The social interactions among crowds via Danmaku (a.k.a., bullet comments) on modern multimedia platforms can facilitate both viewpoint conflicts and consensus, providing fine-grained discriminative social signals that can benefit fake news detection. However, the inherent accumulation latency of Danmaku in real-world scenarios violates the real-time necessity of fake news detection, making the studies of Danmaku-related fake news detection underexplored. To break this violation, we simulate this temporal-aware user interactive process by proposing a novel temporal Generative danmaku framework, called Genda, which consists of: (1) a Danmaku Trigger for predicting the timing and intensity of user reactions; and (2) a Danmaku Generator for synthesizing corresponding semantic and emotional expressions, thereby mutually constructing a temporally aligned and human-like pseudo Danmaku streams. To make the generated Danmaku useful for identifying fake news videos, we further design a Danmaku-guided Temporal Multimodal fake news detection model - DM-FEND, which enables fine-grained multimodal interactions among video, audio, text, and Danmaku, enhancing dynamic modalities alignment and semantic noise inhibition. The experimental results demonstrate that DM-FEND consistently outperforms state-of-the-art baselines across both Chinese (FakeSV) and English (FakeTT) benchmarks. Further ablations validate the crucial role of temporal Danmaku modeling in enhancing robustness and discriminative capability. Finally, this study offers a bright and robust solution for multimodal fake news detection in modern social interactive fashions by bridging the temporal inconsistency between news and user behaviors.
Figures & tables
Figure 1. (a) Temporal distribution of Danmaku across real and fake videos, showing consistent early-stage concentration and distinct evolution patterns. (b) Semantic relatedness comparison between Danmaku and video content.
Figure 2. The workflow of our proposed approach. (A) Genda : a temporal Generative Danmaku module that models user interaction dynamics by performing segment-level video understanding, followed by a Danmaku Trigger to estimate reaction intensity and a Danmaku Generator to produce semantically and temporally aligned Danmaku. (B) DM-FEND : a Danmaku-guided temporal multimodal learning module that enhances unimodal representations via importance-aware and noise-aware masking, and captures fine-grained cross-modal interactions among text, video, audio, and Danmaku. (C) Classification block: temporally aggregated multimodal representations are fed into an MLP for final fake news prediction.
Models
FakeSV
FakeTT
Acc
Mac-F1
Prec
Rec
Acc
Mac-F1
Prec
Rec
Unimodal Baselines
Text(Bert)
81.36
81.29
81.36
81.33
76.54
75.00
74.70
77.24
Image(CLIP-vit)
73.99
73.97
75.65
75.35
68.90
68.49
71.02
73.73
Audio(wav2vec2)
75.52
75.13
76.86
76.37
70.54
69.77
71.75
74.00
Video(VideoMAEv2)
72.86
72.58
72.53
72.71
73.24
71.92
71.81
74.39
Table 1. This paper validates the performance of various baseline methods against our proposed method on the FakeSV and FakeTT datasets. Best results are highlighted in bold , and second-best results are highlighted with underlines . Paired t-tests were used to test the statistical significance of all baselines (p < 0.05), and significant differences are indicated by (*).
DA
EN
CM
FakeSV
FakeTT
Acc
Mis-F1
Acc
Mis-F1
×
×
×
81.49
80.64
77.58
76.95
×
×
83.12
82.87
79.79
79.81
×
×
83.72
82.47
80.39
80.26
×
×
82.95
84.83
80.16
83.42
×
84.32
83.56
81.18
80.20
Table 2. Ablation study of different components. DA: Temporal Danmaku Generation, EN: Danmaku-guided unimodal learning , CM: Danmaku-guided multimodal learning.
Figure 3. Visualized comparison between DM-FEND and its representative variants on the FakeSV dataset using Principal Component Analysis (PCA). Green represents real news sample points, and red represents fake news sample points.
Figure 4. Case study of a short video with the video clip-level pseudo Danmaku generated using our proposed Genda framework.
Generative content is increasingly entering the production and dissemination of news, transforming fake news from manually fabricated or simply manipulated material into complex forms in which native and generated content jointly participate. Existing multimodal fake news detection research primarily focuses on veracity assessment and rarely characterizes how generativity differences affect the reliability of evidence. In contrast, AIGC detection primarily determines whether content is generated or modified by generative models, but it does not by itself establish whether the underlying news event is true. To bridge the separation between these tasks in data and evaluation, we construct Weibo26, a multimodal fake news detection dataset for generative-content scenarios. On this basis, we propose the Generativity-Aware Hierarchical Reasoning (GAHR) framework, which combines global judgment with local correction so that generativity information participates in news-veracity reasoning. Experiments on multiple existing fake news detection benchmarks and Weibo26 show that GAHR achieves competitive veracity-detection performance while effectively identifying generative content.
Wenbin Shen, Guoxuan Qin, Guangxu Yao +3
Nanjing University of Science and Technology, Nanjing, China · University of Chinese Academy of Sciences, Beijing, China
In recent years, multimodal multidomain fake news detection has garnered increasing attention. Nevertheless, this direction presents two significant challenges: (1) Failure to Capture Cross-Instance Narrative Consistency: existing models usually evaluate each news in isolation, fail to capture cross-instance narrative consistency, and thus struggle to address the spread of cluster based fake news driven by social media; (2) Lack of Domain Specific Knowledge for Reasoning: conventional models, which rely solely on knowledge encoded in their parameters during training, struggle to generalize to new or data-scarce domains (e.g., emerging events or niche topics). To tackle these challenges, we introduce Retrieval-Augmented Multimodal Model for Fake News Detection (RAMM). First, RAMM employs a Multimodal Large Language Model (MLLM) as its backbone to capture cross-modal semantic information from news samples. Second, RAMM incorporates an Abstract Narrative Alignment Module. This component adaptively extracts abstract narrative consistency from diverse instances across distinct domains, aggregates relevant knowledge, and thereby enables the modeling of high-level narrative information. Finally, RAMM introduces a Semantic Representation Alignment Module, which aligns the model's decision-making paradigm with that of humans - specifically, it shifts the model's reasoning process from direct inference on multimodal features to an instance-based analogical reasoning process. Extensive experimental results on three public datasets validate the efficacy of our proposed approach. Our code is available at the following link: https://github.com/li-yiheng/RAMM
Yiheng Li, Weihai Lu, Hanyi Yu +1
University of International Business and Economics · Beijing, China · Peking University +4
The rapid advancement of generative AI raises concerns about the misuse of Multimodal LLMs (MLLMs) for large-scale disinformation campaigns on social media. Despite existing research on textual disinformation, a fundamental question remains unanswered: can MLLMs be exploited to fabricate realistic multimodal fake news, and can they reliably detect it? We introduce a multi-agent framework in which a story agent, an image agent, and a critic agent collaborate to produce fake social media posts that plausibly counter true news. We apply the framework to generate over 9,000 paired multimodal news posts across science, health, and entertainment domains, and benchmark 16 open- and closed-source MLLMs for automated detection. We find that most models fall substantially short of human-level accuracy and fail critically on identifying image authenticity. Our research provides a foundation for developing robust defenses against social media fake news. Code and data are available at https: //github.com/xiuzhenzhang/Multimodal.
Jiyao Yang, Yang Liu, Zhenyue Qin +2
Independent Researcher · Carnegie Mellon University · RMIT University +1