Yichen Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yichen Liu has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Imam Akintomiwa Akinlade
Harvard Business School
1 shared publication - Kai Liu
Chongqing University
1 shared publication - Lei Yang
South China University of Technology
1 shared publication - Pan Zhou
Huazhong University of Science and Technology
1 shared publication - Shiqi Yang1 shared publication
- Vinay Venkatesh
Mountain View
1 shared publication
Latest Publications
- 5 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 0620-0626
DP-Prox: A Robust and Differentially Private Framework for Federated Instruction Tuning of Small LLMs on 8 GB Edge Devices
This paper introduces DP-Prox, an end-to-end framework for robust and differentially private federated instruction tuning of small Large Language Models (LLMs) on 8 GB edge devices. DP-Prox is instantiated using the 3.8 billionparameter Phi-3-mini model with 4-bit Quantized Low-Rank Adaptation (QLoRA), and combines (i) FedProx-style proximal regularization on low-rank adapters to stabilize optimization under highly Non-IID smart-home data, and (ii) Differentially Private SGD (DP-SGD) applied directly to adapter gradients to provide user-level $(\epsilon, \delta)$-differential privacy against an honest-but-curious server. In a realistic IoT instruction-following benchmark with Dirichlet-skewed clients ($\alpha=0.3$), DP-Prox improves F1-score from 0.80 (Edge-FIT with FedAvg) to 0.82 while reducing convergence rounds from $T=50$ to $T \approx 40$. Under a strict privacy budget $\left(\epsilon=5, \delta=10^{-5}\right)$, DP-Prox maintains an F1-score of 0.77 with only $+15 \%-20 \%$ local overhead, and preserves the communication efficiency of QLoRA-based federated training. Qualitative analyses further show that DP-Prox yields more consistent instruction-following behavior across heterogeneous homes, especially on multi-device routines and privacy-sensitive queries. Together, these results position DP-Prox as a practical and regulation-ready blueprint for deploying small, edge-resident LLMs in smart-home IoT environments.- 5 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 0620-0626
DP-Prox: A Robust and Differentially Private Framework for Federated Instruction Tuning of Small LLMs on 8 GB Edge Devices
This paper introduces DP-Prox, an end-to-end framework for robust and differentially private federated instruction tuning of small Large Language Models (LLMs) on 8 GB edge devices. DP-Prox is instantiated using the 3.8 billionparameter Phi-3-mini model with 4-bit Quantized Low-Rank Adaptation (QLoRA), and combines (i) FedProx-style proximal regularization on low-rank adapters to stabilize optimization under highly Non-IID smart-home data, and (ii) Differentially Private SGD (DP-SGD) applied directly to adapter gradients to provide user-level $(\epsilon, \delta)$-differential privacy against an honest-but-curious server. In a realistic IoT instruction-following benchmark with Dirichlet-skewed clients ($\alpha=0.3$), DP-Prox improves F1-score from 0.80 (Edge-FIT with FedAvg) to 0.82 while reducing convergence rounds from $T=50$ to $T \approx 40$. Under a strict privacy budget $\left(\epsilon=5, \delta=10^{-5}\right)$, DP-Prox maintains an F1-score of 0.77 with only $+15 \%-20 \%$ local overhead, and preserves the communication efficiency of QLoRA-based federated training. Qualitative analyses further show that DP-Prox yields more consistent instruction-following behavior across heterogeneous homes, especially on multi-device routines and privacy-sensitive queries. Together, these results position DP-Prox as a practical and regulation-ready blueprint for deploying small, edge-resident LLMs in smart-home IoT environments.- 18 June 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 6 (3) , 462-478
Privacy-Preserving Context-Based Electric Vehicle Dispatching for Energy Scheduling in Microgrids: An Online Learning Approach
Electric Vehicles (EVs) are beginning to play a key role in the fast developing area of Internet of Things (IoT). Numerous results have shown the feasibility of vehicle-to-building (V2B) mode of charge/discharge, where EVs are considered as dynamically configurable dispersed energy storage units. When properly incorporated into the building energy system, EVs are able to provide ancillary services to the power grid during high demand periods or outage situations. The arising challenge is how to act intelligent behaviors in complex and changing microgrid environments. With the aim of minimizing the cost and maximizing satisfaction degree, this paper, unlike previous works, jointly considers the building energy need and the safety/willingness of EVs to find and dispatch the optimal vehicle to conduct auxiliary or supportive actions. To realize that, we propose an intelligent Privacy-preserving Context-based Online EV Dispatching System (PCOEDS), using a tree-based structure which supports the ever-increasing big metering datasets with context-awareness. Moreover, privacy preservation and security protection on both sides of the the energy transmission process are well guaranteed in our work. Theoretical results validate that our intelligent dispatching system achieves sublinear regret and differential privacy, which outperforms other online learning method when applied to a huge city-level dataset.
- 18 June 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 6 (3) , 462-478
Privacy-Preserving Context-Based Electric Vehicle Dispatching for Energy Scheduling in Microgrids: An Online Learning Approach
Electric Vehicles (EVs) are beginning to play a key role in the fast developing area of Internet of Things (IoT). Numerous results have shown the feasibility of vehicle-to-building (V2B) mode of charge/discharge, where EVs are considered as dynamically configurable dispersed energy storage units. When properly incorporated into the building energy system, EVs are able to provide ancillary services to the power grid during high demand periods or outage situations. The arising challenge is how to act intelligent behaviors in complex and changing microgrid environments. With the aim of minimizing the cost and maximizing satisfaction degree, this paper, unlike previous works, jointly considers the building energy need and the safety/willingness of EVs to find and dispatch the optimal vehicle to conduct auxiliary or supportive actions. To realize that, we propose an intelligent Privacy-preserving Context-based Online EV Dispatching System (PCOEDS), using a tree-based structure which supports the ever-increasing big metering datasets with context-awareness. Moreover, privacy preservation and security protection on both sides of the the energy transmission process are well guaranteed in our work. Theoretical results validate that our intelligent dispatching system achieves sublinear regret and differential privacy, which outperforms other online learning method when applied to a huge city-level dataset.