DAMMFND: Domain-Aware Multimodal Multi-view Fake News Detection

Authors: Weihai Lu, Yu Tong, Zhiqiu Ye

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments conducted on two real-world datasets demonstrate that the proposed model outperforms state-of-the-art baselines. ... We evaluated our model using two real-world datasets: Weibo (Wang et al. 2018) and Weibo-21 (Nan et al. 2021).
Researcher Affiliation Academia 1Peking University 2Wuhan University 3Anhui University EMAIL, EMAIL, EMAIL
Pseudocode No The paper describes the methodology using prose and mathematical equations but does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes Code https://github.com/luweihai/DAMMFND
Open Datasets Yes We evaluated our model using two real-world datasets: Weibo (Wang et al. 2018) and Weibo-21 (Nan et al. 2021).
Dataset Splits Yes The Weibo dataset contains 7,532 news articles (3,749 true, 3,783 fake) for training and 1,996 articles (996 true, 1,000 fake) for testing. ... Weibo21, a multi-domain dataset, comprises 9,127 articles (4,640 true, 4,487 fake), which we partitioned into training and test sets following established benchmark procedures.
Hardware Specification Yes All codes are developed using PyTorch (Paszke et al. 2019) and executed on an NVIDIA RTX 4090 graphics processing unit.
Software Dependencies No The paper mentions software like PyTorch, BERT, CLIP, and MAE but does not provide specific version numbers for any of these dependencies.
Experiment Setup Yes In the text data encoding section, we set 197 as the maximum length for text input and utilized pre-trained BERT (Devlin et al. 2018) and CLIP models for text encoding. For the visual data encoding, we first resized the input images to 224 224 pixels and employed pre-trained MAE (He et al. 2021) and CLIP models to encode the image data. ... In the DAMMFND framework s loss formula (Eq. 20), the parameter α is set to 0.25. We set the number of channels ktext, kimg and kmm to 18.