SmoothDectector: A Smoothed Dirichlet Multimodal Approach for Combating Fake News on Social Media
Bibliographic record
Abstract
The rapid dissemination of fake news in the digital era has become a pressing concern. The ease of generating and manipulating fake content, including images, text, audio, and videos, has significantly fueled the spread of misinformation on social media platforms. These platforms often lack rigorous editorial scrutiny, exacerbating this problem. Although recent studies have explored multimodal fake news detection to learn shared representations of textual and visual information, they often learn discrete latent representations, merely concatenations of multimodal features. Simple concatenation or summation operations hinder the dynamic interaction of multimodal features. Furthermore, most models rely on additional subtasks, such as reconstruction and event discrimination. The performance of these models depends heavily on subtasks, which can be mathematically complex and time-consuming. This reliance limits the ability of researchers to explore different modeling assumptions freely. This study introduces a novel approach that integrates a probabilistic algorithm with a deep neural network to effectively capture the uncertainties and diversities in the shared latent representation of multimodal data. Specifically, our model utilizes continuous latent representations by leveraging a smoothed Dirichlet distribution, facilitating the identification of shared hidden patterns across textual and visual modalities. In addition, our model demonstrates the powerful properties of generative models when integrated with neural network models. Our results underscore the potential of integrating a probabilistic algorithm with a deep neural network to address the challenges of fake news detection in a multimodal setting. To support further research and reproducibility, we made the code related to this work publicly accessible.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.009 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".