Selected work

Representative Papers

Partitioning Time in Target Trial Emulation

Sep 28, 2026

This study addresses the vulnerability of standard estimators in target trial emulation to immortal time bias and time-varying confounding arising from inappropriate temporal discretization. By subdividing follow-up time windows, this work constructs directed acyclic graphs and ancestral multi-world networks to elucidate how intra-interval causal ordering influences estimation. Building on these insights, we derive the g-formula and rectify the clone-censor-weight estimator, proposing a refined approach that restores its validity under fine-grained time partitions. Ultimately, this research establishes an unbiased effect estimation methodology alongside principled criteria for selecting temporal discretization schemes. These contributions offer actionable guidance for clinical studies, enabling more accurate evaluations of treatment effects while mitigating biases inherent in conventional emulation frameworks.

0 citationsRead paper

Towards Socially Compliant Navigation in Deep Reinforcement Learning via Proxemics-Based Reward Modeling

Aug 13, 2026

This work addresses the tendency of existing deep reinforcement learning–based navigation methods to prioritize task objectives at the expense of social compliance in dense crowds, often resulting in behaviors that violate human social norms. To mitigate this issue, the authors propose a differentiable reward modeling approach grounded in Hall’s proxemic theory, formalizing personal space as a radial Gaussian mixture field. This formulation enables the computation of a local social cost within the robot’s field of view, which is seamlessly integrated into a deep reinforcement learning framework. The method uniquely translates proxemic theory into a dense, interpretable, and differentiable reward signal, significantly improving social compliance across diverse crowd densities and environments while maintaining navigation efficiency comparable to state-of-the-art approaches.

0 citationsRead paper

Recent publications

Latest Papers

Partitioning Time in Target Trial Emulation

Sep 28, 2026

This study addresses the vulnerability of standard estimators in target trial emulation to immortal time bias and time-varying confounding arising from inappropriate temporal discretization. By subdividing follow-up time windows, this work constructs directed acyclic graphs and ancestral multi-world networks to elucidate how intra-interval causal ordering influences estimation. Building on these insights, we derive the g-formula and rectify the clone-censor-weight estimator, proposing a refined approach that restores its validity under fine-grained time partitions. Ultimately, this research establishes an unbiased effect estimation methodology alongside principled criteria for selecting temporal discretization schemes. These contributions offer actionable guidance for clinical studies, enabling more accurate evaluations of treatment effects while mitigating biases inherent in conventional emulation frameworks.

0 citationsRead paper

Towards Socially Compliant Navigation in Deep Reinforcement Learning via Proxemics-Based Reward Modeling

Aug 13, 2026

This work addresses the tendency of existing deep reinforcement learning–based navigation methods to prioritize task objectives at the expense of social compliance in dense crowds, often resulting in behaviors that violate human social norms. To mitigate this issue, the authors propose a differentiable reward modeling approach grounded in Hall’s proxemic theory, formalizing personal space as a radial Gaussian mixture field. This formulation enables the computation of a local social cost within the robot’s field of view, which is seamlessly integrated into a deep reinforcement learning framework. The method uniquely translates proxemic theory into a dense, interpretable, and differentiable reward signal, significantly improving social compliance across diverse crowd densities and environments while maintaining navigation efficiency comparable to state-of-the-art approaches.

0 citationsRead paper