FLODA: Harnessing Vision-Language Models for Deepfake Assessment
Bibliographic record
Abstract
The proliferation of advanced generative AI models has ushered in a new era of image creation, where the distinction between real and synthetic content has become increasingly blurred. This phenomenon poses significant challenges to the integrity and trustworthiness of digital content and AI systems. Recently, Vision-Language Models (VLMs) have shown great potential in addressing these challenges due to their superior performance in vision-language tasks. In this paper, we propose FLODA (FLorence-2 Optimized for Deepfake Assessment), an advanced method designed to surpass existing deepfake detection models. FLODA leverages the VLM model, Florence-2, and extends it by integrating image captioning and authenticity assessment into a single end-to-end architecture. By utilizing caption information, FLODA enhances visual analysis with contextual details, streamlining both caption generation and deepfake detection. To ensure FLODA's efficacy, we performed an ablation study to identify the optimal model configuration and demonstrated its contributions to performance improvements. We also compared our best FLODA model against existing benchmarks through rigorous evaluation. Notably, FLODA demonstrates strong generalization, showcasing its robustness across diverse scenarios with an average accuracy of 97.14%. This research advances digital content integrity and AI trustworthiness, providing a promising direction for future VLM applications. Code and models can be found at https://github.com/byh711/FLODA.
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 | 0.008 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.000 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.005 | 0.003 |
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".