SWRouter: Similarity-Contractive Window Routing for Multi-Turn Large Language Model Conversations
本文提出SWRouter，通过相似度收缩窗口路由解决多轮对话中大语言模型的信息丢失和混淆问题，提高路由质量。
0 citationsRead paper
Selected work
Representative Papers
Diff-DDoS: Realistic Cyber-Physical Attack Synthesis and Robust Detection for 5G-Enabled CPS Using Tabular Diffusion Models
为解决5G系统中DDoS检测器面对的数据稀缺和合成数据不真实的问题，提出Diff-DDoS框架，使用表格扩散模型生成逼真的攻击样本以增强检测器的鲁棒性。
0 citationsRead paper
Joint Power and Mobility Control
To address degraded connectivity in autonomous driving V2X networks caused by dynamic interference, this paper proposes a joint trajectory and transmit power optimization framework. Methodologically, it innovatively models the signal-to-interference-plus-noise ratio (SINR) as a differentiable sigmoid function to explicitly capture power-dependent effects and—uniquely—rigorously proves the concavity of the multi-node network utility maximization (NUM) problem, enabling distributed cooperative optimization. A joint algorithm is designed based on an SINR-based link reception model to co-optimize vehicle trajectories and transmission power. Experimental results demonstrate that, under interference-constrained conditions, the proposed method improves packet reception rate by over 40% compared to baseline schemes. Furthermore, symmetric node placement and balanced power allocation significantly enhance overall network connectivity.
0 citationsRead paper
Recent publications
Latest Papers
SWRouter: Similarity-Contractive Window Routing for Multi-Turn Large Language Model Conversations
本文提出SWRouter，通过相似度收缩窗口路由解决多轮对话中大语言模型的信息丢失和混淆问题，提高路由质量。
0 citationsRead paper
Diff-DDoS: Realistic Cyber-Physical Attack Synthesis and Robust Detection for 5G-Enabled CPS Using Tabular Diffusion Models
为解决5G系统中DDoS检测器面对的数据稀缺和合成数据不真实的问题，提出Diff-DDoS框架，使用表格扩散模型生成逼真的攻击样本以增强检测器的鲁棒性。
0 citationsRead paper
Joint Power and Mobility Control
To address degraded connectivity in autonomous driving V2X networks caused by dynamic interference, this paper proposes a joint trajectory and transmit power optimization framework. Methodologically, it innovatively models the signal-to-interference-plus-noise ratio (SINR) as a differentiable sigmoid function to explicitly capture power-dependent effects and—uniquely—rigorously proves the concavity of the multi-node network utility maximization (NUM) problem, enabling distributed cooperative optimization. A joint algorithm is designed based on an SINR-based link reception model to co-optimize vehicle trajectories and transmission power. Experimental results demonstrate that, under interference-constrained conditions, the proposed method improves packet reception rate by over 40% compared to baseline schemes. Furthermore, symmetric node placement and balanced power allocation significantly enhance overall network connectivity.
0 citationsRead paper