Haoyu Wang

Rongzhe Wei, Eleonora Kreačić, Haoyu Wang et al.

Privacy concerns have led to a surge in the creation of synthetic datasets, with diffusion models emerging as a promising avenue. Although prior studies have performed empirical evaluations on these models, there has been a gap in providing a mathematical characterization of their privacy-preserving capabilities. To address this, we present the pioneering theoretical exploration of the privacy preservation inherent in discrete diffusion models (DDMs) for discrete dataset generation. Focusing on per-instance differential privacy (pDP), our framework elucidates the potential privacy leakage for each data point in a given training dataset, offering insights into how the privacy loss of each point correlates with the dataset's distribution. Our bounds also show that training with $s$-sized data points leads to a surge in privacy leakage from $(ε, O(\frac{1}{s^2ε}))$-pDP to $(ε, O(\frac{1}{sε}))$-pDP of the DDM during the transition from the pure noise to the synthetic clean data phase, and a faster decay in diffusion coefficients amplifies the privacy guarantee. Finally, we empirically verify our theoretical findings on both synthetic and real-world datasets.

Feijie Wu, Xiaoze Liu, Haoyu Wang et al.

Reinforcement learning with human feedback (RLHF) fine-tunes a pretrained large language model (LLM) using user preference data, enabling it to generate content aligned with human preferences. However, due to privacy concerns, users may be reluctant to share sensitive preference data. To address this, we propose utilizing Federated Learning (FL) techniques, allowing large-scale preference collection from diverse real-world users without requiring them to transmit data to a central server. Our federated RLHF methods (i.e., FedBis and FedBiscuit) encode each client's preferences into binary selectors and aggregate them to capture common preferences. In particular, FedBiscuit overcomes key challenges, such as preference heterogeneity and reward hacking, through innovative solutions like grouping clients with similar preferences to reduce heterogeneity and using multiple binary selectors to enhance LLM output quality. To evaluate the performance of the proposed methods, we establish the first federated RLHF benchmark with a heterogeneous human preference dataset. Experimental results show that by integrating the LLM with aggregated client preferences, FedBis and FedBiscuit significantly enhance the professionalism and readability of the generated content.

Haoyu Wang, Jiazhao Wang, Demin Gao et al.

Cross-technology communication(CTC) enables seamless interactions between diverse wireless technologies. Most existing work is based on reversing the transmission path to identify the appropriate payload to generate the waveform that the target devices can recognize. However, this method suffers from many limitations, including dependency on specific technologies and the necessity for intricate algorithms to mitigate distortion. In this work, we present NNCTC, a Neural-Network-based Cross-Technology Communication framework inspired by the adaptability of trainable neural models in wireless communications. By converting signal processing components within the CTC pipeline into neural models, the NNCTC is designed for end-to-end training without requiring labeled data. This enables the NNCTC system to autonomously derive the optimal CTC payload, which significantly eases the development complexity and showcases the scalability potential for various CTC links. Particularly, we construct a CTC system from Wi-Fi to ZigBee. The NNCTC system outperforms the well-recognized WEBee and WIDE design in error performance, achieving an average packet reception rate(PRR) of 92.3% and an average symbol error rate(SER) as low as 1.3%.