for cross-rig generalization and introduce two calibration-based geometric discrepancy metrics—Rig Variance and Rig Contrastive Distance—to quantify inter-rig differences and transfer difficulty. Experimental results demonstrate a strong correlation between geometric discrepancy and performance degradation, with Rig Contrastive Distance effectively predicting the relative difficulty of cross-rig transfer.

0 citationsRead paper

Synchronized Realities: Towards Magic Mobile Experiences through Aligned AR

May 15, 2026

This work addresses the challenge of achieving precise alignment between virtual content and real-world scenes in augmented reality (AR) under uncontrolled physical environments, a limitation that undermines immersion and contextual fidelity. To overcome this, the paper introduces a novel paradigm termed “Synchronized Reality,” which integrates generative artificial intelligence with multimodal context-aware sensing to enable mobile AR experiences dynamically aligned with users’ surroundings. Through a systematic analysis of existing synchronized AR systems, the study identifies commonalities, distinctions, and critical alignment pathways. By examining representative application cases, it demonstrates the paradigm’s efficacy in enhancing environmental consistency while also highlighting emerging opportunities and potential risks for future research and deployment.

0 citationsRead paper

The Renaissance of Repair: A Timely Opportunity for Fabrication Research

May 11, 2026

While existing research on personal fabrication acknowledges sustainability, it rarely centers on repair and lacks systematic technical and methodological support for repair practices. This work proposes the first repair-centered framework for personal fabrication, structuring the repair process into five phases: problem identification, solution exploration, material sourcing, implementation, and testing. Integrating perspectives from human-computer interaction, sustainable manufacturing, and user-driven design, the study analyzes key challenges and research opportunities within each phase. By establishing a novel paradigm that explicitly foregrounds repair in fabrication research, this paper shifts the focus from disposable production toward maintainable and enduring making practices, laying a theoretical foundation for future tools, platforms, and community ecosystems that empower users to independently repair their possessions.

0 citationsRead paper

Behavior-Constrained Reinforcement Learning with Receding-Horizon Credit Assignment for High-Performance Control

Apr 03, 2026

This work addresses the limitations of reinforcement learning, which often deviates from human-like behavior, and imitation learning, which struggles to surpass demonstrator performance. The authors propose a behavior-constrained reinforcement learning framework that models short-horizon future trajectories via receding-horizon prediction and conditions policy learning on reference trajectories. Expert behavioral consistency is enforced at the trajectory distribution level rather than through point-wise matching. By integrating receding-horizon credit assignment with behavior constraints, the approach enables robust generalization under disturbances and varying operating conditions. Evaluated in a high-fidelity racing simulator using professional driver data, the trained policies achieve competitive lap times while closely replicating expert driving styles. Human-in-the-loop assessments further confirm the accurate reproduction of tuning-sensitive driving characteristics.

0 citationsRead paper

Recent publications

Latest Papers

Cautious optimism for deep parameterized quantum circuits

Jul 23, 2026

This study investigates how the generalization performance of parameterized quantum circuits (PQCs) evolves as model size increases. Challenging the conventional wisdom that larger models generalize worse, the work demonstrates that deep PQCs trained with gradient-based methods can exhibit a double descent phenomenon—where increasing the number of parameters actually improves generalization. Through rigorous theoretical analysis grounded in perturbation theory and random matrix spectral theory, combined with systematic experiments on re-uploading PQCs across multiple datasets, this paper provides the first theoretical explanation and empirical validation of double descent in quantum machine learning. The findings consistently hold across diverse datasets and training scales, contesting classical notions of generalization and offering both theoretical grounding and practical confidence for scalable quantum machine learning.

0 citationsRead paper

Understanding Cross-Rig Generalization in Automotive Perception: a Multi-Rig Benchmark and Rig Variation Metrics

Jun 25, 2026

This work addresses the challenge of cross-rig generalization in autonomous driving perception systems, which are typically trained under fixed sensor configurations and struggle to adapt to diverse real-world camera layouts due to geometric domain shifts. To this end, the authors introduce the Plentiful CARLA Camera Rigs benchmark, which renders data from 14 systematically designed camera rigs within identical driving scenes. They propose the first controllable evaluation framework for cross-rig generalization and introduce two calibration-based geometric discrepancy metrics—Rig Variance and Rig Contrastive Distance—to quantify inter-rig differences and transfer difficulty. Experimental results demonstrate a strong correlation between geometric discrepancy and performance degradation, with Rig Contrastive Distance effectively predicting the relative difficulty of cross-rig transfer.

0 citationsRead paper

Synchronized Realities: Towards Magic Mobile Experiences through Aligned AR

May 15, 2026

This work addresses the challenge of achieving precise alignment between virtual content and real-world scenes in augmented reality (AR) under uncontrolled physical environments, a limitation that undermines immersion and contextual fidelity. To overcome this, the paper introduces a novel paradigm termed “Synchronized Reality,” which integrates generative artificial intelligence with multimodal context-aware sensing to enable mobile AR experiences dynamically aligned with users’ surroundings. Through a systematic analysis of existing synchronized AR systems, the study identifies commonalities, distinctions, and critical alignment pathways. By examining representative application cases, it demonstrates the paradigm’s efficacy in enhancing environmental consistency while also highlighting emerging opportunities and potential risks for future research and deployment.

0 citationsRead paper

The Renaissance of Repair: A Timely Opportunity for Fabrication Research

May 11, 2026

While existing research on personal fabrication acknowledges sustainability, it rarely centers on repair and lacks systematic technical and methodological support for repair practices. This work proposes the first repair-centered framework for personal fabrication, structuring the repair process into five phases: problem identification, solution exploration, material sourcing, implementation, and testing. Integrating perspectives from human-computer interaction, sustainable manufacturing, and user-driven design, the study analyzes key challenges and research opportunities within each phase. By establishing a novel paradigm that explicitly foregrounds repair in fabrication research, this paper shifts the focus from disposable production toward maintainable and enduring making practices, laying a theoretical foundation for future tools, platforms, and community ecosystems that empower users to independently repair their possessions.

0 citationsRead paper

Behavior-Constrained Reinforcement Learning with Receding-Horizon Credit Assignment for High-Performance Control

Apr 03, 2026

This work addresses the limitations of reinforcement learning, which often deviates from human-like behavior, and imitation learning, which struggles to surpass demonstrator performance. The authors propose a behavior-constrained reinforcement learning framework that models short-horizon future trajectories via receding-horizon prediction and conditions policy learning on reference trajectories. Expert behavioral consistency is enforced at the trajectory distribution level rather than through point-wise matching. By integrating receding-horizon credit assignment with behavior constraints, the approach enables robust generalization under disturbances and varying operating conditions. Evaluated in a high-fidelity racing simulator using professional driver data, the trained policies achieve competitive lap times while closely replicating expert driving styles. Human-in-the-loop assessments further confirm the accurate reproduction of tuning-sensitive driving characteristics.

0 citationsRead paper