【摘要】 针对网络中频繁出现的谣言，利用高效的方法进行谣言检测具有重要意义。基于此，提出了微调BERT-WWM预训练模型结合Mamba神经网络架构，使用包含8051条数据的公开数据集对模型进行训练，最后进行模型评估与对比，结果显示使用新型的Mamba架构，在准确率上可以进一步提高0.33%,检测准确率达到95.37%,优于TextRCNN,TextRNN-Att等对照模型。更多还原

【Abstract】 Detecting rumors efficiently in online networks is of significant importance. To address this, we propose a fine-tuned BERT-WWM pre-trained model integrated with the Mamba neural network architecture. The model was trained on a public dataset containing 8,051 entries, followed by comprehensive evaluation and comparative analysis. Experimental results demonstrate that the novel Mamba architecture achieves a 0.33% improvement in accuracy, reaching 95.37% detection accuracy, outperforming baseline models such as TextRCNN and TextRNN-Att.更多还原