Yu Yang

Smart assistants increasingly act proactively, yet mistimed or intrusive behavior often causes users to lose trust and disable these features. Learning user preferences for proactive assistance is difficult because real-world studies are costly, limited in scale, and rarely capture how preferences change across multiple interaction sessions. Large language model-based generative agents offer a way to simulate realistic interactions, but existing synthetic datasets remain limited in temporal depth, diverse personas, and multi-dimensional preferences. They also provide little support for transferring population-level insights to individual users under on-device constraints. We present a population-to-individual learning framework for preference-aligned proactive assistants that operates under on-device and privacy constraints. Our approach uses large-scale interaction simulation with 1,000 diverse personas to learn shared structure in how users express preferences across recurring dimensions such as timing, autonomy, and communication style, providing a strong cold start without relying on real user logs. The assistant then adapts to individual users on device through lightweight activation-based steering driven by simple interaction feedback, without model retraining or cloud-side updates. We evaluate the framework using controlled simulations with 1,000 simulated personas and a human-subject study with 34 participants. Results show improved timing decisions and perceived interaction quality over untuned and direct-response baselines, while on-device activation steering achieves performance comparable to reinforcement learning from human feedback. Participants also report higher satisfaction, trust, and comfort as the assistant adapts over multiple interaction sessions.

Last-mile delivery from delivery stations to customers is now mainly finished by dedicated couriers. In practice, each courier generally collects orders destined for one delivery area at the station and delivers them. However, the long distance between the station and the delivery area, due to practical reasons such as the expensive station rental fee in the downtown area, increases the travel time and decreases delivery efficiency. We solve the problem with relay transportation, where one or multiple relay couriers collect orders at the station and send them to delivery couriers, and delivery couriers focus on order delivery in the corresponding areas. We design a mobility-aware single relay courier scheduling system called REDE to minimize the average relay order delivery time (ARODT) for last-mile delivery. First, a heterogeneous task aware route prediction algorithm is proposed to characterize the delivery courier's mobility. After that, a distance-aware greedy algorithm and an ARODT-constrained exchange algorithm are designed to generate the relay route. A hybrid last-mile delivery system REDE+ is then designed to further reduce ARODT of REDE. We also propose REDE-CO to optimize multiple relay courier cooperative scheduling for order relay. Extensive evaluation shows the effectiveness of REDE, REDE+, and REDE-CO in reducing ARODT.

Large-scale human mobility simulation is critical for many science domains such as urban science, epidemiology, and transportation analysis. Recent works treat large language models (LLMs) as human agents to simulate realistic mobility trajectories by modeling individual-level cognitive processes. However, these approaches generate individual mobility trajectories independently, without any population-level coordination mechanism, and thus fail to capture the emergence of collective behaviors. To address this issue, we design M2LSimu, a mobility measures-guided multi-prompt adjustment framework that leverages mobility measures derived from shared data as guidance to refine individual-level prompts for realistic mobility generation. Our framework applies coarse-grained adjustment strategies guided by mobility measures, progressively enabling fine-grained individual-level adaptation while satisfying multiple population-level mobility objectives under a limited budget. Experiments show that M2LSimu significantly outperforms state-of-the-art LLM-based methods on two public datasets.

Large-scale human mobility simulation is critical for applications such as urban planning, epidemiology, and transportation analysis. Recent works treat large language models (LLMs) as human agents to simulate realistic mobility behaviors using structured reasoning, but their high computational cost limits scalability. To address this, we design a mobility-aware cache framework named MobCache that leverages reconstructible caches to enable efficient large-scale human mobility simulations. It consists of: (1) a reasoning component that encodes each reasoning step as a latent-space embedding and uses a latent-space evaluator to enable the reuse and recombination of reasoning steps; and (2) a decoding component that employs a lightweight decoder trained with mobility law-constrained distillation to translate latent-space reasoning chains into natural language, thereby improving simulation efficiency while maintaining fidelity. Experiments show that MobCache significantly improves efficiency across multiple dimensions while maintaining performance comparable to state-of-the-art LLM-based methods.

Smart assistants increasingly act proactively, yet mistimed or intrusive behavior often causes users to lose trust and disable these features. Learning user preferences for proactive assistance is difficult because real-world studies are costly, limited in scale, and rarely capture how preferences change across multiple interaction sessions. Large language model based generative agents offer a way to simulate realistic interactions, but existing synthetic datasets remain limited in temporal depth, diverse personas, and multi-dimensional preferences. They also provide little support for transferring population-level insights to individual users under on-device constraints. We present a population-to-individual learning framework for preference-aligned proactive assistants that operates under on-device and privacy constraints. Our approach uses large-scale interaction simulation with 1,000 diverse personas to learn shared structure in how users express preferences across recurring dimensions such as timing, autonomy, and communication style, providing a strong cold start without relying on real user logs. The assistant then adapts to individual users on device through lightweight activation-based steering driven by simple interaction feedback, without model retraining or cloud-side updates. We evaluate the framework using controlled simulations with 1,000 simulated personas and a human-subject study with 30 participants. Results show improved timing decisions and perceived interaction quality over untuned and direct-response baselines, while on-device activation steering achieves performance comparable to reinforcement learning from human feedback. Participants also report higher satisfaction, trust, and comfort as the assistant adapts over multiple sessions of interactions.

Designing and evaluating personalized and proactive assistant agents remains challenging due to the time, cost, and ethical concerns associated with human-in-the-loop experimentation. Existing Human-Computer Interaction (HCI) methods often require extensive physical setup and human participation, which introduces privacy concerns and limits scalability. Simulated environments offer a partial solution but are typically constrained by rule-based scenarios and still depend heavily on human input to guide interactions and interpret results. Recent advances in large language models (LLMs) have introduced the possibility of generative agents that can simulate realistic human behavior, reasoning, and social dynamics. However, their effectiveness in modeling human-assistant interactions remains largely unexplored. To address this gap, we present a generative agent-based simulation platform designed to simulate human-assistant interactions. We identify ten prior studies on assistant agents that span different aspects of interaction design and replicate these studies using our simulation platform. Our results show that fully simulated experiments using generative agents can approximate key aspects of human-assistant interactions. Based on these simulations, we are able to replicate the core conclusions of the original studies. Our work provides a scalable and cost-effective approach for studying assistant agent design without requiring live human subjects. Additional resources and project materials are available at https://dash-gidea.github.io/.

Human Activity Recognition (HAR) using Inertial Measurement Unit (IMU) sensors is critical for applications in healthcare, safety, and industrial production. However, variations in activity patterns, device types, and sensor placements create distribution gaps across datasets, reducing the performance of HAR models. To address this, we propose LanHAR, a novel system that leverages Large Language Models (LLMs) to generate semantic interpretations of sensor readings and activity labels for cross-dataset HAR. This approach not only mitigates cross-dataset heterogeneity but also enhances the recognition of new activities. LanHAR employs an iterative re-generation method to produce high-quality semantic interpretations with LLMs and a two-stage training framework that bridges the semantic interpretations of sensor readings and activity labels. This ultimately leads to a lightweight sensor encoder suitable for mobile deployment, enabling any sensor reading to be mapped into the semantic interpretation space. Experiments on five public datasets demonstrate that our approach significantly outperforms state-of-the-art methods in both cross-dataset HAR and new activity recognition. The source code is publicly available at https://github.com/DASHLab/LanHAR.