Deepfake detectors are commonly evaluated on real/fake classification alone, but real-world applications require explanations of the visual evidence behind each prediction. Joint vision-language models struggle with authenticity assessment, producing "real" narratives for manipulated images. To address this, we propose a decoupled predict-then-explain pipeline as our submission to the ACM Multimedia 2026 Explainable Deepfake Detection Challenge, with methods trained and evaluated on the XPlainVerse benchmark. A vision-foundation-model ensemble first classifies the image, then a Qwen3-VL-8B model generates a decision-conditioned explanation of the visual evidence. For fake images, a second Qwen3-VL-8B compresses that text into a short non-expert summary without using the image. We further apply metric-aligned Group Relative Policy Optimization (GRPO) to optimize both explainers against the challenge criteria. Under matched conditions, GRPO raises complex BERTScore-F1 from 0.686 to 0.712 and also improves grounding (entity and evidence F1); our submission ranked first on complex BERTScore-F1 and seventh overall.
Figures & tables
Figure 1. Visualization of our architecture. A vision detector classifies a manipulated image as fake; a GRPO-trained vision-language model then generates a textual explanation of the artifacts.
Split
Real
Fake
Total
Training
130,000
320,000
450,000
Validation
50,000
60,000
110,000
Test
90,000
110,000
200,000
Total
270,000
490,000
760,000
Table 1. XPlainVerse dataset splits. Test class counts reflect the hidden evaluation reference; labels are not included in the public test release. Train, validation, and test counts of real and fake images in XPlainVerse, totaling 760,000 images.
#
Team
Overall
Det. F1
Complex
Expl.
1
xz
0.842
0.942
0.706
0.741
7
Ours
0.791
0.860
0.712
0.722
32
Qwen3-VL-8B
0.643
0.634
0.665
0.651
Table 2. Selected final CodaBench scores ( Narang et al., 2026b ) . Complex is Complex BERTScore-F1; Expl. is Explanation Score. Challenge website: https://explainable-deepfake-detection.github.io/ . Leaderboard rows for the overall winner, our seventh-place submission, and the organizers' Qwen3-VL-8B baseline.
Adapter
BERTScore-F1
Entity F1
Evidence F1
Grounding
SFT-1655
0.686
0.362
0.259
0.310
GRPO-26400
0.712
0.387
0.296
0.341
Table 3. Matched SFT vs. GRPO complex-explanation ablation on 10,000 scored test images. SFT and GRPO adapters compared on complex BERTScore-F1, entity F1, evidence F1, and grounding.
Metric
Score
Detection macro F1
0.859709
Complex BERTScore-F1
0.712210
Simple overall
0.727451
Reference explanation score
0.719831
Entity F1
0.389700
Evidence F1
0.300414
Table 4. Post-competition grounding-aware scores on all 200,000 test images. Detection, explanation, grounding, and overall scores of our final submission on the full hidden test set.
Recent advances in generative image models have made many manipulated images highly realistic, raising the need for detectors that are not only accurate but also able to provide visual evidence for their decisions. In this paper, we present our solution to the Explainable Deepfake Detection Challenge [2] on the XPlainVerse dataset [1], where systems are required to predict whether an image is real or fake and generate both complex and simple explanations grounded in visible forensic cues. Our method follows a modular detection-and-explanation design. For the real/fake decision, we build a multi-backbone detector that combines several DINOv3 models with Mesorch manipulation-localization features, bringing together pretrained visual representations, DCT-aware cues, and multi-scale forensic information. To inject explanation evidence into the detector, we use a Grounding-DINO-based pseudo-mask generation pipeline that converts local artifact descriptions from training explanations into weak patch- level supervision for an Artifact Evidence Map. We further introduce a local patch-level contrastive objective that separates artifact and authenticity evidence in the detector feature space without requiring paired images or pixel-level manipulation masks. For language output, we use class-conditional Qwen3-VL models to generate complex explanations for fake and real predictions, followed by a GRPO-optimized text simplification model. The proposed methods were trained and evaluated on the challenge subset of XPlainVerse. On the full test split, our submission achieves 0.9349 detection accuracy, 0.5571 explanation score, and a 0.7456 final challenge score.
Artem Filippov, Aleksandr Gushchin, Kirill Koltsov +2
Lomonosov Moscow State University Moscow, Russia · MSU Institute for Artificial Intelligence Moscow, Russia
Deepfake detection is moving beyond binary classification decisions toward systems that can also explain the visual evidence supporting those decisions. This transition is important for real-world verification settings, where diverse users need to understand not only whether an image is manipulated, but also why it is considered suspicious. The Explainable Deepfake Detection Challenge at ACM Multimedia 2026 is designed to benchmark this joint capability. Built on XPlainVerse, a million-scale benchmark for explainable deepfake detection, the challenge evaluates methods on image classification and grounded natural-language explanation generation. Participants submit a real/fake label together with two explanations for each image: a detailed complex explanation for technical users and a concise simple explanation for general users. The evaluation combines classification metrics with semantic similarity, simplicity, and intent-aware grounding metrics that assess whether explanations identify the relevant manipulated entities and supporting visual evidence. The methodologies developed through the challenge will contribute to the development of next-generation explainable deepfake detectors. Evaluation script, baseline models, and accompanying code are available on https://github.com/Abhijeet8901/XPlainVerse-ACMChallenge.
Abhijeet Narang, Kartik Kuckreja, Shreya Ghosh +4
American University of Sharjah Sharjah, United Arab Emirates
As deepfake detection models increasingly produce natural language explanations, their reasoning often remains weakly grounded in visual artifacts, limiting reliability and user trust. Existing benchmarks mainly evaluate classification accuracy, overlooking whether explanations reflect the actual manipulations. This gap hinders progress toward deployable, explainable deepfake detection systems. To this end, we introduce XPlainVerse, a large-scale benchmark designed for joint deepfake detection and human-centered explanation. XPlainVerse comprises one million real and manipulated images, pairing authentic images from five established sources with forgeries generated by twelve off-the-shelf image editing and synthesis models. We further propose a multi-stage filtering pipeline, Edit-Check, to verify if manipulations satisfy their intended edits, enabling reliable reasoning supervision at scale. Beyond dataset scale, XPlainVerse provides two complementary explanation styles: technical explanations for expert analysis and simplified explanations optimized for non-technical users. To evaluate explanation quality beyond surface similarity, we propose novel metrics, EntityScore and EvidenceScore, that measure reasoning fidelity by checking whether explanations correctly identify manipulated entities and visual evidence. Human annotations on 2,000 explanation pairs validate our dataset quality against human judgment. We believe XPlainVerse will establish grounded explanation quality as a measurable dimension of deepfake detection and support scalable research on trustworthy, interpretable models.
Abhijeet Narang, Kartik Kuckreja, Shreya Ghosh +3
Monash University · MBZUAI · The University of Queensland