LLLimeng Lu

Limeng Lu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Limeng Lu has a total of 15 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • HateMediator: Fine-Tuning Large Language Models for Counter-Hate Speech via Multiturn Mediations

    by Xiao-Kun Wu, Lejun Ai, Yulong Li, Limeng Lu, Jixuan Xie, Yue Wang, Jiaxin Luo, Delu Zeng, Min Chen, Giancarlo Fortino

    The proliferation of hate speech on social media presents an escalating threat to both public discourse and individual mental well-being. Traditional strategies that prioritize detection and removal often neglect to engage directly with hate speakers or address the underlying causes of their hostility. This article proposes HateMediator, a dialogue-based intervention framework that fine-tunes large language models (LLMs) to generate persuasive, context-aware counter-hate speech. The framework emphasizes two core aspects: the generation of effective counter-hate responses and their evaluation through multiturn dialogues. Our fine-tuning approach integrates tutorial-based learning with critical token guidance, enabling LLMs to recognize and reproduce strategic rhetorical patterns observed in expert interventions. To support training and evaluation, we introduce the MedHate dataset, grounded in social science theory, comprising complete dialogue records from 85 real-world hate incidents (including 255 dialogues), expert-crafted counter-responses, and feedback from the original hate speakers. Experimental results show that HateMediator consistently outperforms baseline LLMs across multiple evaluation dimensions. This study advances both the technical frontier of hate speech intervention and the ethical deployment of LLMs in addressing complex social issues.

  • A Novel Agent-Based Approach for Dynamic Emotion Modeling in Social Networks

    by Xiaokun Wu, Limeng Lu, Mariagrazia Dotoli, Giancarlo Fortino, Min Chen

    In a socially tense environment with rising emotional pressure, understanding the spread patterns of group emotions—particularly negative emotions—is crucial for identifying social risks. Extensive research has explored emotion contagion, often using propagation models where node state transitions rely on preset probabilities. However, these methods introduce randomness, making them less reflective of real-world dynamics by failing to capture individual node behaviors and interactions in emotional networks. To address this, our study introduces a novel approach integrating text-based emotion recognition with propagation models, reconstructing emotion contagion at an individual level. This model enhances traditional nodes with multihop agents driven by text emotion analysis, where agents record and respond to neighbors’ emotional states. As a result, emotion spread becomes a deterministic process, with individualized infection rates reflecting node variability. We categorized nodes based on emotional states, creating corresponding agent types to form the dynamic agent-based emotion model (AEmo). Tests on real-world and scale-free networks show this method effectively predicts group negative emotion spread and provides insight into individual emotion evolution, validating the model’s effectiveness.


  • A Method of Self-Supervised Denoising and Classification for Sensor-Based Human Activity Recognition

    by Limeng Lu, Tao Deng

    Human activity recognition (HAR) is an important sub-field of pervasive computing and pattern recognition. While researchers have achieved remarkable results in feature extraction and classification for sensor-based HAR, they have encountered performance bottlenecks. Sensor signal denoising has emerged as an excellent approach to enhance the performance of sensor-based HAR architectures. In this study, we propose a novel self-supervised blind denoising method for sensor signals, which serves as a new module in the HAR task and significantly improves the overall system performance. Our method applies the idea of masking from image inpainting to temporal signal processing, and utilizes the neighboring signal to predict the signal at the center, leveraging independent noisy measurements and temporal relationships between real neighboring signals, without the need to understand the noise distribution. The denoising function is learned from samples, and the denoised signals are then fed into the classification model. Experimental results conducted on benchmark datasets WISDM, UCI-HAR, and PAMAP2 demonstrate the effectiveness of our denoising method. After denoising, the inner-class divergence of actions decreases, while the outer-class divergence increases. As a result, the recognition accuracy of the system significantly improves to 98.6%, 97.64%, and 97.12% on the respective datasets.


  • Open Access

    A Multichannel CNN-GRU Model for Human Activity Recognition

    by Limeng Lu, Chuanlin Zhang, Kai Cao, Tao Deng, Qianqian Yang

    Human activity recognition (HAR) is one of the important research areas in pervasive computing. Among HAR, sensor-based activity recognition refers to acquiring a high-level knowledge about human activities from readings of many low-level sensors. In recent years, although the existing methods of deep learning (DL) have been widely used for sensor-based HAR with some good performance, they still face such challenges as feature extraction and characterization, continuous action segmentation in dealing with time series problems. In this study, a multichannel fusion model is proposed with the idea of dividing. In this proposed architecture, a multichannel convolutional neural network (CNN) is used to enhance the ability to extract features at different scales, and then the fused features are fed into the gated recurrent unit (GRU) for feature labeling and enhanced feature representation, through the learning of temporal relationships. Finally, the multichannel CNN-GRU model is designed using global average pooling (GAP) to connect the feature maps with the final classification. The model performance was conducted on three benchmark datasets of WISDM, UCI-HAR, and PAMAP2 with the accuracy of 96.41%, 96.67%, and 96.25% respectively. The results show that the proposed model demonstrates better activity detection capability than some of the reported results.